Remote ice detection equipment — RIDE

Cold Regions Science and Technology 72 (2012) 7–16

Contents lists available at SciVerse ScienceDirect

Cold Regions Science and Technology journal homepage: www.elsevier.com/locate/coldregions

Remote ice detection equipment — RIDE R.E. Gagnon ⁎, J. Groves, W. Pearson Institute for Ocean Technology, National Research Council of Canada, St. John's, NL, Canada, A1B 3T5

a r t i c l e

i n f o

Article history: Received 2 June 2011 Accepted 10 November 2011 Keywords: Remote ice detection Rime and glaze ice Icing layer thickness measurement

a b s t r a c t A new instrument for remote ice detection and thickness measurement is described. It incorporates two optically based technologies that give it capability to measure the thickness of clear or foggy layers of solid or liquid on surfaces. The device is capable of measurements on moving surfaces such as wind turbines, and aircraft propellers and rotors. Details of the prototype are presented along with thickness data acquired during two test programs, one study on clear ice layers where measurements were obtained from a distance of ~15 m and the other on foggy ice layers at a distance of approximately 13 m. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.

1. Introduction Atmospheric icing on aircraft and space vehicles poses hazards in terms of aerodynamic efficiency and damage to components due to impacts from shedding ice pieces. Icing is also a problem in cold marine environments leading to large buildups that affect vessel stability. Even moderate accumulations affect production and pose hazards for crews working on decks and structures. Relatively slight accumulations can affect communications and safety equipment. Wind turbines operating in cold regions can suffer loss of efficiency and damage due to atmospheric ice buildup. Power lines also experience damage due to atmospheric icing. Furthermore, commonly experienced road ice (e.g. black ice) is responsible for significant loss of life and equipment. Various strategies and methods have been developed to combat icing in its variety of forms. For example, in the case of aircraft wing icing, electric heaters, redirected aircraft engine heat and electroexpulsive methods have been utilized. Marine icing, on the other hand, typically accumulates in much greater amounts over large areas and the state of the art for its removal consists of baseball bats and mallets. Avoidance strategies, such as changing course for ships or changing altitude for aircraft, are also sometimes used at the first indication of icing before it becomes a more serious issue. In many cases there is a need to detect and measure the thickness of icing accumulations in a timely manner in order to intelligently initiate mitigating measures. A variety of icing sensing technologies have been introduced such as electromagnetic and electro-acoustic

⁎ Corresponding author. E-mail address: [email protected] (R.E. Gagnon).

devices, resonant beam sensors, infrared absorption sensors, etc. to address this need. Most systems involve mechanical contact of some sort while a few are non-contact. Non-contact devices are desirable since there are costs associated with altering the surface and structural components of an aircraft wing, for example, to deploy a sensor. Contact type devices may also introduce risk due to their electrical wires being in the vicinity of fuel tanks and communications equipment. Furthermore in several instances it is desirable to make measurements on moving components, such as turbine blades and propellers, or from a moving car, which makes contact type devices impractical or impossible. A number of optically based remote detection systems have been described in the literature. Most are based on absorptive and reflective properties of the ice and also its birefringence. For example Pernick (1999) describes a method that involves scanning a surface of an aircraft with laser light of two wavelengths and using the absorptive characteristics of water, deicing fluid and ice to determine the composition of a layer. Pernick mentions the possibility of using absorption to determine the thickness of layers, however the focus is mainly on detecting the presence of the materials and thickness measurement is not included in the patent claims. It also appears that Pernick's system would only work on clear ice, i.e. glaze ice, and would not be suitable for rime ice or frosted glaze ice. A similar method, based on absorption of infrared light, is described by Sinnar (1989) and in this instance the capability of measuring thickness appears in the patent claims. We speculate that the effectiveness of the technique could be reduced by the effect of a frosty surface on the ice or inclusions (air bubbles) in the ice that influence the intensity of radiation detected by the sensor. The same considerations would apply to the Road Surface State Sensor DSC111 device briefly described by Bridge (2008). The technology appears to work on the same principles as that of Pernick (1999) and Sinnar (1989) and it is stated that the technology can detect and measure thicknesses of layers of water, ice or snow.

0165-232X/$ – see front matter. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.coldregions.2011.11.004

8

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

Christian et al. (1993) describe an optically based system for detecting the presence of ice on aircraft wings during ground inspection based on the reflectivity of ice at differing wavelengths. Similarly, Gregoris et al. (2004) describe a system that utilizes spectral contrast associated with differing reflectivity of ice with varying wavelength to detect and measure ice thickness. We speculate that the method could suffer from inaccuracy in thickness measurement when the surface is at substantial angles to the sensor line of sight and when the ice has air bubble inclusions. Blackwood (1993) uses the birefringent properties of ice for detection where polarized light is directed at a surface with a layer and the returning light, after passing through the layer, is analyzed for altered polarization that would indicate the presence of a birefringent material, e.g. ice. A method that utilizes laser interferometry to measure ice thickness has been described by Gagnon (1997). Interference fringes are counted as the optical path of a laser beam through an ice layer is altered in a controlled manner by changing the angle of incidence or alternately by changing the wavelength of the laser in a continuous manner while the geometry remains fixed. The method would be limited to transparent materials with no air bubble inclusions, such as glaze ice. Here we describe a remote non-contact icing thickness sensing device capable of accurate measurements at considerable distances (>20 m). The device, known as RIDE (Remote Ice Detection Equipment), incorporates two optical methods to detect and measure ice and fluid layers. One method is for relatively clear layers, such as glaze ice and water layers, and the other technique is suited to foggy ice, such as rime ice, and translucent liquid layers.

Fig. 1. Illustration of the remote thickness measurement technique, Optical method 1, used for clear solid or liquid layers.

2. Technical description Above we noted that the system incorporates two optical methods for detecting and measuring layers, whether clear or foggy. The two methods are utilized within the same apparatus. We now describe the two methods, followed by a detailed description of the physical hardware and software used for the analysis.

where H is the thickness of the ice, D is the diameter of the dark circle and Θ is the critical angle for total internal reflection, given by:

2.1. Optical method 1

where n is the refractive index of the medium. For ice, the critical angle Θ is approximately equal to 50°. To measure the thickness of a layer of glaze ice one would only need to know the diameter of the circular boundary between the light and dark regions in order to calculate the thickness using expressions (1) and (2). Hence, the essential hardware components are a laser and a camera to capture the image from which a measurement is made. If the camera is viewing from a direction other than normal to the surface then the circle will appear as an ellipse, where the major axis is the diameter of the actual circle. Test programs utilizing this method are described below, including sample images. Essentially, an image analysis program fits an ellipse to the boundary between the dark elliptical region and its corresponding peripheral light region in the image. The software fits the ellipse by first locating the approximate center of the brightest spot in the image. Next the software computes the brightness gradient along a radius extending outward from the aforementioned center. The steepest gradient occurs at the boundary between the dark elliptical region and its corresponding peripheral light region. This operation is repeated for a set of equally spaced radii extending outward from the center of the bright spot. A set of points is thereby obtained to which an ellipse is fit (Fig. 2). The major axis of the ellipse corresponds to the diameter (D) of the circle shown in Fig. 1. Note that the beam diameter affects the sharpness of the border between the light and dark regions, that is, as the beam diameter becomes greater the brightness gradient at the boundary decreases. Hence it is desirable to have the smallest diameter possible. However, while the beam diameter we use (~1 mm) causes slight ‘fuzziness’ or ‘smearing’ of the boundary it nevertheless holds true that the location

The technique used for clear ice and liquid layers makes use of the refractive property of the material to remotely detect and measure its thickness on a surface. A beam of focused light, such as from a laser of any visible wavelength outside the region where ice strongly absorbs (i.e., not greater than 1 μm) is used. The beam is directed towards the surface on which a layer of ice is present (Fig. 1). The angle of incidence is not important and can be very large. The beam traverses the ice layer and impinges on the underlying surface, producing an intense bright spot from which light scatters in all directions (for diffuse reflecting surfaces). All of the light scattered from the spot and striking the ice/air interface at an angle of incidence less than the critical angle (theta) passes through the interface. Those rays incident at angles greater than theta get internally reflected from the ice/air interface to strike and illuminate the surface again. Consequently the view from above, normal to the surface, shows a bright spot where the laser first strikes the surface, in the center of a dark circular region that is surrounded by a light region that gives the dark region a sharply defined perimeter (diameter D). The circular boundary, with diameter D, marks the distinction between light from the bright spot that was internally reflected inside the layer and light from the bright spot that passed back out through the top ice/air interface. The thickness of the layer is related to the diameter of the circular boundary and the refractive index of the layer material by the simple expression: H¼

D 4  TanðΘÞ

ð1Þ

−1

Θ ¼ sin


 1 n

ð2Þ

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

Fig. 2. Image from an ice layer measurement showing the elliptical fit at the light/dark boundary and the major axis of the ellipse. Note that the present system uses a much higher resolution still camera than the video camera that was used in the earlier system. Hence the fairly large size of the bright spot in the middle is now much smaller due to the higher image quality of the camera. The iced surface was tilted relative to the view direction, hence the pattern of illumination is elliptical rather than circular.

of the maximal value of the brightness gradient at the boundary region still corresponds to the location of the very sharp boundary that would result from using an extremely small beam diameter. Hence the value for D (Fig. 1) is still reasonably accurate in spite of the fact that the beam has a finite diameter.

9

Fig. 4. Detailed schematic for Optical method 2. Note that the blob ‘perimeter’ is not a well-defined line but rather, for descriptive purposes in discussing the figure, it refers to an ellipse defined by a line of arbitrary equal brightness.

According to the simplified geometry of the situation the separation of the two spots (S) as viewed by the sensor is related to the thickness (H) of the layer according to the expression: h     i −1 −1 H ¼ S= tan sin ðsinðVÞ=nÞ þ tan sin ðsinðLÞ=nÞ cosðVÞ

ð3Þ

2.2. Optical method 2 The basic operation of the second technique is illustrated in Figs. 3 and 4. The method can be used for clear or foggy layers of solids or liquids. We will first consider the case of a clear layer on a surface. A focussed beam of light (of suitable wavelength as in the first method), such as from a laser, is directed at the surface with the clear layer. The beam first strikes the layer top surface and a small fraction of the beam is diffusely reflected from this point (Spot1, Fig. 3). The rest of the beam refracts through the layer and strikes the underlying surface and is diffusely reflected from that spot (Spot2 in Fig. 3). A sensor placed at an angle to the incident beam would detect both spots. Fig. 4 shows the geometry of the situation in more detail where L is the impingement angle of the beam relative to the normal to the surface as measured positive to the right of the normal and negative to the left, V is the angle of view relative to the normal to the surface as measured positive to the left of the normal and negative to the right and the angle of refraction is r. Note that Fig. 4 includes details associated with both the clear and foggy layer cases where for the moment we are considering the former. The sensor (a digital camera attached to a telescope) is aligned at an angle to the laser so that both spots, as diffusely reflected from the top and bottom interfaces of the layer, are visible due to the geometry (Figs. 3 and 4).

Fig. 3. The basic operation of the second technique, Optical method 2. The method can be used for clear or foggy layers of solids or liquids.

where V is the telescope view angle, L is the laser impingement angle and n is the refractive index of the layer material. The distinctiveness of the spots will be enhanced if the incident laser beam is focused to a small diameter on the surface (≤1 mm). The laser beam can be focused to this diameter even from significant distances (>20 m) using a beam expander and focussing optics. Hence the laser and detector can be large distances from the surface with the layer. The case described above was for an optically clear material layer where the two spots are distinctly visible to the viewer. In the case of an optically diffuse layer of material, such as rime ice, or cold clear ice with a frosty surface, the first spot will still be visible on the top of the layer, however, the second spot, at the bottom of the layer, will be replaced by a relatively large circular or elliptical blob of light (Fig. 4). This results from scattering of the beam by the diffusive aspect of the bulk layer material, i.e. the many tiny air bubbles in the case of rime ice and/or by its frosty top surface. However, the center of the blob of light, that is, where the light is approximately the brightest, will still correspond to the point where the second spot would have appeared if the layer had been optically clear. The center of the blob of light can be determined by relatively simple image analysis software (below), or manually, and then expression (3) above can be used to determine the layer thickness as before for the clear layer case. In general the surface with the layer on it will be tilted relative to the plane containing the laser, the telescope/camera and the measurement location on the surface (Fig. 5). Consequently expression (3) will have a more complex form (below). An important aspect of this method is that the geometry of the situation must be known. The three keys parameters that have to be determined are the angles a and b (Fig. 5) and the angle of tilt T of the surface. The angles can be determined through various means as discussed by Gagnon (2005) if they are not known beforehand. We note that D in Fig. 5 is the angle between the laser and the viewer. This is determined from their pointing directions as indicated by compasses or similar devices. The angles A and B are related to angles

10

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

The angles a and b are computed directly from this expression and the expression relating a, b, A and B, along with the measured values for A and B. The tilt angle T for the surface (Fig. 6) is given by: −1

T ¼ tan



 M sinðaÞ D sinða þ bÞ

ð5Þ

Where M is the horizontal separation between the highest and lowest points on the ellipse. With a, b and T defined, the fully generalized expression for the thickness (H) of the layer is given by the expression: H¼

S h sinðbÞ cosðRV Þ tanðAV Þ 1 þ

cosðRLÞ tanðALÞ cosðRV Þ tanðAV Þ

i

ð6Þ

RV = tan − 1(tan(b)sin(T)); where: RL = tan − 1(tan(a)sin(T)); −1 −1 AL = sin (sin(L)/n); and AV = sin (sin(V)/n), where: 2

Fig. 5. Schematic of the system setup where a measurement is made on a surface that is tilted relative to the plane defined by the camera, the laser and the measurement point location. Relevant angles used in the calculations are shown. Note that the perspective in the drawing is such that the line defining the base of the vertically oriented freestanding surface and the line connecting the laser and camera lie in the horizontal plane. Consequently the laser and camera are tilted upwards at an angle from the horizontal plane.

a and b by the geometric expression (a + b) = (A + B). A and B are known from the pan and tilt data from the camera and laser tripods. The present configuration of RIDE uses an expedient way of determining a and b. By adjusting the focus on the laser optics an expanded cylindrical beam is projected onto the surface, as shown in Fig. 6. The image formed at the detector of the cylindrical beam is skewed due to the differences in angles a and b, and also by any tilt angle T of the surface, so that the beam is rendered as an ellipse. Using the same analysis software as used for Optical method 1, the geometric properties of the ellipse are computed. That is, a least squares elliptical fit is obtained for the elliptically-shaped image. The diameter D of the cylindrical beam corresponds to the extent of the ellipse in the vertical direction. The distance, E, separating the two points formed by intersecting the ellipse boundary with a horizontal line passing through its center, is related to the diameter D and angles a and b, according to the following expression: D=E ¼

sinðaÞ sinðbÞ

3   tan ð a Þ cos ð T Þ −1 6 −1 7 L ¼ 90− tan 4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5− a− tan ð tanðaÞÞ 1 þ ð tanðaÞ sinðT ÞÞ2 2 V ¼ 90− tan

3

  tanðbÞ cosðT Þ −1 7 4qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5− b− tan ð tanðbÞÞ 1 þ ð tanðbÞ sinðT ÞÞ2

−1 6

ð7Þ

ð8Þ

Once the relevant angles are known then the image analysis software locates the bright spot corresponding to the laser initially impinging on the top surface of the ice and the approximate bright central region of the large blob that illuminates the underlying surface created by the diffuse propagation of the laser beam through the ice thickness. A sample image from a test program described below is shown in Fig. 7. As an aid to the reader before proceeding we reiterate what is meant by the terms ‘spot’, ‘blob’ and ‘light region’. A ‘spot’ is created when the beam strikes the top surface of the ice layer and also the underlying surface in the case where the ice is clear and does not have a frosty surface (Fig. 3). A ‘spot’ has a well-defined perimeter and relatively small size whereas a ‘blob’ is a larger area created by diffuse expansion of the laser beam as it scatters off small air bubbles while propagating through the ice layer. A ‘blob’ may also be produced on

ð4Þ

Fig. 6. A perspective drawing illustrating the projection of an expanded cylindrical laser beam onto a rectangular surface that is tilted backwards relative to the viewer by an angle T from vertical. The image formed is skewed due to the differences in angles a and b (Fig. 5), and also by the tilt angle T of the surface, so that the beam is rendered as an ellipse. Relevant parameters of the ellipse are shown that are used to determine angles a, b and T.

Fig. 7. An image captured by the RIDE system during a test program where the thickness of a foggy ice layer on a metallic surface was determined. The bright spot at the right corresponds to the location where the low-power laser beam impinges on the top surface of the ice. The large and more diffuse illuminated region taking up the majority of the image corresponds to the light blob on the underlying surface caused by scattering of the laser beam by the tinny air bubbles in the ice as it propagates through the layer. The center of the right spot and the compensated ‘center’ of the left blob are indicated by ‘+’.

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

an underlying surface if the top surface of a clear ice layer is frosty. The blob ‘perimeter’ in Fig. 4 is not a well-defined line but rather, for descriptive purposes in discussing the figure, it refers to an ellipse defined by a line of arbitrary equal brightness. In Fig. 7, for example, we refer to the small well-defined bright area on the right as a ‘spot’ whereas the much larger illuminated area covering most of the image, with intensity decreasing from the center outwards, we designate as a ‘blob’. The ‘light region’ in Figs. 1 and 2 is distinct from the designations ‘spot’ and ‘blob’ since it is annular in shape and is not caused by scattering and diffusion of the beam while passing through the layer, but rather it is due to internal reflection within the clear ice layer of light scattering from the small bright spot where the laser impinges on the underlying surface. The first step in the automated analysis of an image is to convert the image to gray scale and then take a summation of the pixel intensities for each column in the image (curve #1, Fig. 8). Then a large Hanning window is applied to the curve representing column intensity sum versus column number to eliminate the noise on the curve. Next a Butterworth high-pass filter is applied (curve #3, Fig. 8). The Butterworth filter enhances the central peak of the curve and diminishes the values at the sides of the central peak. The resulting plot exhibits a main central peak, corresponding to the left center of illumination, with a bump on the right side corresponding to the bright spot at the right. Generally the peak associated with the left light blob is not symmetric so a ‘center’ of the peak is determined using the following method. The shape of the peak is skewed slightly because the cone of light illuminating the surface that creates the blob is at an angle to the surface (Fig. 4). This tends to pull the perceived center of illumination (from the summed columns) to the right. That is, more light is ‘lost’ on the left than on the right for two reasons: on the left the cone of light is spread out more so more of it falls outside the image than on the right, and since the cone of light is spread out more on the left it is less intense and therefore some light is possibly ‘lost’ due to thresholding sensitivity associated with camera settings. The skewing effect would ultimately lead to positive offsets and reduced slopes in linear plots of measured ice thickness versus actual ice thickness. To compensate for this small but noticeable effect we use a method involving taking the derivative of a spline fit to curve #3 (curve #4, Fig. 8). The spline fit served to produce a smooth derivative curve. The rest of the procedure is illustrated in Fig. 9 where the relevant portion of the derivative curve needed to find the left light blob is shown. That portion is from the first peak (peak #1) in curve #4 to the first trough (trough #1). The absolute value of this curve is then taken (plot #2, Fig. 9). The plot is then inverted by subtracting all the values from the maximum value (plot #3, Fig. 9). The center of

11

mass (CM, Fig. 9) of the resulting plot is taken as the compensated ‘center’ of illumination. We found this method made the best use of the visual information and resulted in a good linear fit with negligible zero offset (see below). The computer routine identifies the location of the summed columns with the highest value for the image and this always corresponds to the blob at the left in the image. Once the compensated ‘center’ of illumination for the left blob has been identified the portion of the image to the left of that point is cropped and a similar routine runs on the remaining portion of the image to locate the center of the right spot, but this time without using the Butterworth filter. This portion of the image generates curve #2 and curve #5 in Fig. 8. Then the same procedure that generated Fig. 9 is applied to curve #5 from peak #2 to trough #2 (Fig. 8) after the curve is adjusted upwards so that the ‘zero’ amplitude corresponds to what was formerly the average of the peak and trough values. This secondary routine performs well, as was confirmed by comparing the calculated position of the center of the right spot with manual estimates obtained directly from the images. We note that in the case of the right spot that the CM and the peak on plot #3 closely correspond since the right spot is created by reflection of laser light from the top surface of the ice and is not subject to the effect of the angled light cone discussed above for the left blob. 3. Capabilities, limitations and accuracy The minimum thickness that the system can measure depends on the nature of the layer. If it is a clear layer then the minimum thickness, using Optical method 1, is about 0.5 mm. The diameter of the dark circular region (Fig. 1) begins to approach the diameter of the bright spot produced by the laser beam, as perceived by the camera, for thinner layers. For the case of foggy ice with air bubble inclusions the minimum measurement thickness is about 1.5 mm, as empirically determined from our tests. For thinner layers in the present tests Optical method 2 could not distinguish between the left blob and right spot. It is expected, however, that for certain cases where the angle between the laser and the camera are set to greater values, such as by increasing angle L in Fig. 4, then thinner layer thickness measurements could be made because this arrangement creates greater separation between the blob at the left and spot at the right. In principle if there is no limit on the laser output power then both optical methods can handle measurements of arbitrarily thick layers where ambient light is filtered out. In practical terms, however, it is often desirable for safety reasons to keep the laser output at Class IIIa (American National Standards Institute, Z136), i.e. b5 mW. While we did not investigate the upper limit for thickness measurements, our tests suggest that when using the normal laser output

Fig. 8. Chart that illustrates various primary stages of automated processing that a typical image (Fig. 7) undergoes in order to locate the right and left centers of illumination.

12

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

Fig. 9. Chart that illustrates the secondary stages of automated processing that the image in Fig. 7 undergoes in order to locate the left compensated ‘center’ of illumination. Note that plot #1 and plot #2 overlap on the left-hand portion of the chart.

(~15 mW) of the RIDE system either optical method would be suitable for measurements of layers several centimeters thick. Another test program would be necessary to fully investigate what the upper limit would be for thickness measurements in the case of the normal laser output and when it is limited to 5 mW. Ice surface roughness, or roughness of the underlying surface (e.g. Shuttle foam, see below) does not hinder the system's ability to measure ice thickness for the case of Optical method 1. What is obtained in those cases is an average value for the ice thickness in that particular region. For Optical method 2 if the scale of the ice roughness is greater than the diameter of the laser beam then the angle of the beam's entry through the ice surface would be affected. In that case it would be advisable to take a number of measurements at the location of interest where the position of the beam on the surface is changed slightly for each measurement and then to take the average value as a reasonable estimate for the ice layer thickness. If the roughness features are on the scale of 1 mm and if the ice layer is several millimeters or more in thickness then another strategy would be to slightly expand the beam diameter to a few millimeters so that it is greater than the scale of the roughness of the ice surface. Optical method 2 will still work in this case. For example, in Fig. 7 the effect of expanding the beam diameter would be to increase the size of the bright spot at the right while its center would remain fixed. Similarly the position of the light blob to the left of the spot would remain fixed. Hence the image analysis and calculations would yield the same result for the thickness. In principle this strategy would work as long as the beam is not expanded to the extent to cause the perimeter of the right spot to encroach upon the central area of the light blob. Therefore the strategy would work best for relatively thick layers of ice. If the ice surface is not parallel to the underlying surface the geometry of the situation is somewhat self-correcting so that the level of error in a thickness measurement is acceptable for moderate non-parallelness. For example, if the ice surface has a 7° nonparallelness to the underlying surface it would introduce less than 3% error in a thickness measurement for the case of Optical method 1. The same degree of non-parallelness in the case of Optical method 2, where the laser impingement and viewing angles are 30°, leads to about 5% error. The diameter of the laser beam is well controlled by the focusing optic, achieving an approximate 1 mm diameter throughout the operating range of the system. Similarly the diameter of the expanded

beam can be accurately controlled using the optic. The laser intensity received at the surface where a measurement is to be made is the same regardless of the distance since the beam is always focused to a diameter of ~1 mm for a thickness measurement. The level of uncertainty in the laser and camera pan and tilt angles (±1°), that similarly affects the angles used in the calculations, results in only a 1% uncertainty in a thickness measurement. In addition to the magnifying power of the telescope the digital camera itself has optical and digital zoom capability to allow well-sized images even for the longer distance measurements. In other words, the scale of the images does not vary much within the system's operating range. For these reasons the system does not exhibit substantial variation in accuracy of measurements over its working range (approximately 6.5–30 m). We reiterate that the system works in the same manner using any laser wavelength in the visible region. Other than when the camera tripod and laser tripod mechanisms are moving during pan and tilt maneuvers there is no mechanical jitter in the system to cause blurring of the images. It is anticipated, however, that in the outdoor environment wind could cause some relative shaking that could affect measurements. Wind protection enclosures would be a simple solution if warranted. In situations where the platform(s) for the tripods itself is vibrating then the tripods would have to be isolated from those vibrations. Regarding Optical method 2, the small air bubbles in the ice used in our tests did not affect the angle of refraction as the beam passed through the top surface of the layer. Note that the beam diameter was roughly an order of magnitude greater than the typical bubble diameter. The layer surface was relatively flat and the laser wavelength was much smaller than the air bubble size so the beam necessarily refracted according to the refractive index of bubble-free solid ice. In our case the only effect of the bubbles was to cause scattering of the beam after it passed through the solid surface. However the center of the cone of light produced by the scattering followed the same path as a beam passing through clear ice since the bubbles were symmetrical, i.e. spherical, and fairly uniformly distributed. The presence of small air bubbles is typical for rime ice and any ice that has frozen rapidly from the liquid phase (e.g. Kermani et al., 2007; Pflaum, 1984). If thickness measurements are to be made on a layer of ice that is known to have significant numbers of air bubbles with diameters approaching 1 mm then a mitigating strategy would be to expand the beam to a few millimeters in diameter, similar to the strategy mentioned above where the ice surface roughness is on the order of 1 mm.

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

The technology is capable of measurements on moving objects. For rapidly moving objects such as aircraft propeller blades that intersect the field of view frequently a high-resolution video camera can be used or still camera with suitable exposure duration. When viewing normal to the plane of the propeller rotation blades will intersect the field of view a number of times within a single video image, or a single still camera image, to produce a stationary image of the laser pattern produced on the surface since the laser spot is stationary relative to the camera. One obtains an average thickness of the icing layer in this case for the portion of the iced surface on the blades that passes through the field of view. In the case of a road ice measurement from a moving vehicle the camera captures an image where the surface has moved during the exposure so one obtains an average ice thickness for the length of iced road in the exposure. In the case of slower moving rotating objects such as wind turbines a video camera, or still camera, would capture an image of a blade passing through the field of view at regular intervals. This is very similar to the previous study where icing measurements were made on a model rotating helicopter rotor (Gagnon and Marcotte, 1995) where a video camera was used to capture the image data. In those tests the images were analyzed after the tests. For the present automated analysis system a straightforward synchronization technology can be envisaged to synchronize image capture when moving objects are in the field of view, such as optical sensors that trigger the still camera when a blade is in a suitable position. Similar devices can also be used to open a shutter briefly for the laser in cases where a continuous beam of laser light is undesirable for eye safety reasons. Finally we mention the technology's capability to distinguish between water and ice layers. In the case of a rime layer the laser beam creates a fairly large blob of light on the surface, rather than a small bright spot, due to scattering by the bubbles in the ice. This is an obvious indicator of ice. For glaze ice the situation is a little more complicated. According to Gagnon (1996) if the layer is liquid water the light annular region around the dark region (Fig. 1) is perfectly regular in appearance around its circumference, with diminishing intensity as the radius increases. However if the layer is natural clear ice (i.e. glaze), its interior will have inherent irregularities such as grain boundaries and some small air bubbles. Surface irregularities will also be present. If very fine surface irregularities are present they will give the ice a frosty texture whereas larger irregularities will make the surface bumpy. Whatever the nature of the inherent irregularities of the ice they will cause shadowing and reflective effects that will cause distinct disruptions in the appearance of the light annular region from its appearance when the medium is liquid, while maintaining its overall annular shape. Additionally, the circular border between the dark and light regions will be somewhat fuzzy if the ice surface has a frosty texture. Note that the distinctiveness of the irregularities in the light region can be enhanced by slight movements of the laser beam, deliberately applied by the user or from ambient movement. Hence, localized irregularities in the intensity of illumination of the light annular region and/or reduction in the sharpness of the border between the light and dark regions, compared to its known sharpness for liquid layers, indicate that the medium is relatively clear ice. Software that could make this qualitative assessment would not be difficult to code and implement in the RIDE technology, however this has not been done yet. At present the user would be able to manually do the assessment by viewing the captured images.

13

The first of the two major components is the laser, situated on the tripod at the right in Fig. 10. The laser (Fig. 11A) is a 15 mW HeNe model (Melles Griot™) with a wavelength of 633 nm. Attached to the front of the laser is a beam expansion and focusing optic (Melles Griot™) (Fig. 11B) that enables focusing of the laser beam to a small point (≤ 1 mm diameter) at variable distances up to 30 m. The same optical attachment is used to expand the laser beam to create a uniform cylinder of light with variable diameter up to about 3 cm that is utilized to determine geometry angles for the calculations (see Fig. 6). The focus dial on the laser is controlled by a user-operated servomotor with belt drive (Fig. 11C). The operator, who would normally be positioned near the other tripod, is able to control the servomotor from a distance using a control box to rotate the focusing dial. Note that laser beam intensities b5 mW are sufficient for the RIDE technology to function and the output of the laser can be diminished in a controlled manner using a pair of polarizing filters with different rotation settings. Laser output power in the 1–5 mW range corresponds to Class IIIa (American National Standards Institute, Z136). Many laser pointers fall within this classification. The laser and its peripheral components are mounted on a flat base made of Lexan. The base plate is attached to the top of a commercial motorized pan/tilt device (Quickset ™) that allows the user to control the pan and tilt in order to point the laser at desired locations where measurements are to be made. During setup the base of the Quickset device is leveled and oriented on the tripod using the tripod's own pan/tilt mechanisms. This Quickset pan/tilt device relays its pan and tilt information to the data acquisition box. The box is connected to the computer that samples the various device orientations and acquires camera images in order to perform a thickness calculation using the above expressions. The motorized pan/tilt device is mounted on top of the rugged tripod that serves as a stable platform. A Lexan weather housing (Fig. 10) fits over the laser assembly and clamps to the Lexan base. A small video camera with a zoom lens, mounted onto the base plate beside the laser, enables the operator to see where the laser is pointing. The camera is connected to a monitor situated at the operator's location. The monitor and zoom controller with extension cable, enable the operator to view the laser spot location with sufficient precision while executing changes of position using the motorized pan/tilt device. 4.2. Detector — Camera/telescope The other major component is the detector. It consists of a digital camera (Canon Power Shot S2IS 5 megapixels) (Fig. 12A) coupled to an eyepiece (Scopetronix MaxView™ 40) that is attached to a

4. System hardware components 4.1. Laser Fig. 10 shows the two main system components situated on their respective heavy-duty tripods (Quickset™). Also visible in the photo are the Lexan™ weather housings for the components resting at the bases of the tripods.

Fig. 10. The two main system components situated on their respective tripods. The laser tripod is on the right. Also visible in the photo are the Lexan™ weather housings resting at the bases of the tripods.

14

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

at a location where a measurement is desired. Once the beam is in position then the user manually adjusts the pan and tilt wheel mechanisms on the detector tripod to point the camera and telescope at the spot where the laser impinges on the surface. When the laser and detector have been properly oriented the user captures an image with the laser optic set in the expanded beam mode in order to determine the precise geometrical relationships between the detector, the laser and the surface with the layer as outlined above, if not known beforehand. Then an image is captured with the laser optic set in the focused beam mode. In the case of a foggy ice layer the geometry information is then used with the perceived distance (S) between the center of the initial bright spot on the top of the layer and the approximate center of the blob of light on the underlying surface, as determined from the captured image and the scale, to compute the layer thickness via expression (6). Fig. 11. HeNe laser (A) and optic (B) for beam expansion and focusing. The optic is controlled by a servomotor (C).

5. Test programs

telescope (Schmidt-Cassegrain Celestron C5 Spotting Scope) (Fig. 12B). The near focus limit of the system is roughly 6.5 m. For shorter distances the telescope would have to be replaced with one of lower power. A servomotor (Fig. 12C) is attached to the focus dial on the telescope to allow the operator to remotely control the focus. A potentiometer attached to the focus dial relays the information about the dial setting to the data acquisition box. Note that the telescope focus dial was calibrated against the scale of a sharplyfocused grid pattern as it was moved through a range of distances (~11–30 m) from the detector. Therefore, the focus dial setting can be used to determine the associated scale of captured images. Potentiometers are also attached to the tripod's pan and tilt mechanisms to relay pan/tilt information to the data acquisition box. This information, along with that from other sources (above), is used to determine the precise geometrical relationships of the system components during any particular measurement. In certain lighting conditions, such as bright sunlight, an optical filter is required to increase the signalto-noise ratio for the captured images. The optical filter is a narrow band-pass filter centered on the laser wavelength. If light conditions warrant it the operator can choose to remotely deploy the filter situated on a servomotor-driven slider (Fig. 12D) located between the eyepiece of the telescope and the camera. The telescope and camera are mounted on a Lexan base plate that directly attaches to the tripod. A Lexan weather housing (Fig. 10) fits over the telescope/camera assembly and clamps to the Lexan base. In operation, while standing near the detector tripod, the user would remotely position the laser beam to impinge on the surface

Fig. 12. Digital camera (A) and telescope (B). The telescope's focusing knob is controlled by a servomotor (C). A filter, for use in bright sunlight conditions, can be deployed by a second servomotor (D).

Test programs using earlier versions of the technology for measuring clear ice and liquid layers have been conducted. Bench top measurements of a clear ice layer on a flat surface were conducted by Hermanto and Gagnon (1996). The technology was also used to measure icing layers on an airfoil and on a rotating model rotor in an NRC icing facility (Gagnon and Marcotte, 1995). The early system was furthermore used to measure the thickness of de-icing fluid during tests of wing performance in the presence of contaminated fluid in a wind tunnel facility (Oleskiw et al., 1995). More recently, the clear ice measurement technology was used at the US Army TARDEC Facility in Warren, Michigan to determine clear ice layer thicknesses on foam samples of the type used on the Space Shuttle external fuel tank using Optical method 1. The system performed well for both flat foam samples and samples that had an irregular surface texture (Fig. 13). Successful measurements were made over a range of distances up to ~28 m. Fig. 14 shows results of thickness measurement tests that were conducted on a flat foam sample with varying ice thickness at a distance of ~15 m. The standard deviation for the linear fit to the data is ~0.08 mm. Fig. 2 shows a sample image from the tests conducted using a flat foam surface and Fig. 15 shows an image from the tests using the foam with the irregular surface texture. In both cases the surfaces were tilted relative to the camera. Note that the present system uses a much higher resolution still camera than the video camera that was used in the earlier system. Hence the fairly large size of the bright spot in the middle of the images in Figs. 2 and 15 is now much smaller due to the higher image quality of the camera.

Fig. 13. Foam sample of the type used on the external fuel tank of the Space Shuttle. A rectangular-shaped clear ice layer is present on the central region of the foam's surface.

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

Fig. 14. Plot of detected ice thickness versus actual ice thickness for tests conducted on a locally flat foam sample, at a distance of ~ 15 m. The actual measurements were obtained using a tactile dial gauge that was accurate to within 0.025 mm. In preparation for the tests a thin sheet of clear ice (roughly 4 mm thick) was bonded onto the top of a piece of foam by freezing a thin layer of water situated between the foam and the ice. Then the top surface of the ice was melted with a warm metal plate to ensure it was flat and close to parallel with the base of the foam (within ~ 1°). Consequently the foam, that had a series of pre-cut descending steps of slightly diminishing height, formed a series of plateaus (each ~ 25 mm wide) that had incrementally thicker ice layers from which the remote ice thickness measurements were obtained. The standard deviation for the fitted line on the plot is ~ 0.08 mm.

15

Fig. 16. A photo of the iced lifeboat hook used for the ice detection tests. The laser beam can be seen where it impinges on the rectangular region where the ice layer has been melted flat. Slots in the ice can be seen at the sides of the melted surface where the bare metal of the hook is visible. The slots correspond to the areas where the stacks of spacers, affixed near the ends of the slightly warm aluminum rectangular column that was used for melting the ice, eventually touched the hook metal briefly at the end of each controlled melt step.

Most recently the present system, which has clear and foggy layer capability, was used to measure thicknesses of accumulations of foggy freshwater ice layers on a metallic surface. The tests were conducted in the Model Preparation Area and Cold Room of the Institute for Ocean Technology, NRC. The metal surface was that of a lifeboat hook (Fig. 16) that was installed in the Cold Room to test icing effects on its performance. The Cold Room was equipped with two windows to provide optical access for the RIDE technology (Fig. 17). The detector and laser were located ~ 14 m and ~ 12 m respectively from the iced metal surface with an angle of about 51° between the laser beam and the viewing direction. The icing layer was applied to the metal surface by spraying fresh water at 0 °C onto the cold metal surface until a layer in access of 1 cm thickness had accumulated. The air temperature and initial metal temperature during the spraying was −15 °C. When enough ice had accumulated the Cold Room air temperature was brought to −10 °C. Then a region of the accumulation was melted flat with a slightly warm aluminum rectangular column in a sequence of controlled steps of 0.75 mm each. This was achieved by sequentially removing a single spacer (0.75 mm in thickness) from

a stack of spacers affixed near each end of the aluminum rectangular column that kept it off the lifeboat hook surface by the prescribed amount at the end of each melting step when the stacks eventually, and briefly, touched the bare metal of the hook (Fig. 16). A number of thickness measurements were made using the RIDE technology following each melting step for a variety of exposure settings of the camera. The results are shown in Fig. 18. The standard deviation for the linear fit to the data is ~0.40 mm. Prior to the present tests a few preliminary images were captured as the lifeboat hook was iced by spraying saline water, with the approximate salinity of seawater, on its surface. The purpose of the test was to see the impact of saline icing on the lifeboat hook release mechanism. The layer was about 2 mm thick. One could see from the images that the detection system was behaving in a similar manner as that in the tests reported here for freshwater ice. However, since measurements were obtained for only one thickness that was near the low end of the system's measurement capability, where scatter is expected to be greater, no rigorous assessment of accuracy could be made. Manual analysis of the images yielded a thickness of roughly 2 mm. It was noted that the bright spot at the right was not as distinct as in the freshwater ice layer cases, because of the spongy wet texture of the ice that formed as the spray liquid was freezing over a period of several minutes. The determination of the accuracy of the RIDE

Fig. 15. An image from the tests using the foam with the irregular surface texture (e.g. Fig. 13). Note that the iced surface was tilted relative to the view direction, otherwise the pattern of illumination would have been circular. The image has been enhanced so that the reader can more easily see the border between the dark and light regions that generated the elliptical fit.

Fig. 17. View of the Model Preparation Area and Cold Room at the Institute for Ocean Technology, NRC, where ice detection tests were performed to measure the thickness of an icing layer on a lifeboat hook. The Cold Room was equipped with two windows to provide optical access for the RIDE technology. The laser tripod is at the right and the camera tripod is at the left.

16

R.E. Gagnon et al. / Cold Regions Science and Technology 72 (2012) 7–16

of ice or liquid on surfaces at distances in the range of 6.5–30 m. The essential components of the prototype are a low-power laser with a long-distance focusing optic, a compact telescope and a digital camera. The system has many potential applications such as measuring the thickness of ice layers on aerospace vehicles, wind turbines, road surfaces, power lines, and safety and communications equipment on marine structures and ships. The prototype can be easily miniaturized for short-distance measurement applications. Acknowledgments

Fig. 18. Plot of detected ice thickness versus actual ice thickness for the tests conducted on the lifeboat hook that had a layer of foggy ice. The actual ice thickness was determined from controlled melting of an ice accumulation in a series of steps of 0.75 mm ± 0.03 mm. The standard deviation for the fitted line is ~ 0.40 mm.

The authors are grateful to IOT/NRC and PERD for financial support of this project. The authors would also like to thank Dr. Tom Meitzler and his group at the US Army TARDEC facility in Warren, Michigan where the tests on clear ice layers on Shuttle foam samples were carried out. References

system for measurements on saline ice layers and the investigation of optical effects of the liquid brine inclusions requires a full set of tests similar to that reported here. 6. Considerations The present configuration of the RIDE technology is suited to measurements at fairly large distances from structures such as aircraft during on-ground inspection, and space vehicles (e.g. Space Shuttle) prior to launch. It is also suited to measurements of icing on large wind turbines. There are several other situations where fairly localized measurements of icing are needed such as safety and communications equipment on marine vessels and structures, in-flight icing measurements on fixed and moving aircraft components, road ice measurements from motor vehicles, and small-scale wind turbines. In these cases the RIDE technology can be easily reduced in size to a compact intelligent unit for measurements at preset locations. At short distances, and with predefined geometry, all of the bulky components such as the telescope, tripods, and servomotors, are no longer necessary. Essentially a small diode laser and compact camera, with electronic peripherals for computation and communication, and a weather housing are all that is required. 7. Conclusions A remote ice detection prototype device has been described that is capable of accurate thickness measurements of foggy or clear layers

Blackwood, C.I.R., 1993. Apparatus and Method for Ice Detection. US Patent 5,243,185, September 7. Bridge, P., 2008. Non-invasive Road Weather Sensors. Paper No: Weather2008-002. 4th National Conference on Surface Transportation Weather. Indianapolis, Indiana (June 16–17, 2008). Christian, D.J., Robertson, G.D., Tucci, J.M., Westcott, J., 1993. Apparatus for deicng. US Patent 5,180,122, January 19. Gagnon, R.E., 1996. Method and Apparatus for Remote Detection and Thickness Measurement of Ice or Liquid Layer. US Patent 5,541,733, July 30. Gagnon, R.E., 1997. Speckle Interference Method for Remote Thickness Measurement of a Sheet and Layer and an Apparatus Therefor. US Patent 5,650,851, July 22. Gagnon, R.E., 2005. Method and Apparatus for Layer Thickness Measurement. US Patent Provisional Application 60/721,996, September. Gagnon, R.E., Marcotte, D.L., 1995. Icing Thickness Measurements from an Airfoil and Model Helicopter Rotor Using a New Remote Technique. Proceedings of the American Helicopter Society International Icing Symposium '95, Montreal, Canada, September, 1995, pp. 417–428. Gregoris, D., Yu, S., Teti, F., 2004. Multispectral Imaging of Ice. IEEE Canadian Conference on Electrical and Computer Engineering, Niagara Falls, May 2004. Hermanto, I., Gagnon, R.E., 1996. Machine Vision for Ice Layer Thickness Measurements. SPIE's International Symposium on Intelligent Systems and Advanced Manufacturing, Boston, Massachusetts USA, November, 1996. Kermani, M., Farzaneh, M., Gagnon, R., 2007. Compressive strength of atmospheric ice. Cold Regions Science and Technology 49, 195–205 (2007). Oleskiw, M.M., Penna, P.J., Crabbe, R.S., Beyers, M.E., 1995. Full-Scale Wind-Tunnel Simulation of Takeoff Performance Degradation with Contaminated Fluid Runback. Proceedings of the American Helicopter Society International Icing Symposium '95, Montreal, Canada, September, 1995, pp. 179–190. Pernick, B.J., 1999. Aircraft Ice Detecting System. US Patent 5,921,501, July 13. Pflaum, J.C., 1984. New clues for decoding hailstone structure. Bulletin of the American Meteorological Society 65 (6), 583–593. Sinnar, A.M., 1989. Compositional State Detection System and Method. US Patent 4,808,824, February 28.