- Email: [email protected]
An optical fiber intrusion detection system for railway security
Sensors and Actuators A 253 (2017) 91–100
Contents lists available at ScienceDirect
Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna
An optical fiber intrusion detection system for railway security Angelo Catalano 1 , Francesco Antonio Bruno 1 , Carlo Galliano, Marco Pisco, Giovanni Vito Persiano, Antonello Cutolo, Andrea Cusano ∗ Optoelectronic Division – Engineering Dept., University of Sannio, c.so Garibaldi 107, 82100 Benevento, Italy
a r t i c l e
i n f o
Article history: Received 29 July 2016 Received in revised form 17 November 2016 Accepted 18 November 2016 Available online 21 November 2016 Keywords: Intrusion detection system Field installation Optical fiber sensors Fiber bragg grating Intruder walking test Railway
a b s t r a c t In this paper, we report the in-field demonstration of an intrusion detection system implemented using Fiber Bragg Grating (FBG) sensors. The system is conceived to protect the perimeter of an area from unauthorized accesses to railways assets. As a case study, we installed and tested the system in a real scenario, and several field trials were performed to validate the system’s ability to recognize any human intrusion within the protected area. The proposed sensing system is a powerful tool to detect intruders, especially in real scenarios where protection fences are not practicable. The proposed intrusion detection system represents a valuable solution to improve railway security. © 2016 Elsevier B.V. All rights reserved.
1. Introduction In the 21st century, railways are as vulnerable as airports and government buildings to vandalism and terrorism. Transport operators have growing concerns about the safety and security of railway facilities, workers, and passengers [1]. The peculiarities of railway environments make many conventional solutions to protect them impractical. A railway asset − such as a service area, a depot for trains or a railway station − cannot be closed with barriers because they would obstruct the passage of authorized staff or passengers. An additional difficulty arises from the protection of access to rail tracks that cannot be physically obstructed by commercial intrusion detection systems such as sensorized fences or taut wire sensors. Additionally, electronics-based technologies, such as microwave sensors, electric field sensors, ported coaxial cables, and infrared sensors, suffer from electromagnetic interferences associated with train transit. Over the past few years, to meet the growing demand for improved security, worldwide research on intrusion detection sensing systems has grown significantly [2]. Commercially available unattended ground sensor (UGS) systems make use of several sensing modalities (e.g., acoustic, seismic,
∗ Corresponding author. E-mail address: [email protected] (A. Cusano). 1 These authors contributed equally to this work. http://dx.doi.org/10.1016/j.sna.2016.11.026 0924-4247/© 2016 Elsevier B.V. All rights reserved.
passive infrared, magnetic, electrostatic, and video). The efficacy of UGS systems is often limited by high false alarm rates, and attempts to reduce false alarms could decrease the probability of detection [3,4]. As the false alarms are typically associated with the typology of the transduction principle, sensing solutions based on multiple technologies for intrusion detection offer benefits over a single technology approach in terms of their robustness and reliability [5,6]. Most modern intrusion detection systems employ multiple intrusion sensors to maximize their trustworthiness [7] because the usage of multiple technologies for intrusion detection reduces the false alarm rate while simultaneously improving the probability of detection [8–11]. A number of researchers have demonstrated that multiple sensing technologies offer information redundancy, reliability and complementarity. Lester et al. [12] show that dual technology sensors (i.e., passive infrared and microwave sensors) offer superior intruder detection and better performance for rejecting false/nuisance alarms than do single-technology systems. Similarly, to improve the performance of a system for personnel identification, Huang et al. [13] proposed a multi-sensor system based on ultrasonic, seismic and acoustic sensors to reduce the number of false alarms. To date, there are also several solutions involving visual content analysis [14–16]. Ngoc et al. [14,15] proposed a video and a multiple-sensors platform to detect and classify mobile objects. J.L. Castro et al. [16] proposed an intelligent surveillance system for the identification of dangerous intrusions based on information from video, audio and sensors.
92
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Despite the necessity of using several solutions based on different technologies, there is still minimal literature available regarding the use of optical fiber sensing technologies for intrusion detection [17]. Nonetheless, compared to conventional technologies, optical fiber sensing technology is potentially well suited for protecting large areas from unauthorized access. Additionally, optical fibers are ideal for use in railway areas due to the intrinsic advantages associated with their use, such as their immunity to electromagnetic interference, high sensitivity, compactness, remote sensing ability and stability in harsh environments. Allwood et al. [17] have recently reviewed the fiber optic techniques used in physical intrusion detection systems. They envisaged fundamentally three optical fiber technological approaches: interferometry, scattering and Fiber Bragg Gratings (FBGs) based detection. Interferometric approaches (ranging from classical Sagnac, Michelson or Mach-Zehnder interferometric schemes [17,18] up to speckle pattern configurations [19–21]) are able to offer high sensitivities but they are also extremely sensitive to external factors such as ambient conditions. An extensive signal processing is thus continuously required to remove external noise caused by the environment. Optical fiber scattering techniques have been also proposed for distributed sensing [22,23]. Such techniques demonstrated to be well suited for large distances, where the costs of the interrogation equipment are depreciated on the distance. FBG-based systems have already been used in security applications for monitoring entry points such as windows and doors [24] or in fence perimeter systems [25]. FBG sensors, indeed, offer high multiplexing capability, versatility, and simplicity of use, allowing the deployment of a cost-effective interrogation strategy [26,27]. Furthermore, FBG sensors have the potential to form a complete security system based entirely on fiber optic technology. Wide functionalities can be included in a single system based on the same core technology ranging from intrusion detection up to smart diagnostics in various railway applications such as switch monitoring, axle counting, weighing in motion, and wheel flat detection [28–30]. In our previous work [27,31,32], we also demonstrated using inlab trials that a technological solution based on FBG strain sensors is potentially able to detect human intrusions through a perimeter. In particular, we proposed a sensorized mat composed of a ribbed rubber mat with a low Young modulus integrated with FBGs [27]. The experimental analysis was focused on the characterization of the elastic response of the sensorized mat in the presence of applied static loads to evaluate the mat’s sensitivity curve, response times, and required number of sensors per unit area. Based on the results, we fabricated a mat as large as 2 m2 with eight FBGs, and we then verified in lab the functionality of the system in the presence of intruders walking on the mat. The small-scale system also allowed for the development of a simple signal processing strategy to neglect the effects of elastic hysteresis and of thermal drifts, affecting the response of the FBGs bonded to the mat. In this paper, expanding on the results of the previous in-lab activity, we report on the demonstration of an optical fiber intrusion detection system in a real railway scenario. Specifically, we installed and tested a sensorized rubber mat over a large square of approximately 20 m2 at the entrance (shown in Fig. 1) of the service area of Ente Autonomo Volturno (EAV) Railway in Naples (Italy). The sensing system is intended to be used in combination with a television close circuit (TVCC) system. As previously mentioned, the use of multiple sensing technologies offers information redundancy and complementarity. In this specific case, the TVCC system is used to cover both the access on the rail tracks and access to the
Fig. 1. Ponticelli Service Area of Ente Autonomo Volturno Railway Naples, Italy (a); service area in Ponticelli − Napoli (b).
Fig. 2. Schematics and pictures of the lower surface of the rubber mat.
“walkable” area, while the optical fiber intrusion detection system is used to protect the “walkable” area. In the following section, we briefly describe the architecture of the intrusion detection system. In section III, we report on the system installation and on the hardware/software setup. In the fourth section, we show the experimental results obtained during the onfield tests performed at the service area in Ponticelli. Finally, the paper ends with a conclusion section. 2. System architecture The sensing system is composed of a rubber mat with FBG strain sensors integrated on the lower surface of the mat. The sensorized mat is a scaled-up version of the prototype already demonstrated in reference [27]. Here, we briefly recall the structure of the sensorized mat and its principles of operation for the sake of completeness. The rubber mat has a smooth upper surface and a ribbed lower surface as shown in Fig. 2. The FBGs are bonded within the ribs of the lower surface to minimize the possibility of damage to or breakage of the optical fiber.
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
93
Fig. 3. System Architecture.
The principle of operation of the sensing system is quite simple. An intruder walking on the mat represents a load on its upper surface. The applied pressure due to the human footstep leads to a deformation of the mat according to its elastic properties, and an FBG bonded within the mat ribs experiences a strain, resulting in a Bragg wavelength shift. By deploying a net of equally spaced FBGs under the mat, the detection of a human walking on the mat can be simply achieved by monitoring the respective FBG wavelengths. The total number and the interspace among the sensors installed on the mat have been chosen based on the results of tests performed in the laboratory on a prototype mat [27]. In our previous work [27], we estimated the size of the influence zone of each sensor. Using this information, we determined that four FBGs per square meter were required to detect a walking human without any dark zones. Starting from this sensing configuration and by taking into account the mentioned design rules, we implemented a sensing mat solution that is suitable for protecting access to the Ponticelli service area shown in Fig. 1. The entry point features a walkable opening that is approximately 4 m wide. To control access by non-authorized persons, we chose to cover an area of approximately 20 m2 , creating a sensorized rubber mat of approximately 4 × 5 m2 . The extended area prevents the intruder from jumping over the mat and allows for identification of the intruder’s walking direction. The rubber mat used to protect this specific area is composed of three rectangular sections of approximately 1 × 5 m2 and one trapezoidal section with a larger base of 0.750 m, smaller base of 0.250 m and length of 5 m; the thickness is 3 mm. This division of the mat into smaller pieces allows for easier transportation, easier installation and more practical maintenance operations. To sensorize the entire rubber mat, 75 FBGs have been placed on the lower surface of the mat. Therefore, the entire mat area is covered in terms of its detection capability by at least one FBG. We selected the Micron Optics SM125-500 (Micron Optics, Inc., Atlanta, GA, USA) as the interrogation unit. It offers a scan frequency of 2 Hz in a wavelength range of 1510–1590 nm. Additionally, it is equipped with four independent channels, and its wavelength stability and accuracy are 1 pm [33]. Adopting the classical wavelength division multiplexing strategy, we located up to 20 FBGs on each channel by guaranteeing a spectral separation of approximately 4 nm. Specifically, we bonded twenty FBGs on the lower surface of each rectangular section of the mat, while on the trapezoidal section we bonded fifteen FBGs as schematically shown in Fig. 3. The FBG sensors for each section are not arranged in a single array configuration because we avoided having the fiber go across the ribs of the mat. Therefore, on each section of the mat, we real-
ized two arrays of 10 FBGs and carefully positioned them along the ribs of the lower surface of the mat. As schematically shown in Fig. 3, the eight arrays of FBG sensors installed on the mat are thus connected to eight optical fibers enclosed in an armored cable 25 m long. The armored cable brings the fiber bundle in a safe control room positioned 20 m from the intrusion detection mat. The armored cable is connected to the inputs FC/APC connectors of an optical coupler box, containing four fiber optic couplers (2 × 1 biconic module, single mode, with a wavelength of 1550 nm and a bandwidth of ± 40 nm) to allow the correct FBG multiplexing and fibers connection to the interrogation unit. The interrogator is connected to the Ethernet port of a personal computer. Finally, a management software package (SW) has been developed and installed on the personal computer to allow for “real-time” monitoring of intrusion events. The SW reads the FBG wavelengths, which are continuously detected by the interrogation system, and generates alarms associated with intrusion events by analyzing the FBG wavelength shifts. The SW implements a simple signal conditioning strategy (already investigated in our previous work23 by means of in-lab activities). Basically, we found that the Bragg wavelength shift cannot be considered to be a reliable observable signal for intrusion events. Even if the applied pressure due to the intruders steps is directly related to the Bragg wavelength shift, the elastic hysteresis of the mat as well as the daily thermal cycles (drifts) affect the Bragg wavelength in a “slow” temporal regime [27]. Consequently, we used the derivative of the Bragg wavelength versus time as the intruder activity indicator. Then, a simple comparison with respect to a predetermined threshold allows us to establish when the intrusion event occurs and, at the same time, the slow variations due to thermal drifts or to a slow elastic relaxation are easily rejected. In Fig. 4, a screenshot of the Graphical User Interface (GUI) of the software is displayed. This interface schematically displays 75 square-shaped boxes labeled with a number ranging from 1 to 75. These labels exactly identify the FBG sensors physically installed on the mat. Specifically, the square boxes with labels ranging from 1 to 60 correspond to the FBG sensors installed on the rectangular sections of the mat, whereas the square boxes with labels 61 to 75 correspond to the sensors physically placed on the trapezoidal section of the mat. According to the desired behavior, when nobody is walking on the mat, the square-shaped boxes must be green; otherwise, they should change their color to indicate that an intrusion has occurred. In principle, the intrusion events can be classified with different alarm levels (corresponding to different colors) by using multiple thresholds to classify the intrusion events. Information about
94
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Fig. 4. GUI of intrusion detection system.
Fig. 6. A sensorized mat rolled on a cylindrical support (a). Unrolling procedure (b). Single section of mat (c). Mat laid out (d).
Fig. 5. Comparison of the test site before (a) and after (b) adjustments.
detected intrusions is sent to a remote server to allow for management of the alarm (for example by alerting a remote operator and activating the TVCC system). Furthermore, the software implements auto-diagnostic features and provides the ability to record and save all of the data and alarms generated. 3. Fabrication and on-field installation In this section, we describe the on-field installation procedure and system setup. 3.1. Preliminary activities to the installation As shown in Fig. 5a, the test site was initially characterized by highly irregular ground and by the presence of debris, including small rocks and vegetation. To safeguard the integrity of the optical fiber arrays, it is necessary to lay the mat on a smooth surface, devoid of roughness or hazards that could result in injury to or breakage of the optical fiber. Therefore, we produced a concrete surface with building materials (a mixture of concrete and lime), devoid of grass and debris, to create a smooth surface as shown in Fig. 5b. 3.2. Installation procedure Both the rubber mat and the FBGs are commercially available. The FBGs are already produced in an array (spaced 50 cm from each other). The bonding of the FBGs to the mat is performed (in lab) in a few steps, briefly described here: 1. The surface of the mat is cleaned to remove any impurities, such as dust, that may affect the FBG bonding.
2. The FBG arrays are placed on the ribbed surface of the mat in the positions where the FBGs must be bonded. 3. The FBG is tensioned and kept strained (the induced wavelength shift is approximately 300 pm) in the assigned position. 4. Each FBG sensor is bonded with a glue layer that uniformly covers the FBG sensor. 5. We wait for approximately 120 seconds (with the pre-strain applied) to allow for crystallization of the glue. 6. We repeat steps 3, 4, and 5 until all FBGs are installed on the mat. As shown in Fig. 6a, each section was wound around a cylindrical support for transportation to the installation site. When each section of the mat was in position on the cemented surface (Fig. 6b), it was unrolled, as shown in Fig. 6c. This procedure was repeated for all four sections of the mat until the entire mat was correctly laid out (Fig. 6d). A plastic duct that allows for the passage of the arrays to a collecting box and protects the fiber was installed on one side of the cemented surface. In the box, the eight fiber arrays are connected to an armored optical fiber cable. A second plastic duct that contains the previously mentioned armored fiber branches off from this box and leads to the utility room. Before locking the duct, it was filled with polyurethane foam to ensure maximum protection for the fiber. Then, the perimeter of each section was fixed to the ground with a two-component resin to waterproof the underside of the mat. In addition, a metal skeleton was used to fix all sections of the mat to the cement to avoid the possibility of accidental lifting. Fig. 7 shows the intrusion detection mat installed in the field. The spectra acquired by the interrogation system after the installation was completed are shown in Fig. 8. As the graph shows, each FBG is characterized by a narrow reflection peak with an optimal amplitude and a baseline at low noise levels, confirming the successful installation of the sensing mat. 3.3. System setup After installation, the sensing system is ready to be used. The tracking of the Bragg wavelength shifts is performed by the interrogation system. The procedure is automatic and can eventually be
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Fig. 7. Intrusion detection system. Inset: Sensorized Mat and TCVV system. Inset: Sensorized Mat and TCVV.
Fig. 8. FBG sensor spectra.
improved by manually setting the power threshold for Bragg peak detection as well as the peak wavelength range. The basic threshold that is useful for intrusion event detection must be manually set to establish the sensitivity of the intruder sensing system. We determined this basic threshold by preliminarily analyzing the system noise floor when the mat was not trampled by intruders or subject to any pressure. The measurement was performed over 60 min of continuous acquisition for all FBG sensors. The standard deviation of the absolute value of the derivative of the Bragg wavelength shift versus time ranges from 0.6 pm/s to 2.2 pm/s for the 75 installed FBGs. The maximum value of the noise floor was measured to be approximately 15 pm/s. Therefore, we used 20 pm/s as the threshold to enable the alarm generation. It is worth noting that, by assuming a FBG thermal sensitivity of 10pm/◦ C, by using such a threshold, only a thermal change of 2 ◦ C/s is able to generate an alarm. 4. Experimental tests and results We performed several tests on the installed intrusion detection system to verify its ability to detect a human walking on the mat. In the following section, we report a selection of the results that are representative of the performed field tests.In the first reported test, a man walks in straight line along the sensorized mat, as shown in Fig. 9a. Fig. 9b shows the absolute values of the derivative of the Bragg wavelength shifts of the twenty FBGs installed in Section 3 of the sensorized mat during two successive walks of an intruder crossing Section 3 of the mat. The subplots are located in the same relative positions of the respective sensors to simplify the under-
95
standing of the data. In each subplot, we also indicated the 20 pm/s threshold for reference. During the walk, the sensors close to the man’s steps respond to the applied pressure by exhibiting a peak in the derivative of the Bragg wavelength shift. The sequence of the peaks clearly reveals the walking direction, and a velocity of approximately 1.1 m/s can also be inferred. The same behavior can also be observed in the second walk, despite discrepancies in the peak amplitudes attributed to inevitable differences in the position as well as in the pressure applied on the mat during the two walks. The pressure induced by a step can be sensed by more than one sensor simultaneously. Specifically, during the walk reported in Fig. 9b, the man took 7 steps to cross the mat, while for each walk approximately 18 peaks are discernable. By carefully correlating the video and FBG data, we ascertained that each step induces multiple peaks, whereas the highest peak corresponds to the closest sensor and the smaller peaks to the adjacent sensors. The y-axis scales are auto-normalized according to the y-scale reported in the respective subplots, so while the peaks appear similar in amplitude, they have significantly different values. The graphical interface provides an immediate view in real time of the status of the mat by coloring yellow the boxes where the derivative of Bragg wavelength shift versus time exceeds the threshold of 20 pm/s. In Fig. 9c, we show three representative frames from the GUI, which correspond to the times of 92.5, 95 and 97 s, respectively. The sensorized mat easily detects the intrusion of the man crossing section 3 of the mat, as indicated by the boxes turning yellow. The other sensors, which are not triggered by the man walking, are not alarmed. Indeed, by comparing the subplots in Fig. 9b and the corresponding frames in Fig. 9c, in some cases (i.e., FBG 42 in frame 1), the sensor is not in alarm despite the peaks visible in the subplot of Fig. 9b, but this circumstance occurs only when the sensors are far from the steps (so another sensor is detecting the intrusion). In all of the tests performed, we never observed a missed detection. Even when we attempted to elude the system. As an example, we report in Fig. 10 the data from a test in which a slender woman walks on the separation bar between Sections 3 and 4 of the mat. By increasing the contact surface between the intruder and the mat, in fact, the weight is distributed on a higher area, so that a lower pressure is exerted on the mat. However, since the mat sensitivity is not spatially uniform and it strongly increases close to the FBGs [27], the load should not be spread towards the FBG. For this reason, to attempt to elude the detection system, we used the separation bar that lies along the borders of the detection zone of the FBGs installed under the mat. As shown in Fig. 10b and Fig. 10c, in both the time plots and the three representative frames from the GUI, corresponding to times of 56, 59.5 and 61.5 s, respectively, the intruder is easily detected by more than one sensor. In principle, the random noise may exceed the threshold, resulting in a false alarm, while progressively increasing the threshold could result in a detrimental desensitization of the mat. An alternative effective strategy could exploit the peak amplitude as a reliable indicator of the confidence level of the presence of the intruder with respect to the threshold excess due to noise. In this regard, we show the amplitudes of several peaks versus their time extents in Fig. 11a. These results include tens of walking tests performed on the mat at the Ponticelli service area on different days with different intruders and, thus, changing walking directions, speeds, and gaits. Each relative maximum is considered to be a peak, and the peak time duration is evaluated as the full width at half maximum. For comparison and to elucidate the difference between peaks due to random noise and peaks due to the intruder footsteps, we display the results of the same analysis measured during 1 h of noise acquisition in Fig. 11b.
96
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Fig. 9. Photographs of the walking test on the mat (a), derivative of the Bragg wavelength shift versus time for the twenty FBGs fixed on section 3 of the mat (b), and frames from the GUI for the first walk of the intruder crossing section 3 of the mat (c).
Fig. 10. Photograph of the walking test on the mat (a), derivative of the Bragg wavelength shift versus time for the FBGs fixed under the mat near the separation bar between section 3 and 4 of the mat (b), and frames from the GUI for a test in which a slender woman walks on the separation bar between Sections 3 and 4 of the mat (c).
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
97
Fig. 11. Peak height versus peak width considering tens of walking tests performed on the mat (a), peak height versus peak width considering 1 h of noise acquisition (b).
As previously mentioned, the noise amplitude is below 20 pm/s. A green line delimiting the noise region is reported in both graphs for reference. During the walking tests, the amplitude typically settles around several tens of pm/s or a few hundred pm/s. Maximum values as high as 2 nm/s can also be observed. The analysis confirms what we already observed in the two particular cases reported in Figs. 9 and 10. Peaks due to footsteps can have amplitudes slightly above the noise floor or can significantly exceed the threshold and are well-distinguishable from the noise background. These results suggest that generating alarms by also indicating to what extent the threshold has been overridden can be useful. This strategy can be implemented by accompanying the alarm with the value of the peak amplitude or more simply by using multiple thresholds for alarm generation. For example, by exploiting the direct representation offered by the GUI, beyond the low threshold of 20 pm/s, we can set two additional thresholds of 50 pm/s and 250 pm/s to represent the intensity of the pressure experienced by the sensorized mat. Using these thresholds, in Fig. 12 we displayed the behavior of the GUI during a diagonal crossing of a man over the entire mat (Fig. 12a). The intrusion is again well detected, but the additional thresholds offer immediate information regarding the localization of the intruder on the mat, and allows for visualization of the intruder’s path. The colors of the boxes give a more precise indication of the location of the intruder by providing a higher confidence level on the correctness of the alarm. Please note that, during this
Fig. 12. Photograph of an intruder walking on the mat (a), frames of the GUI for the intrusion detection mat during the walking test (b–d).
98
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
test, the mat is partially exposed to sun and partially in the shadow, but the system response is not affected by this condition, confirming that the “slow” thermal variations, occurring daily, are not able to have an effect on the normal system operation. To confirm this system functionality, we also reported multiple experimental intrusion tests on the sensorized mat. In this specific case, the mat was simultaneously crossed by two intruders as shown in Fig. 13a. In particular, one of the two intruders crossed section 2 of the mat in a straight line, while the second intruder crossed with a parallel path on section 3. The frames, shown in Fig. 13b–d, illustrate the frames of the graphical interface in this scenario. The system is able to detect intrusion events and generate alarms in real time that are always coherent with the pressures applied in different areas of the mat. By taking advantage of the spatial distribution of the sensors and the alarms’ simultaneity, it is also possible to infer the presence ofP multiple intruders and to trace the path taken by the intruders. It is not always trivial or possible to retrieve the number of intruders and their directions from the behavior of the GUI (or from the information detected by the FBGs); however, the fusion of the data coming from all of the sensors can be efficiently used to obtain high-level information about the events triggering the alarms.
5. Discussion and conclusions
Fig. 13. Photograph of two intruders walking on Sections 2 and 3 of the mat a, behavior of the GUI for the intrusion detection mat during the walking test (b–d).
In this work, we have demonstrated the effective field implementation of an intrusion detection system based on FBG sensors fixed under a ribbed rubber mat to protect the train entry area of the Ponticelli service area of the Ente Autonomo Volturno Railway Naples, Italy. The installed sensing system represents a real-scale prototype of a perimeter protection system that is able to protect railway areas from unauthorized activities. We carefully described the modular architecture of the sensing system, the preparatory activities and the effective installation procedure. The implementation of an intrusion detection mat system with FBG sensors fixed on the mat offers a wide variety of features and benefits. FBG sensors do not require electrical power, they offer immunity to electromagnetic interferences, and their high multiplexing capability allows the deployment of a cost-effective interrogation strategy. Other basic components constituting the system architecture, such as the rubber mat, fiber optic coupler box and a personal computer, are low-cost components. As a possible drawback, we underline that, while the sensorization and laying-out of the mat require a few simple operations, the complete installation requires a preliminary preparation of the area dedicated to the installation to safeguard the integrity of the optical fiber arrays. In this specific installation, the intrusion detection system still allows rail track access because the mat cannot be placed on the rail tracks and positioning under the ballast (of any sensing system) is unpractical. Therefore, integration with another sensing system such as a video analysis system (or another optical fiber detection system) is required to allow the detection of a human walking across the rail tracks. The experimental results demonstrated that the intrusion detection mat system is always able to detect a human walking through the protected perimeter without exhibiting false negative alarms. Additionally, the sensing system is able to provide further information about the intrusion event. The presence of multiple sensors, such as those installed on the intrusion detection mat, enables the localization of the intruder, the recognition of the cross-
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
ing pattern and the determination of the number of intruders and their directions. Actually the assessment of probability of detection of a intrusion detection system requires a combined and statistically accurate analysis of nuisance alarms, because a trade-off between true and false alarms can be expected in relation to the set system sensitivity. Discrimination among different intrusion events requires the creation of a database in which each event typology is statistically characterized. The classification of events (including disturbance events) enables the recognition of the intruder features (within a certain probability) as well as the reduction of potential nuisance alarms related to meteorological phenomena or to other interfering events (i.e., animal intrusion). Basically, a complete definition of system performances would require an accurate statistical characterization of the events in different operating and whether conditions, forming a population of events statistically meaningful. However, such a study goes beyond the scope of this work and is currently the subject of an additional investigation. At this stage, we demonstrated the capability of the sensing system to successfully accomplish the detection of intrusion events in a real and “open” scenario. Additionally, the system demonstrated its ability to provide additional information (based on the features of the FBG responses as well as on the overall mat response) that can be used to infer high-level characteristics about intrusion events. Considering its features, the effectiveness of the installation and the detection performances, this intrusion sensing system represents a valuable solution for railway security applications. Acknowledgments The work is supported by the Italian Ministry of University and research under the national projects PON-SICURFER (Tecnologie innovative per la SICURezza della circolazione dei veicoli FERroviari) PON01 00142, CUP: B81C1100017000. We would like to thank Ansaldo STS and EAV srl for their support during the system installation. References [1] T.S.K. Chan, K.S.M. Chung, Applications and selections of intelligent surveillance system in railway industry, in: Railway Engineering – Challenges for Railway Transportation in Information Age, 2008. ICRE 2008, International Conference on,1,6, 25–28 March, 2008. [2] J.D. Williams, Advanced technologies for perimeter intrusion detection sensors, Security and Detection, 1995, in: European Convention, on,133,137, 16–18 May, 1995, http://dx.doi.org/10.1049/cp:19950484. [3] Xin Jin, A. Sarkar, S. Ray, T. Damarla, Target detection and classification using seismic and PIR sensors, Sensors Journal, IEEE 12 (2012) 1709, http://dx.doi. org/10.1109/JSEN.2011.217725 (no.6 1718, June). [4] G.L. Goodman, Detection and classification for unattended ground sensors, Information, Decision and Control, 1999. IDC 99. Proceedings.1999, 419,424, 1999 (2016), http://dx.doi.org/10.1109/IDC.1999.754194. [5] S.L. Rose-Pehrsson, C. Minor, K. Johnson, J. Owrutsky, S. Wales, D. Steinhurst, D. Gottuk, Multi-sensory, data fusion detection system for improved situational awareness, ECS Trans. 11 (16) (2008) 1–5, http://dx.doi.org/10. 1149/1.2890236. [6] B. Ying-Wen, X. Zi-Li, L. Zong-Han, Design and implementation of a home embedded surveillance system with ultra-low alert power, in: Consumer Electronics, IEEE Transactions, on, 57,1, 153,159, February, 2011, http://dx. doi.org/10.1109/TCE.2011.5735496. [7] A. Siraj, R.B. Vaughn, S.M. Bridges, Intrusion sensor data fusion in an intelligent intrusion detection system architecture,System Sciences, in: Proceedings of the 37th Annual Hawaii International Conference, on, 10, 5–8 Jan., 2004, p. 2004, http://dx.doi.org/10.1109/HICSS.2004.1265658. [8] R.C. Luo, Y. Chih-Chen, S. KuoLan, Multisensor fusion and integration: approaches, applications, and future research directions, Sensors Journal, IEEE 2 (2) (2002), http://dx.doi.org/10.1109/JSEN.2002.1000251 (107,119, Apr). [9] D.L. Hall, J. Llinas, An introduction to multisensor data fusion, Proc. IEEE 85 (1997) 6–23, http://dx.doi.org/10.1109/5.554205. [10] B.V. Dasarathy, Sensor fusion potential exploitation-innovative architectures and illustrative applications, Proc. IEEE 85 (1997) 24–38. [11] P.K. Varshney, Multisensor data fusion, Electron. Comm. Eng. J. 9 (6) (1997) 245–253.
99
[12] A.J. Lester, C.L. Smith, Analyses of performance of volumetric intrusion detection technologies, Security Technology, in: Proceedings. IEEE 33rd Annual 1999 International Carnahan Conference, on,101,111, 1999, p. 1999, http://dx.doi.org/10.1109/CCST.1999.797902. [13] H. Po-Sen, T. Damarla, M. Hasegawa-Johnson, Multi-sensory features for personnel detection at border crossings, Information Fusion (FUSION), in: 2011 Proceedings of the 14th International Conference, on,1,8, 5–8 July, 2011. [14] H.-B.B. Ngoc, F. Bremond, M. Thonnat, J.C. Faure, Shape recognition based on a video and multi-sensor system, in: Advanced Video and Signal Based Surveillance, 2005. AVSS 2005. IEEE Conference, on, 230,235, 15–16 Sept., 2005, http://dx.doi.org/10.1109/AVSS.2005.157727. ´ [15] D. Frejlichowski, K. Go´sciewska, P. Forczmanski, R. Hofman, Smart monitor— an intelligent security system for the protection of individuals and small properties with the possibility of home automation, Sensors 14 (2014) 9922–9948. [16] J.L. Castro, M. Delgado, J. Medina, M.D. Ruiz-Lozano, Intelligent surveillance system with integration of heterogeneous information for intrusion detection, Expert Syst. Appl. 38 (9 September) (2011) 11182–11192, http:// dx.doi.org/10.1016/j.eswa.2011.02.165 (ISSN 0957-4174). [17] G. Allwood, G. Wild, S. Hinckley, Optical fiber sensors in physical intrusion detection systems: a review, IEEE Sens. J. 16 (14) (2016) 5497–5509 (July 15). [18] Fiber Optic Sensors: An Introduction for Engineers and Scientists, Eric Udd, William B. Spillman, John Wiley & Sons, 11 Oct 2011. [19] W.B. Spillman, B.R. Kline, L.B. Maurice, P.L. Fuhr, Statistical-mode sensor for fiber optic vibration sensing uses, Appl. Opt. 28 (1989) 3166–3176. [20] A. Dhall, J.K. Chhabra, N.S. Aulakh, Intrusion detection system based on speckle pattern analysis, Exp. Tech. 29 (2005) 25–31. [21] L. Rodriguez-Cobo, M. Lomer, J. Lopez-Higuera, Fiber specklegram-multiplexed sensor, J. Lightwave Technol. 33 (2015) 2591–2597. [22] B. Liu, Q. Jin, Y. Bai, D. Wang, Y. Wang, Distributed fiber-optic sensors for vibration detection, Sensors 16 (8) (2016) 1164, http://dx.doi.org/10.3390/ s16081164. [23] X. Bao, L. Chen, Recent progress in distributed fiber optic sensors, Sensors 12 (2012) 8601–8639. [24] G. Allwood, S. Hinckley, G. Wild, Optical fiber Bragg grating based intrusion detection systems for homeland security, Proceedings IEEE Sensors Applications Symposium (2013) 66–70. [25] G. Allwood, G. Wild, S. Hinckley, Fibre optic acoustic sensing for intrusion detection systems, Proceedings Acoustics 2011: Breaking New Ground. Australian Acoustical Society (2011), 89–1–89–6. [26] M. Moccia, M. Pisco, A. Cutolo, V. Galdi, P. Bevilacqua, A. Cusano, Opto-acoustic behavior of coated fiber Bragg gratings, Opt. Express 19 (20) (2011) 18842–18860. [27] A. Catalano, F.A. Bruno, M. Pisco, A. Cutolo, A. Cusano, An intrusion detection system for the protection of railway assets using fiber bragg grating sensors, Sensors 14 (2014) 18268–18285. [28] Cusano, A. Laudati, A. Iele, A. Bruno, F.A. Parente, G. Giordano, M. Mazzino, N. Bocchetti, G, Smart railway in Italy, in: Proceedings of the 3th Asia Pacific Workshop on Structural Health Monitoring APWSHM, Tokyo, Japan, 30 November–2 December, 2010 (Sensors 2014, 14 18285). [29] C. Wei, C. Lai, S. Liu, W.H. Chung, T.K. Ho, H. Tam, S.L. Ho, A. McCusker, J. Kam, K.Y. Lee, A fiber bragg grating sensor system for train axle counting, Sens. J. IEEE 10 (2010) 1905–1912. [30] M.L. Filograno, P. Corredera, M. Rodríguez-Plaza, A. Andrés-Alguacil, M. González-Herráez, Wheel flat detection in high-speed railway systems using fiber bragg gratings, IEEE Sens. J. 13 (12) (2013) 4808–4816 (art. no. 6563101,). [31] A. Catalano, F.A. Bruno, M. Pisco, A. Cutolo, A. Cusano, Intrusion detection system for the protection of railway assets by using Fiber Bragg Grating sensors: a Case Study, Photonics Conference, 2014 Third Mediterranean, Trani (2014) 1–3, http://dx.doi.org/10.1109/MePhoCo.2014.6866483. [32] A. Catalano, F.A. Bruno, M. Pisco, A. Cutolo, A. Cusano, An intrusion detection system based on the optical fiber technology for the protection of railway assets, in: AISEM Annual Conference, 2015 XVIII, Trento, 2015, pp. 1–4, http:// dx.doi.org/10.1109/AISEM.2015.7066796. [33] Datasheet Micron Optics, Inc., SM125. Retrieved on 14/11/2016 from: http:// www.micronoptics.com/wp-content/uploads/2015/11/sm125 1502. pdf.
Biographies
AngeloCatalano was born in Benevento, Italy on 1985. He received the Electronic Engineering Master’s degree from University of Sannio (BN, Italy) with a final dissertation dealing with Development and testing of an intrusion detection system based on Fiber Optic Sensors for railway application. He collaborates with the Optoelectronic Group of University of Sannio. Since June 2014 he is a Ph.D student in Information Technologies for the Engineering at the University of Sannio. His current research activities essentially focuses on the fiber optic sensors. He is author and co-author of international publications, including international journals, and conferences.
100
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Francesco Antonio Bruno was born on January 30, 1982 in Foggia, Italy. He received the B.S. and M.S. degree in Telecommunications Engineering from University of Sannio, Italy, in July 2005 and October 2008 respectively. In 2012, he received his Ph.D. in Information Engineering from the University of Naples Parthenope, defending the thesis Optoelectronic Sensors for Smart Railways. His field of interest is in the area of optoelectronic sensors and the industrial application of optical fiber sensors. He is author and co-author of international publications, including international journals, and conferences.
Giovanni Vito Persiano received the degree in physics from the University “Federico II”, Napoli, Italy. In June 1991 he won a two-year scholarship by C.N.R. in Napoli. In the semester June-December 1993 he was supported by a research contract from SGS- Thomson, when he was admitted to the Ph.D. course in electronic engineering at the University of “Federico II” and got the Ph.D. degree in 1997. From 1996 to 2001 he had been a Research Assistant at the Faculty of Engineering of the University of Sannio, Benevento, where he has been, since April 2001, Associate professor of Electronic Engineering.
Carlo Galliano was born on April, 21, 1982 in Benevento, Italy. He received the B.S. in Computer Engineering from University of Sannio, Benevento, Italy, in December 2011. He is currently collaborating with Unisannio Optoelectronic Group to develop control interfaces for new optoelectronic system prototypes. In July 2012, he developed an automatic speech to text server to help deaf students to take lessons, based on open source speech to text engine Julius.
Antonello Cutolo was born in Napoli, Italy, on November 7, 1955. He received the doctoral degree in electronic engineering from the University Federico II of Napoli. He was with the Applied Mathematics-Physics Laboratory, Technical University of Denmark (1980–1981). From 1981 to 1983, he was with Adone Storage Ring of Frascati. From 1983 to 1986, he was with the High Energy Physics Laboratory, Stanford, CA. In 1986, he was a Professor of Quantum Electronics at the University of Naples, where he became a Professor of Optoelectronics in 1993. He is currently a Full Professor at the University of Sannio, Benevento. Since 2002 he has been leading the Optoelectronic Pole of the research consortium “Centro Regionale Information Communication Technology – CeRICT”. He has co-founded two high tech companies, Optosmart srl and OptoAdvance srl (spin off companies of the University of Sannio) Furthermore, he has founded the research Consortium TOP-IN, whose main aim is the exploitation of advanced optoelectronics technologies for industrial applications.
Marco Pisco was born in Naples, Italy in 1977. He received a Master’s Degree in Information and telecommunication engineering (110/110 magna cum laude) in 2003 from the University of Naples Federico II, Italy. In 2007, he received an international Ph.D. Degree in Information Engineering at the University of Sannio, Italy, jointly with the Faculty of Electrical Engineering (FER) at the University of Zagreb, defending the thesis Optoelectronic Devices based on 1-D and 2-D Photonic Bandgap Structures for Sensing and Communication Applications. He is currently a postdoctoral research fellow at the Optoelectronic Division − Engineering Department of the University of Sannio. He received the National Scientific Habilitation to be Associate Professor. His field of interest is in the area of optoelectronics and photonics. Specifically, he addresses research and development of optical fiber sensors and photonic bandgap-based materials and devices for sensing and communication applications. He is currently a member of the International Program Committee of the International Conference on Sensors and Electronic Instrumental Advances (SEIA) and of the Technical Program Committee of the European Workshop of Optical Fiber Sensors. He is the author and co-author of several publications, including books, international journals, national and international conferences and book chapters as well as a reviewer for IEEE, OSA and Elsevier journals.
Andrea Cusano received Master’s degree cum Laude in Electronic and Telecommunication Engineering and the Ph. D in Optoelectronics from University of Naples “Federico II”, Italy. He is currently Associate Professor at the Engineering Department of University of Sannio, where he and Prof. Cutolo co-founded the Optoelectronic Group since 2002. He is author of over 120 journal articles and 150 refereed conference communications related to the development of new fiber optic and photonic sensors for physical, chemical and biological sensing applications. He currently serves as Editor-in-Chief of the Journal of Optics and Laser Technology (Elsevier) and as Associate Editor for the Journal of Photonic Sensors (Springer) and he is a member of the technical committee of several international conferences.
Contents lists available at ScienceDirect
Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna
An optical fiber intrusion detection system for railway security Angelo Catalano 1 , Francesco Antonio Bruno 1 , Carlo Galliano, Marco Pisco, Giovanni Vito Persiano, Antonello Cutolo, Andrea Cusano ∗ Optoelectronic Division – Engineering Dept., University of Sannio, c.so Garibaldi 107, 82100 Benevento, Italy
a r t i c l e
i n f o
Article history: Received 29 July 2016 Received in revised form 17 November 2016 Accepted 18 November 2016 Available online 21 November 2016 Keywords: Intrusion detection system Field installation Optical fiber sensors Fiber bragg grating Intruder walking test Railway
a b s t r a c t In this paper, we report the in-field demonstration of an intrusion detection system implemented using Fiber Bragg Grating (FBG) sensors. The system is conceived to protect the perimeter of an area from unauthorized accesses to railways assets. As a case study, we installed and tested the system in a real scenario, and several field trials were performed to validate the system’s ability to recognize any human intrusion within the protected area. The proposed sensing system is a powerful tool to detect intruders, especially in real scenarios where protection fences are not practicable. The proposed intrusion detection system represents a valuable solution to improve railway security. © 2016 Elsevier B.V. All rights reserved.
1. Introduction In the 21st century, railways are as vulnerable as airports and government buildings to vandalism and terrorism. Transport operators have growing concerns about the safety and security of railway facilities, workers, and passengers [1]. The peculiarities of railway environments make many conventional solutions to protect them impractical. A railway asset − such as a service area, a depot for trains or a railway station − cannot be closed with barriers because they would obstruct the passage of authorized staff or passengers. An additional difficulty arises from the protection of access to rail tracks that cannot be physically obstructed by commercial intrusion detection systems such as sensorized fences or taut wire sensors. Additionally, electronics-based technologies, such as microwave sensors, electric field sensors, ported coaxial cables, and infrared sensors, suffer from electromagnetic interferences associated with train transit. Over the past few years, to meet the growing demand for improved security, worldwide research on intrusion detection sensing systems has grown significantly [2]. Commercially available unattended ground sensor (UGS) systems make use of several sensing modalities (e.g., acoustic, seismic,
∗ Corresponding author. E-mail address: [email protected] (A. Cusano). 1 These authors contributed equally to this work. http://dx.doi.org/10.1016/j.sna.2016.11.026 0924-4247/© 2016 Elsevier B.V. All rights reserved.
passive infrared, magnetic, electrostatic, and video). The efficacy of UGS systems is often limited by high false alarm rates, and attempts to reduce false alarms could decrease the probability of detection [3,4]. As the false alarms are typically associated with the typology of the transduction principle, sensing solutions based on multiple technologies for intrusion detection offer benefits over a single technology approach in terms of their robustness and reliability [5,6]. Most modern intrusion detection systems employ multiple intrusion sensors to maximize their trustworthiness [7] because the usage of multiple technologies for intrusion detection reduces the false alarm rate while simultaneously improving the probability of detection [8–11]. A number of researchers have demonstrated that multiple sensing technologies offer information redundancy, reliability and complementarity. Lester et al. [12] show that dual technology sensors (i.e., passive infrared and microwave sensors) offer superior intruder detection and better performance for rejecting false/nuisance alarms than do single-technology systems. Similarly, to improve the performance of a system for personnel identification, Huang et al. [13] proposed a multi-sensor system based on ultrasonic, seismic and acoustic sensors to reduce the number of false alarms. To date, there are also several solutions involving visual content analysis [14–16]. Ngoc et al. [14,15] proposed a video and a multiple-sensors platform to detect and classify mobile objects. J.L. Castro et al. [16] proposed an intelligent surveillance system for the identification of dangerous intrusions based on information from video, audio and sensors.
92
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Despite the necessity of using several solutions based on different technologies, there is still minimal literature available regarding the use of optical fiber sensing technologies for intrusion detection [17]. Nonetheless, compared to conventional technologies, optical fiber sensing technology is potentially well suited for protecting large areas from unauthorized access. Additionally, optical fibers are ideal for use in railway areas due to the intrinsic advantages associated with their use, such as their immunity to electromagnetic interference, high sensitivity, compactness, remote sensing ability and stability in harsh environments. Allwood et al. [17] have recently reviewed the fiber optic techniques used in physical intrusion detection systems. They envisaged fundamentally three optical fiber technological approaches: interferometry, scattering and Fiber Bragg Gratings (FBGs) based detection. Interferometric approaches (ranging from classical Sagnac, Michelson or Mach-Zehnder interferometric schemes [17,18] up to speckle pattern configurations [19–21]) are able to offer high sensitivities but they are also extremely sensitive to external factors such as ambient conditions. An extensive signal processing is thus continuously required to remove external noise caused by the environment. Optical fiber scattering techniques have been also proposed for distributed sensing [22,23]. Such techniques demonstrated to be well suited for large distances, where the costs of the interrogation equipment are depreciated on the distance. FBG-based systems have already been used in security applications for monitoring entry points such as windows and doors [24] or in fence perimeter systems [25]. FBG sensors, indeed, offer high multiplexing capability, versatility, and simplicity of use, allowing the deployment of a cost-effective interrogation strategy [26,27]. Furthermore, FBG sensors have the potential to form a complete security system based entirely on fiber optic technology. Wide functionalities can be included in a single system based on the same core technology ranging from intrusion detection up to smart diagnostics in various railway applications such as switch monitoring, axle counting, weighing in motion, and wheel flat detection [28–30]. In our previous work [27,31,32], we also demonstrated using inlab trials that a technological solution based on FBG strain sensors is potentially able to detect human intrusions through a perimeter. In particular, we proposed a sensorized mat composed of a ribbed rubber mat with a low Young modulus integrated with FBGs [27]. The experimental analysis was focused on the characterization of the elastic response of the sensorized mat in the presence of applied static loads to evaluate the mat’s sensitivity curve, response times, and required number of sensors per unit area. Based on the results, we fabricated a mat as large as 2 m2 with eight FBGs, and we then verified in lab the functionality of the system in the presence of intruders walking on the mat. The small-scale system also allowed for the development of a simple signal processing strategy to neglect the effects of elastic hysteresis and of thermal drifts, affecting the response of the FBGs bonded to the mat. In this paper, expanding on the results of the previous in-lab activity, we report on the demonstration of an optical fiber intrusion detection system in a real railway scenario. Specifically, we installed and tested a sensorized rubber mat over a large square of approximately 20 m2 at the entrance (shown in Fig. 1) of the service area of Ente Autonomo Volturno (EAV) Railway in Naples (Italy). The sensing system is intended to be used in combination with a television close circuit (TVCC) system. As previously mentioned, the use of multiple sensing technologies offers information redundancy and complementarity. In this specific case, the TVCC system is used to cover both the access on the rail tracks and access to the
Fig. 1. Ponticelli Service Area of Ente Autonomo Volturno Railway Naples, Italy (a); service area in Ponticelli − Napoli (b).
Fig. 2. Schematics and pictures of the lower surface of the rubber mat.
“walkable” area, while the optical fiber intrusion detection system is used to protect the “walkable” area. In the following section, we briefly describe the architecture of the intrusion detection system. In section III, we report on the system installation and on the hardware/software setup. In the fourth section, we show the experimental results obtained during the onfield tests performed at the service area in Ponticelli. Finally, the paper ends with a conclusion section. 2. System architecture The sensing system is composed of a rubber mat with FBG strain sensors integrated on the lower surface of the mat. The sensorized mat is a scaled-up version of the prototype already demonstrated in reference [27]. Here, we briefly recall the structure of the sensorized mat and its principles of operation for the sake of completeness. The rubber mat has a smooth upper surface and a ribbed lower surface as shown in Fig. 2. The FBGs are bonded within the ribs of the lower surface to minimize the possibility of damage to or breakage of the optical fiber.
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
93
Fig. 3. System Architecture.
The principle of operation of the sensing system is quite simple. An intruder walking on the mat represents a load on its upper surface. The applied pressure due to the human footstep leads to a deformation of the mat according to its elastic properties, and an FBG bonded within the mat ribs experiences a strain, resulting in a Bragg wavelength shift. By deploying a net of equally spaced FBGs under the mat, the detection of a human walking on the mat can be simply achieved by monitoring the respective FBG wavelengths. The total number and the interspace among the sensors installed on the mat have been chosen based on the results of tests performed in the laboratory on a prototype mat [27]. In our previous work [27], we estimated the size of the influence zone of each sensor. Using this information, we determined that four FBGs per square meter were required to detect a walking human without any dark zones. Starting from this sensing configuration and by taking into account the mentioned design rules, we implemented a sensing mat solution that is suitable for protecting access to the Ponticelli service area shown in Fig. 1. The entry point features a walkable opening that is approximately 4 m wide. To control access by non-authorized persons, we chose to cover an area of approximately 20 m2 , creating a sensorized rubber mat of approximately 4 × 5 m2 . The extended area prevents the intruder from jumping over the mat and allows for identification of the intruder’s walking direction. The rubber mat used to protect this specific area is composed of three rectangular sections of approximately 1 × 5 m2 and one trapezoidal section with a larger base of 0.750 m, smaller base of 0.250 m and length of 5 m; the thickness is 3 mm. This division of the mat into smaller pieces allows for easier transportation, easier installation and more practical maintenance operations. To sensorize the entire rubber mat, 75 FBGs have been placed on the lower surface of the mat. Therefore, the entire mat area is covered in terms of its detection capability by at least one FBG. We selected the Micron Optics SM125-500 (Micron Optics, Inc., Atlanta, GA, USA) as the interrogation unit. It offers a scan frequency of 2 Hz in a wavelength range of 1510–1590 nm. Additionally, it is equipped with four independent channels, and its wavelength stability and accuracy are 1 pm [33]. Adopting the classical wavelength division multiplexing strategy, we located up to 20 FBGs on each channel by guaranteeing a spectral separation of approximately 4 nm. Specifically, we bonded twenty FBGs on the lower surface of each rectangular section of the mat, while on the trapezoidal section we bonded fifteen FBGs as schematically shown in Fig. 3. The FBG sensors for each section are not arranged in a single array configuration because we avoided having the fiber go across the ribs of the mat. Therefore, on each section of the mat, we real-
ized two arrays of 10 FBGs and carefully positioned them along the ribs of the lower surface of the mat. As schematically shown in Fig. 3, the eight arrays of FBG sensors installed on the mat are thus connected to eight optical fibers enclosed in an armored cable 25 m long. The armored cable brings the fiber bundle in a safe control room positioned 20 m from the intrusion detection mat. The armored cable is connected to the inputs FC/APC connectors of an optical coupler box, containing four fiber optic couplers (2 × 1 biconic module, single mode, with a wavelength of 1550 nm and a bandwidth of ± 40 nm) to allow the correct FBG multiplexing and fibers connection to the interrogation unit. The interrogator is connected to the Ethernet port of a personal computer. Finally, a management software package (SW) has been developed and installed on the personal computer to allow for “real-time” monitoring of intrusion events. The SW reads the FBG wavelengths, which are continuously detected by the interrogation system, and generates alarms associated with intrusion events by analyzing the FBG wavelength shifts. The SW implements a simple signal conditioning strategy (already investigated in our previous work23 by means of in-lab activities). Basically, we found that the Bragg wavelength shift cannot be considered to be a reliable observable signal for intrusion events. Even if the applied pressure due to the intruders steps is directly related to the Bragg wavelength shift, the elastic hysteresis of the mat as well as the daily thermal cycles (drifts) affect the Bragg wavelength in a “slow” temporal regime [27]. Consequently, we used the derivative of the Bragg wavelength versus time as the intruder activity indicator. Then, a simple comparison with respect to a predetermined threshold allows us to establish when the intrusion event occurs and, at the same time, the slow variations due to thermal drifts or to a slow elastic relaxation are easily rejected. In Fig. 4, a screenshot of the Graphical User Interface (GUI) of the software is displayed. This interface schematically displays 75 square-shaped boxes labeled with a number ranging from 1 to 75. These labels exactly identify the FBG sensors physically installed on the mat. Specifically, the square boxes with labels ranging from 1 to 60 correspond to the FBG sensors installed on the rectangular sections of the mat, whereas the square boxes with labels 61 to 75 correspond to the sensors physically placed on the trapezoidal section of the mat. According to the desired behavior, when nobody is walking on the mat, the square-shaped boxes must be green; otherwise, they should change their color to indicate that an intrusion has occurred. In principle, the intrusion events can be classified with different alarm levels (corresponding to different colors) by using multiple thresholds to classify the intrusion events. Information about
94
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Fig. 4. GUI of intrusion detection system.
Fig. 6. A sensorized mat rolled on a cylindrical support (a). Unrolling procedure (b). Single section of mat (c). Mat laid out (d).
Fig. 5. Comparison of the test site before (a) and after (b) adjustments.
detected intrusions is sent to a remote server to allow for management of the alarm (for example by alerting a remote operator and activating the TVCC system). Furthermore, the software implements auto-diagnostic features and provides the ability to record and save all of the data and alarms generated. 3. Fabrication and on-field installation In this section, we describe the on-field installation procedure and system setup. 3.1. Preliminary activities to the installation As shown in Fig. 5a, the test site was initially characterized by highly irregular ground and by the presence of debris, including small rocks and vegetation. To safeguard the integrity of the optical fiber arrays, it is necessary to lay the mat on a smooth surface, devoid of roughness or hazards that could result in injury to or breakage of the optical fiber. Therefore, we produced a concrete surface with building materials (a mixture of concrete and lime), devoid of grass and debris, to create a smooth surface as shown in Fig. 5b. 3.2. Installation procedure Both the rubber mat and the FBGs are commercially available. The FBGs are already produced in an array (spaced 50 cm from each other). The bonding of the FBGs to the mat is performed (in lab) in a few steps, briefly described here: 1. The surface of the mat is cleaned to remove any impurities, such as dust, that may affect the FBG bonding.
2. The FBG arrays are placed on the ribbed surface of the mat in the positions where the FBGs must be bonded. 3. The FBG is tensioned and kept strained (the induced wavelength shift is approximately 300 pm) in the assigned position. 4. Each FBG sensor is bonded with a glue layer that uniformly covers the FBG sensor. 5. We wait for approximately 120 seconds (with the pre-strain applied) to allow for crystallization of the glue. 6. We repeat steps 3, 4, and 5 until all FBGs are installed on the mat. As shown in Fig. 6a, each section was wound around a cylindrical support for transportation to the installation site. When each section of the mat was in position on the cemented surface (Fig. 6b), it was unrolled, as shown in Fig. 6c. This procedure was repeated for all four sections of the mat until the entire mat was correctly laid out (Fig. 6d). A plastic duct that allows for the passage of the arrays to a collecting box and protects the fiber was installed on one side of the cemented surface. In the box, the eight fiber arrays are connected to an armored optical fiber cable. A second plastic duct that contains the previously mentioned armored fiber branches off from this box and leads to the utility room. Before locking the duct, it was filled with polyurethane foam to ensure maximum protection for the fiber. Then, the perimeter of each section was fixed to the ground with a two-component resin to waterproof the underside of the mat. In addition, a metal skeleton was used to fix all sections of the mat to the cement to avoid the possibility of accidental lifting. Fig. 7 shows the intrusion detection mat installed in the field. The spectra acquired by the interrogation system after the installation was completed are shown in Fig. 8. As the graph shows, each FBG is characterized by a narrow reflection peak with an optimal amplitude and a baseline at low noise levels, confirming the successful installation of the sensing mat. 3.3. System setup After installation, the sensing system is ready to be used. The tracking of the Bragg wavelength shifts is performed by the interrogation system. The procedure is automatic and can eventually be
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Fig. 7. Intrusion detection system. Inset: Sensorized Mat and TCVV system. Inset: Sensorized Mat and TCVV.
Fig. 8. FBG sensor spectra.
improved by manually setting the power threshold for Bragg peak detection as well as the peak wavelength range. The basic threshold that is useful for intrusion event detection must be manually set to establish the sensitivity of the intruder sensing system. We determined this basic threshold by preliminarily analyzing the system noise floor when the mat was not trampled by intruders or subject to any pressure. The measurement was performed over 60 min of continuous acquisition for all FBG sensors. The standard deviation of the absolute value of the derivative of the Bragg wavelength shift versus time ranges from 0.6 pm/s to 2.2 pm/s for the 75 installed FBGs. The maximum value of the noise floor was measured to be approximately 15 pm/s. Therefore, we used 20 pm/s as the threshold to enable the alarm generation. It is worth noting that, by assuming a FBG thermal sensitivity of 10pm/◦ C, by using such a threshold, only a thermal change of 2 ◦ C/s is able to generate an alarm. 4. Experimental tests and results We performed several tests on the installed intrusion detection system to verify its ability to detect a human walking on the mat. In the following section, we report a selection of the results that are representative of the performed field tests.In the first reported test, a man walks in straight line along the sensorized mat, as shown in Fig. 9a. Fig. 9b shows the absolute values of the derivative of the Bragg wavelength shifts of the twenty FBGs installed in Section 3 of the sensorized mat during two successive walks of an intruder crossing Section 3 of the mat. The subplots are located in the same relative positions of the respective sensors to simplify the under-
95
standing of the data. In each subplot, we also indicated the 20 pm/s threshold for reference. During the walk, the sensors close to the man’s steps respond to the applied pressure by exhibiting a peak in the derivative of the Bragg wavelength shift. The sequence of the peaks clearly reveals the walking direction, and a velocity of approximately 1.1 m/s can also be inferred. The same behavior can also be observed in the second walk, despite discrepancies in the peak amplitudes attributed to inevitable differences in the position as well as in the pressure applied on the mat during the two walks. The pressure induced by a step can be sensed by more than one sensor simultaneously. Specifically, during the walk reported in Fig. 9b, the man took 7 steps to cross the mat, while for each walk approximately 18 peaks are discernable. By carefully correlating the video and FBG data, we ascertained that each step induces multiple peaks, whereas the highest peak corresponds to the closest sensor and the smaller peaks to the adjacent sensors. The y-axis scales are auto-normalized according to the y-scale reported in the respective subplots, so while the peaks appear similar in amplitude, they have significantly different values. The graphical interface provides an immediate view in real time of the status of the mat by coloring yellow the boxes where the derivative of Bragg wavelength shift versus time exceeds the threshold of 20 pm/s. In Fig. 9c, we show three representative frames from the GUI, which correspond to the times of 92.5, 95 and 97 s, respectively. The sensorized mat easily detects the intrusion of the man crossing section 3 of the mat, as indicated by the boxes turning yellow. The other sensors, which are not triggered by the man walking, are not alarmed. Indeed, by comparing the subplots in Fig. 9b and the corresponding frames in Fig. 9c, in some cases (i.e., FBG 42 in frame 1), the sensor is not in alarm despite the peaks visible in the subplot of Fig. 9b, but this circumstance occurs only when the sensors are far from the steps (so another sensor is detecting the intrusion). In all of the tests performed, we never observed a missed detection. Even when we attempted to elude the system. As an example, we report in Fig. 10 the data from a test in which a slender woman walks on the separation bar between Sections 3 and 4 of the mat. By increasing the contact surface between the intruder and the mat, in fact, the weight is distributed on a higher area, so that a lower pressure is exerted on the mat. However, since the mat sensitivity is not spatially uniform and it strongly increases close to the FBGs [27], the load should not be spread towards the FBG. For this reason, to attempt to elude the detection system, we used the separation bar that lies along the borders of the detection zone of the FBGs installed under the mat. As shown in Fig. 10b and Fig. 10c, in both the time plots and the three representative frames from the GUI, corresponding to times of 56, 59.5 and 61.5 s, respectively, the intruder is easily detected by more than one sensor. In principle, the random noise may exceed the threshold, resulting in a false alarm, while progressively increasing the threshold could result in a detrimental desensitization of the mat. An alternative effective strategy could exploit the peak amplitude as a reliable indicator of the confidence level of the presence of the intruder with respect to the threshold excess due to noise. In this regard, we show the amplitudes of several peaks versus their time extents in Fig. 11a. These results include tens of walking tests performed on the mat at the Ponticelli service area on different days with different intruders and, thus, changing walking directions, speeds, and gaits. Each relative maximum is considered to be a peak, and the peak time duration is evaluated as the full width at half maximum. For comparison and to elucidate the difference between peaks due to random noise and peaks due to the intruder footsteps, we display the results of the same analysis measured during 1 h of noise acquisition in Fig. 11b.
96
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Fig. 9. Photographs of the walking test on the mat (a), derivative of the Bragg wavelength shift versus time for the twenty FBGs fixed on section 3 of the mat (b), and frames from the GUI for the first walk of the intruder crossing section 3 of the mat (c).
Fig. 10. Photograph of the walking test on the mat (a), derivative of the Bragg wavelength shift versus time for the FBGs fixed under the mat near the separation bar between section 3 and 4 of the mat (b), and frames from the GUI for a test in which a slender woman walks on the separation bar between Sections 3 and 4 of the mat (c).
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
97
Fig. 11. Peak height versus peak width considering tens of walking tests performed on the mat (a), peak height versus peak width considering 1 h of noise acquisition (b).
As previously mentioned, the noise amplitude is below 20 pm/s. A green line delimiting the noise region is reported in both graphs for reference. During the walking tests, the amplitude typically settles around several tens of pm/s or a few hundred pm/s. Maximum values as high as 2 nm/s can also be observed. The analysis confirms what we already observed in the two particular cases reported in Figs. 9 and 10. Peaks due to footsteps can have amplitudes slightly above the noise floor or can significantly exceed the threshold and are well-distinguishable from the noise background. These results suggest that generating alarms by also indicating to what extent the threshold has been overridden can be useful. This strategy can be implemented by accompanying the alarm with the value of the peak amplitude or more simply by using multiple thresholds for alarm generation. For example, by exploiting the direct representation offered by the GUI, beyond the low threshold of 20 pm/s, we can set two additional thresholds of 50 pm/s and 250 pm/s to represent the intensity of the pressure experienced by the sensorized mat. Using these thresholds, in Fig. 12 we displayed the behavior of the GUI during a diagonal crossing of a man over the entire mat (Fig. 12a). The intrusion is again well detected, but the additional thresholds offer immediate information regarding the localization of the intruder on the mat, and allows for visualization of the intruder’s path. The colors of the boxes give a more precise indication of the location of the intruder by providing a higher confidence level on the correctness of the alarm. Please note that, during this
Fig. 12. Photograph of an intruder walking on the mat (a), frames of the GUI for the intrusion detection mat during the walking test (b–d).
98
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
test, the mat is partially exposed to sun and partially in the shadow, but the system response is not affected by this condition, confirming that the “slow” thermal variations, occurring daily, are not able to have an effect on the normal system operation. To confirm this system functionality, we also reported multiple experimental intrusion tests on the sensorized mat. In this specific case, the mat was simultaneously crossed by two intruders as shown in Fig. 13a. In particular, one of the two intruders crossed section 2 of the mat in a straight line, while the second intruder crossed with a parallel path on section 3. The frames, shown in Fig. 13b–d, illustrate the frames of the graphical interface in this scenario. The system is able to detect intrusion events and generate alarms in real time that are always coherent with the pressures applied in different areas of the mat. By taking advantage of the spatial distribution of the sensors and the alarms’ simultaneity, it is also possible to infer the presence ofP multiple intruders and to trace the path taken by the intruders. It is not always trivial or possible to retrieve the number of intruders and their directions from the behavior of the GUI (or from the information detected by the FBGs); however, the fusion of the data coming from all of the sensors can be efficiently used to obtain high-level information about the events triggering the alarms.
5. Discussion and conclusions
Fig. 13. Photograph of two intruders walking on Sections 2 and 3 of the mat a, behavior of the GUI for the intrusion detection mat during the walking test (b–d).
In this work, we have demonstrated the effective field implementation of an intrusion detection system based on FBG sensors fixed under a ribbed rubber mat to protect the train entry area of the Ponticelli service area of the Ente Autonomo Volturno Railway Naples, Italy. The installed sensing system represents a real-scale prototype of a perimeter protection system that is able to protect railway areas from unauthorized activities. We carefully described the modular architecture of the sensing system, the preparatory activities and the effective installation procedure. The implementation of an intrusion detection mat system with FBG sensors fixed on the mat offers a wide variety of features and benefits. FBG sensors do not require electrical power, they offer immunity to electromagnetic interferences, and their high multiplexing capability allows the deployment of a cost-effective interrogation strategy. Other basic components constituting the system architecture, such as the rubber mat, fiber optic coupler box and a personal computer, are low-cost components. As a possible drawback, we underline that, while the sensorization and laying-out of the mat require a few simple operations, the complete installation requires a preliminary preparation of the area dedicated to the installation to safeguard the integrity of the optical fiber arrays. In this specific installation, the intrusion detection system still allows rail track access because the mat cannot be placed on the rail tracks and positioning under the ballast (of any sensing system) is unpractical. Therefore, integration with another sensing system such as a video analysis system (or another optical fiber detection system) is required to allow the detection of a human walking across the rail tracks. The experimental results demonstrated that the intrusion detection mat system is always able to detect a human walking through the protected perimeter without exhibiting false negative alarms. Additionally, the sensing system is able to provide further information about the intrusion event. The presence of multiple sensors, such as those installed on the intrusion detection mat, enables the localization of the intruder, the recognition of the cross-
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
ing pattern and the determination of the number of intruders and their directions. Actually the assessment of probability of detection of a intrusion detection system requires a combined and statistically accurate analysis of nuisance alarms, because a trade-off between true and false alarms can be expected in relation to the set system sensitivity. Discrimination among different intrusion events requires the creation of a database in which each event typology is statistically characterized. The classification of events (including disturbance events) enables the recognition of the intruder features (within a certain probability) as well as the reduction of potential nuisance alarms related to meteorological phenomena or to other interfering events (i.e., animal intrusion). Basically, a complete definition of system performances would require an accurate statistical characterization of the events in different operating and whether conditions, forming a population of events statistically meaningful. However, such a study goes beyond the scope of this work and is currently the subject of an additional investigation. At this stage, we demonstrated the capability of the sensing system to successfully accomplish the detection of intrusion events in a real and “open” scenario. Additionally, the system demonstrated its ability to provide additional information (based on the features of the FBG responses as well as on the overall mat response) that can be used to infer high-level characteristics about intrusion events. Considering its features, the effectiveness of the installation and the detection performances, this intrusion sensing system represents a valuable solution for railway security applications. Acknowledgments The work is supported by the Italian Ministry of University and research under the national projects PON-SICURFER (Tecnologie innovative per la SICURezza della circolazione dei veicoli FERroviari) PON01 00142, CUP: B81C1100017000. We would like to thank Ansaldo STS and EAV srl for their support during the system installation. References [1] T.S.K. Chan, K.S.M. Chung, Applications and selections of intelligent surveillance system in railway industry, in: Railway Engineering – Challenges for Railway Transportation in Information Age, 2008. ICRE 2008, International Conference on,1,6, 25–28 March, 2008. [2] J.D. Williams, Advanced technologies for perimeter intrusion detection sensors, Security and Detection, 1995, in: European Convention, on,133,137, 16–18 May, 1995, http://dx.doi.org/10.1049/cp:19950484. [3] Xin Jin, A. Sarkar, S. Ray, T. Damarla, Target detection and classification using seismic and PIR sensors, Sensors Journal, IEEE 12 (2012) 1709, http://dx.doi. org/10.1109/JSEN.2011.217725 (no.6 1718, June). [4] G.L. Goodman, Detection and classification for unattended ground sensors, Information, Decision and Control, 1999. IDC 99. Proceedings.1999, 419,424, 1999 (2016), http://dx.doi.org/10.1109/IDC.1999.754194. [5] S.L. Rose-Pehrsson, C. Minor, K. Johnson, J. Owrutsky, S. Wales, D. Steinhurst, D. Gottuk, Multi-sensory, data fusion detection system for improved situational awareness, ECS Trans. 11 (16) (2008) 1–5, http://dx.doi.org/10. 1149/1.2890236. [6] B. Ying-Wen, X. Zi-Li, L. Zong-Han, Design and implementation of a home embedded surveillance system with ultra-low alert power, in: Consumer Electronics, IEEE Transactions, on, 57,1, 153,159, February, 2011, http://dx. doi.org/10.1109/TCE.2011.5735496. [7] A. Siraj, R.B. Vaughn, S.M. Bridges, Intrusion sensor data fusion in an intelligent intrusion detection system architecture,System Sciences, in: Proceedings of the 37th Annual Hawaii International Conference, on, 10, 5–8 Jan., 2004, p. 2004, http://dx.doi.org/10.1109/HICSS.2004.1265658. [8] R.C. Luo, Y. Chih-Chen, S. KuoLan, Multisensor fusion and integration: approaches, applications, and future research directions, Sensors Journal, IEEE 2 (2) (2002), http://dx.doi.org/10.1109/JSEN.2002.1000251 (107,119, Apr). [9] D.L. Hall, J. Llinas, An introduction to multisensor data fusion, Proc. IEEE 85 (1997) 6–23, http://dx.doi.org/10.1109/5.554205. [10] B.V. Dasarathy, Sensor fusion potential exploitation-innovative architectures and illustrative applications, Proc. IEEE 85 (1997) 24–38. [11] P.K. Varshney, Multisensor data fusion, Electron. Comm. Eng. J. 9 (6) (1997) 245–253.
99
[12] A.J. Lester, C.L. Smith, Analyses of performance of volumetric intrusion detection technologies, Security Technology, in: Proceedings. IEEE 33rd Annual 1999 International Carnahan Conference, on,101,111, 1999, p. 1999, http://dx.doi.org/10.1109/CCST.1999.797902. [13] H. Po-Sen, T. Damarla, M. Hasegawa-Johnson, Multi-sensory features for personnel detection at border crossings, Information Fusion (FUSION), in: 2011 Proceedings of the 14th International Conference, on,1,8, 5–8 July, 2011. [14] H.-B.B. Ngoc, F. Bremond, M. Thonnat, J.C. Faure, Shape recognition based on a video and multi-sensor system, in: Advanced Video and Signal Based Surveillance, 2005. AVSS 2005. IEEE Conference, on, 230,235, 15–16 Sept., 2005, http://dx.doi.org/10.1109/AVSS.2005.157727. ´ [15] D. Frejlichowski, K. Go´sciewska, P. Forczmanski, R. Hofman, Smart monitor— an intelligent security system for the protection of individuals and small properties with the possibility of home automation, Sensors 14 (2014) 9922–9948. [16] J.L. Castro, M. Delgado, J. Medina, M.D. Ruiz-Lozano, Intelligent surveillance system with integration of heterogeneous information for intrusion detection, Expert Syst. Appl. 38 (9 September) (2011) 11182–11192, http:// dx.doi.org/10.1016/j.eswa.2011.02.165 (ISSN 0957-4174). [17] G. Allwood, G. Wild, S. Hinckley, Optical fiber sensors in physical intrusion detection systems: a review, IEEE Sens. J. 16 (14) (2016) 5497–5509 (July 15). [18] Fiber Optic Sensors: An Introduction for Engineers and Scientists, Eric Udd, William B. Spillman, John Wiley & Sons, 11 Oct 2011. [19] W.B. Spillman, B.R. Kline, L.B. Maurice, P.L. Fuhr, Statistical-mode sensor for fiber optic vibration sensing uses, Appl. Opt. 28 (1989) 3166–3176. [20] A. Dhall, J.K. Chhabra, N.S. Aulakh, Intrusion detection system based on speckle pattern analysis, Exp. Tech. 29 (2005) 25–31. [21] L. Rodriguez-Cobo, M. Lomer, J. Lopez-Higuera, Fiber specklegram-multiplexed sensor, J. Lightwave Technol. 33 (2015) 2591–2597. [22] B. Liu, Q. Jin, Y. Bai, D. Wang, Y. Wang, Distributed fiber-optic sensors for vibration detection, Sensors 16 (8) (2016) 1164, http://dx.doi.org/10.3390/ s16081164. [23] X. Bao, L. Chen, Recent progress in distributed fiber optic sensors, Sensors 12 (2012) 8601–8639. [24] G. Allwood, S. Hinckley, G. Wild, Optical fiber Bragg grating based intrusion detection systems for homeland security, Proceedings IEEE Sensors Applications Symposium (2013) 66–70. [25] G. Allwood, G. Wild, S. Hinckley, Fibre optic acoustic sensing for intrusion detection systems, Proceedings Acoustics 2011: Breaking New Ground. Australian Acoustical Society (2011), 89–1–89–6. [26] M. Moccia, M. Pisco, A. Cutolo, V. Galdi, P. Bevilacqua, A. Cusano, Opto-acoustic behavior of coated fiber Bragg gratings, Opt. Express 19 (20) (2011) 18842–18860. [27] A. Catalano, F.A. Bruno, M. Pisco, A. Cutolo, A. Cusano, An intrusion detection system for the protection of railway assets using fiber bragg grating sensors, Sensors 14 (2014) 18268–18285. [28] Cusano, A. Laudati, A. Iele, A. Bruno, F.A. Parente, G. Giordano, M. Mazzino, N. Bocchetti, G, Smart railway in Italy, in: Proceedings of the 3th Asia Pacific Workshop on Structural Health Monitoring APWSHM, Tokyo, Japan, 30 November–2 December, 2010 (Sensors 2014, 14 18285). [29] C. Wei, C. Lai, S. Liu, W.H. Chung, T.K. Ho, H. Tam, S.L. Ho, A. McCusker, J. Kam, K.Y. Lee, A fiber bragg grating sensor system for train axle counting, Sens. J. IEEE 10 (2010) 1905–1912. [30] M.L. Filograno, P. Corredera, M. Rodríguez-Plaza, A. Andrés-Alguacil, M. González-Herráez, Wheel flat detection in high-speed railway systems using fiber bragg gratings, IEEE Sens. J. 13 (12) (2013) 4808–4816 (art. no. 6563101,). [31] A. Catalano, F.A. Bruno, M. Pisco, A. Cutolo, A. Cusano, Intrusion detection system for the protection of railway assets by using Fiber Bragg Grating sensors: a Case Study, Photonics Conference, 2014 Third Mediterranean, Trani (2014) 1–3, http://dx.doi.org/10.1109/MePhoCo.2014.6866483. [32] A. Catalano, F.A. Bruno, M. Pisco, A. Cutolo, A. Cusano, An intrusion detection system based on the optical fiber technology for the protection of railway assets, in: AISEM Annual Conference, 2015 XVIII, Trento, 2015, pp. 1–4, http:// dx.doi.org/10.1109/AISEM.2015.7066796. [33] Datasheet Micron Optics, Inc., SM125. Retrieved on 14/11/2016 from: http:// www.micronoptics.com/wp-content/uploads/2015/11/sm125 1502. pdf.
Biographies
AngeloCatalano was born in Benevento, Italy on 1985. He received the Electronic Engineering Master’s degree from University of Sannio (BN, Italy) with a final dissertation dealing with Development and testing of an intrusion detection system based on Fiber Optic Sensors for railway application. He collaborates with the Optoelectronic Group of University of Sannio. Since June 2014 he is a Ph.D student in Information Technologies for the Engineering at the University of Sannio. His current research activities essentially focuses on the fiber optic sensors. He is author and co-author of international publications, including international journals, and conferences.
100
A. Catalano et al. / Sensors and Actuators A 253 (2017) 91–100
Francesco Antonio Bruno was born on January 30, 1982 in Foggia, Italy. He received the B.S. and M.S. degree in Telecommunications Engineering from University of Sannio, Italy, in July 2005 and October 2008 respectively. In 2012, he received his Ph.D. in Information Engineering from the University of Naples Parthenope, defending the thesis Optoelectronic Sensors for Smart Railways. His field of interest is in the area of optoelectronic sensors and the industrial application of optical fiber sensors. He is author and co-author of international publications, including international journals, and conferences.
Giovanni Vito Persiano received the degree in physics from the University “Federico II”, Napoli, Italy. In June 1991 he won a two-year scholarship by C.N.R. in Napoli. In the semester June-December 1993 he was supported by a research contract from SGS- Thomson, when he was admitted to the Ph.D. course in electronic engineering at the University of “Federico II” and got the Ph.D. degree in 1997. From 1996 to 2001 he had been a Research Assistant at the Faculty of Engineering of the University of Sannio, Benevento, where he has been, since April 2001, Associate professor of Electronic Engineering.
Carlo Galliano was born on April, 21, 1982 in Benevento, Italy. He received the B.S. in Computer Engineering from University of Sannio, Benevento, Italy, in December 2011. He is currently collaborating with Unisannio Optoelectronic Group to develop control interfaces for new optoelectronic system prototypes. In July 2012, he developed an automatic speech to text server to help deaf students to take lessons, based on open source speech to text engine Julius.
Antonello Cutolo was born in Napoli, Italy, on November 7, 1955. He received the doctoral degree in electronic engineering from the University Federico II of Napoli. He was with the Applied Mathematics-Physics Laboratory, Technical University of Denmark (1980–1981). From 1981 to 1983, he was with Adone Storage Ring of Frascati. From 1983 to 1986, he was with the High Energy Physics Laboratory, Stanford, CA. In 1986, he was a Professor of Quantum Electronics at the University of Naples, where he became a Professor of Optoelectronics in 1993. He is currently a Full Professor at the University of Sannio, Benevento. Since 2002 he has been leading the Optoelectronic Pole of the research consortium “Centro Regionale Information Communication Technology – CeRICT”. He has co-founded two high tech companies, Optosmart srl and OptoAdvance srl (spin off companies of the University of Sannio) Furthermore, he has founded the research Consortium TOP-IN, whose main aim is the exploitation of advanced optoelectronics technologies for industrial applications.
Marco Pisco was born in Naples, Italy in 1977. He received a Master’s Degree in Information and telecommunication engineering (110/110 magna cum laude) in 2003 from the University of Naples Federico II, Italy. In 2007, he received an international Ph.D. Degree in Information Engineering at the University of Sannio, Italy, jointly with the Faculty of Electrical Engineering (FER) at the University of Zagreb, defending the thesis Optoelectronic Devices based on 1-D and 2-D Photonic Bandgap Structures for Sensing and Communication Applications. He is currently a postdoctoral research fellow at the Optoelectronic Division − Engineering Department of the University of Sannio. He received the National Scientific Habilitation to be Associate Professor. His field of interest is in the area of optoelectronics and photonics. Specifically, he addresses research and development of optical fiber sensors and photonic bandgap-based materials and devices for sensing and communication applications. He is currently a member of the International Program Committee of the International Conference on Sensors and Electronic Instrumental Advances (SEIA) and of the Technical Program Committee of the European Workshop of Optical Fiber Sensors. He is the author and co-author of several publications, including books, international journals, national and international conferences and book chapters as well as a reviewer for IEEE, OSA and Elsevier journals.
Andrea Cusano received Master’s degree cum Laude in Electronic and Telecommunication Engineering and the Ph. D in Optoelectronics from University of Naples “Federico II”, Italy. He is currently Associate Professor at the Engineering Department of University of Sannio, where he and Prof. Cutolo co-founded the Optoelectronic Group since 2002. He is author of over 120 journal articles and 150 refereed conference communications related to the development of new fiber optic and photonic sensors for physical, chemical and biological sensing applications. He currently serves as Editor-in-Chief of the Journal of Optics and Laser Technology (Elsevier) and as Associate Editor for the Journal of Photonic Sensors (Springer) and he is a member of the technical committee of several international conferences.