- Email: [email protected]
Optimization of QR code readability in movement state using response surface methodology for implementing continuous chain traceability
Computers and Electronics in Agriculture 139 (2017) 56–64
Contents lists available at ScienceDirect
Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag
Original papers
Optimization of QR code readability in movement state using response surface methodology for implementing continuous chain traceability Jianping Qian a,b, Xiaowei Du a,b, Baoyan Zhang c, Beilei Fan a,b, Xinting Yang a,b,⇑ a
National Engineering Research Center for Information Technology in Agriculture, Beijing Academy of Agriculture and Forestry Science, Beijing 100097, China National Engineering Laboratory for Agri-product Quality Traceability, Beijing 100097, China c Tianjin Rural Affairs Committee Information Center, Tianjin 3300061, China b
a r t i c l e
i n f o
Article history: Received 6 May 2016 Received in revised form 9 May 2017 Accepted 11 May 2017
Keywords: QR code Traceability Response surface methodology Barcode readability Optimization
a b s t r a c t Logistics and storage is the main processing for agro-food supply chain. Because of disconnection information between the two processing, it is difficult to trace continuously. An intelligent conveyer belt provides an effective method to associate storage and logistics by QR code scanning and information recording. Improving the QR code readability in movement state is the core of implementing continuous chain traceability with this belt. In this paper, a intelligent conveyer belt including automatic conveyer unit, barcode scanning unit, fault remove unit and control display unit was designed. Four factors affected QR readability were selected and the value range was confirmed, which was reading distance, code size, coded characters and belt moving speed. Based on the belt, an Central Composite Inscribed (CCI) experiment of four factors with five levels was designed using Response Surface Methodology (RSM) to obtain the optimal reading parameters. The result shows that the main factors of reading distance, belt moving speed and the interaction between reading distance and code size have the significant effect on QR code readability. Under the optimization condition of 141.45 mm reading distance, 34.58 mm code size, 100 bytes coded characters and 2.98 m/min belt moving speed, the average value of QR code readability was 95%. With the optimization parameters, the intelligent conveyer belt was used in an apple marketing enterprise. The result shows that the continuous traceability between storage and logistic can be implemented with the extended breadth, deepened depth and improved precision. Ó 2017 Elsevier B.V. All rights reserved.
1. Introduction Traceability is an effective method to ensure food safety and quality and to reduce the costs associated with recalls (Regattieri et al., 2007; Yang et al., 2016). In the last three decades, some astounding events, such as the BSE crisis and the problems posed by foods ingredients from genetically modified (GM) crops, have strongly focused attention on the topic of agro-food traceability (Bertolini et al., 2006). The ISO 22005:2007 food traceability standard requires that each company knows both its suppliers and customers, based on the principle of one-up and one-down (International Organization for Standardization, 2007). To achieve traceability, traceable unit identification and information record are very important. Identification technologies such as barcode and Radio Frequency Identification (RFID) as
⇑ Corresponding author at: National Engineering Research Center for Information Technology in Agriculture, Beijing Academy of Agriculture and Forestry Science, Beijing 100097, China. E-mail address: [email protected] (X. Yang). http://dx.doi.org/10.1016/j.compag.2017.05.009 0168-1699/Ó 2017 Elsevier B.V. All rights reserved.
distinguishing traceable units method can be integrated into a supply chain system (Cunha et al., 2010; Ruiz-Garcia and Lunadei, 2011). Ampatzidis et al. (2009) evaluated using RFID and barcode technologies in manual fruit harvesting to improve traceability. A platform for livestock management was also formulated by deploying mobile computing, RFID technology and wireless/mobile networking (Voulodimos et al., 2010). With the development of mobile communication technology, the information collecting and uploading time using portable devices (mobile phone, PDA, tablet PC) has become an effective means for farming operation information collection (So-In et al., 2014; Qian et al., 2015). Steinberger et al. (2009) developed mobile farming information collection equipment that transmitted the information to the server through the internet. Under the framework of from farm to table, it including many supply chain nodes, such as product, storage, logistic and sale. In the inner of supply chain enterprise, it is convenient to trace continuously through traceable unit identification and information record. For example, Qian et al. (2012) describe a study with a primary goal to develop a Wheat Flour Milling Traceability System
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
(WFMTS), incorporating 2D barcode and RFID technologies, and to validate the system in a wheat flour mill enterprise in China. Actually, because of disarticulated traceable unit identification and absent information record, it is difficult to trace continuously between supply chain nodes. The existing studies and applications provide a good reference for constructing traceability system framework or implementing supply chain inner traceability. But implementing continuous traceability between supply chain nodes faces some challenges with the demand of efficiency: (1) Transformation information can be recorded synchronously with the traceable unit identification; (2) High intelligent level can improve efficiency and reduce error. A intelligent conveyer belt including automatic conveyer unit, barcode scanning unit, fault remove unit and control display unit provided a effective solution for continuous traceability between storage and logistics. Difference with the static state (Tarjan et al., 2014), optimization of barcode reading in movement state with the intelligent conveyer belt need to further study. Response Surface Methodology (RSM) is a compilation of mathematical and statistical methods suitable to optimize, develop and improve processes (Bingol et al., 2012). In this paper, RSM was used for obtaining the optimal reading parameters and the index of breath, depth, precision was evaluated for continuous traceability effects. 2. QR code and the intelligent conveyer belt 2.1. QR code QR code is a two-dimensional barcode defined by the industrial standard ISO/IEC18004:2006 (International Organization for Standardization, 2006). QR code structure is shown in Fig. 1. Each QR code is structured by dark (logical ‘‘1”) and light (logical ‘‘0”) modules. The modules are evenly distributed in a square net of fields, where the size of a field is the size of a single module. By the standard ISO/IEC18004 one module should be sized 4 4 px (pixels) with the print resolution of 300 dpi (dots per inch). This size ensures readability by the majority of optical devices. 2.2. Intelligent conveyer belt Common conveyer belts have the main function of convey goods and less for recording information. The modified intelligent conveyer belt developed by our research team consists of automatic conveyer unit, barcode scanning unit, fault remove unit and control display unit. The structure of the belt is shown in Fig. 2.
57
The speed of conveyer unit can be adjusted from 0 m/min to 3.5 m/min. NSL-FM420 scanning module (Newland Auto-ID Tech., 2016) by Newland Auto-ID Technology was used as barcode scanning unit. The module supports QR code scanning and identification of mobile barcode. In order to fitting to scan from different directions of the packages, a holder with six free degrees was designed to support the barcode scanning module. Fault remove module adopted a pneumatic mode, easy to eliminate large packaging. If the product on the conveyer belt was not accord with the product list of order, the product was removed. A industrial personal computer was used as control display unit. The unit can control the other units according to the business and display the device status and order list. The device was also integrated with the WiFi module, which can realize wireless data transmission from the remote system, such as (Enterprise Resource Planning (ERP) system.
3. Materials and method 3.1. Factors selection and value range A QR code consists of black modules (square dots) arranged in a square grid on a white background, which can be read by an imaging device (such as a camera and scanner) and processed using Reed–Solomon error correction until the image can be appropriately interpreted. The required data are then extracted from patterns that are present in both horizontal and vertical components of the image. The amount of data that can be stored in the QR code symbol depends on encoding modes (numeric, alphanumeric, byte/ binary, and kanji), version (1, . . . , 40, indicating the overall dimensions of the symbol), and error correction level (L, M, Q, H). According to the test result in static sate and fixed distance (15 cm) with different mobile phones by Tarjan et al. (2014), QR code readability is not directly influenced by the number of coded characters, or by the error correction level, but by the size of modules. Since the data storage is not so many and the error correction level is not very high for traceable encoding requirement, it is suitable with the fixed value of 7 version and M error correction level in this research. In order to test the readability with different distance, the factor of reading distance was selected. Code size was a test factor corresponding with module size under the set printer resolution. Although the coded characters not affects directly the reading rate under fixed distance, the factor also was considered under condition of variable distance and moving processing in this research. Difference with static state, moving speed was the important factor
Fig. 1. QR code structure (International Organization for Standardization, 2006).
58
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
Fig. 2. Structure diagram of the intelligent conveyer belt.
in motion process. Four factors were selected and a preliminary laboratory trial which objective was to obtain factors value range, was presented. The main factors and value range was listed on the below. (1) Reading distance Reading distance is the important factor, since every reader has different reading distance and it has a good reading rate in the special distance. For the selected scanning module, the technology handbook shows it has a reading range from 35 to 210 mm. (2) Code size Suitable size label is not only benefit to read but also reasonable to paste in the package. For actual agro-food traceability label, the size of 40 60 mm and 60 90 mm are usually applied. Therefore, the maximum value of QR code size sets to 60 60 mm. If the code size is less than 10 10 mm, it is difficult to read. (3) Coded characters QR code has bigger characters capacity than 1D barcode. For implementing traceability, some important information such as traceability ID, producer, date and check result, are encoded to the code. Other information can be queried through loading to the database with traceability ID. The encoded information is about 20–100 characters.
or in combination, and their interactions on response variables in a process allowing develop, improve and optimize such process (Bas and Boyaci, 2007). Moreover, RSM determines the effect of independent variables on the process and generates a mathematical model that accurately describes the process (Wang et al., 2016a). RSM can be helpful to quantitatively and routinely adjust parameters that influence QR code readability. Central Composite Design (CCD) and Box-Behnken Design are always used in RSM (DonisGonzález et al., 2012). For CCD, there are three types which is Central Composite Circumscribed (CCC), Central Composite Inscribed (CCI) and Central Composite Face-centered (CCF) according to different value (Wang et al., 2016b). Generally, RSM includes three steps. First step is to confirm a approximate range included optimum condition. Second step is to establish the relationship model between response and the set of independent factors. Third step is to optimize the process using the established model. (2) Experiment design According to the preliminary trial results, a CCI design of four numerical factors (reading distance, code size, coded characters and belt moving speed) with five levels. As a rule of CCI experiment, six repetitions at the center point, sixteen repetitions at the cube point and eight repetitions at the axial point were used in this study. With a a value of 0.5, the factors and its levels were shown in Table 1. (3) Data analysis
(4) Belt moving speed With the modified conveyer belt, moving speed can be adjusted. Through adjusting the belt speed with different gradient in the preliminary trial, it was found that the reading rate has a drastic decline in case of exceeding 3.5 m/min. On the other side, if the speed is less than 2.0 m/min, it affects the operating efficiency. The moving speed is set from 2.0 to 3.5 m/min. 3.2. Optimization experiment with RSM (1) RSM RSM, originally described by Box and Wilson, is a statistical tool that enables to assess the effect of the independent variables, alone
Every experiment point was repeated twenty times and the average of QR code readability acted as response value. The experimental results of the CCI were fitted with a second-order polynomial equation by a multiple regression technique. The quadratic model for predicting the optimal point was expressed as follow:
Y ¼ b0 þ
k X i¼1
bi vi þ
k k1 X k X X bii vi vi þ bij vi vj þ eij i¼1
ð1Þ
i¼1 j¼jþ1
where Y is the response value, b0 ,bi , bii , bij are regression coefficients for intercept, linear, quadratic and interaction coefficients, respectively, vi and vj are the independent variables and k is number of factors.
59
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64 Table 1 Factors levels coding. Levels
Factors
Upper cube point Upper axial point Center point (0) Lower axial point Lower cube point
(1) (0.5) (+0.5) (+1)
Reading distance (mm)
Code size (mm)
Coded characters (byte)
Belt moving speed (m/min)
35 78.75 122.5 166.25 210
10 22.5 35 47.5 60
20 40 60 80 100
2.0 2.375 2.75 3.125 3.5
The data were statistically analyzed using Design Expert 8.06 (Stat-Ease, Inc., Minneapolis, Minnesota) for optimizing reading parameters. Significant effects of independent variables on each response was determined by ANOVA. The multiple linear regression analysis of the experimental data yielded the second order polynomial models for predicting QR code readability.
QR code readability ¼ þ105:89060 þ 0:18347X 1 þ 1:94653X 2 1:01801X 3 þ 2:51080X 4 þ 8:71429E 003X 1 X 2 8:92857E 005X 1 X 3 þ 1:4286E 002X 1 X 4 1:5625E 003X 2 X 3 þ 0:25000X 2 X 4 3:1250E 002X 3 X 4
4. Results and discussion The experiments were performed according to the design matrix in Table 2. The statistic test factor, F-Value, was used to evaluate the significance of the model at the 95% confidence level. The results of quadratic model in the form of ANOVA are given in Table 3. The F-value of 16.77 implied the model is significant. There is only a 0.01% chance that F-value could occur due to noise. Also, values of P less than 0.05 indicated model terms were significant. The Lack of Fit F-value of 0.85 implied the lack of fit is not significant relative to the pure error. There is a 61.32% chance that a Lack of Fit F-value could occur due to noise. Non-significant lack of fit indicated the model have good fit. Through test data regression, the quadratic model in terms of actual factors was expressed as follows:
Table 2 Designed matrix by RSM and obtained value of response. Run order
Reading distance (mm)
Code size (mm)
Coded characters (byte)
Belt moving speed (m/min)
QR code readability (%)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
122.5 210 122.5 35 122.5 210 166.25 35 210 210 210 122.5 122.5 122.5 210 122.5 35 122.5 122.5 210 122.5 122.5 35 35 35 35 35 122.5 78.75 210
35 60 35 60 47.5 10 35 10 10 60 10 35 35 35 60 35 60 35 22.5 10 35 35 10 10 60 10 60 35 35 60
80 100 60 20 60 20 60 100 100 20 100 60 60 60 20 60 20 60 60 20 60 40 20 100 100 20 100 60 60 100
2.75 2 2.75 3.5 2.75 3.5 2.75 2 3.5 2 2 2.375 3.125 2.75 3.5 2.75 2 2.75 2.75 2 2.75 2.75 2 3.5 3.5 3.5 2 2.75 2.75 3.5
95 90 95 0 90 25 85 100 20 90 75 95 80 90 55 75 35 70 70 55 95 90 100 40 10 35 30 80 85 45
1:83538E 003X 21 5:4483E 002X 22 þ 9:96743E 003X 23 7:20375X 24
ð2Þ
4.1. Comparison of RSM method and ANN method ANN have high learning ability and capability of identifying and modeling the complex non-linear relationships between the input and output of a system (Nourbakhsh et al., 2014; Hamid et al., 2016). In order to compare RSM method and ANN method, an ANN model with three layers was employed. As the structure shown in Fig. 3, there were four neurons in the input layer including reading distance, code size, coded characters and belt moving speed. The hidden layer contained 10 neurons. QR code readability was used as only one output neuron. From input layer to hidden layer, Levenberg-Marquardt backpropagation (Trainlm Function) was used in Matlab R2014a. In 30 groups dataset, 20 groups were
Table 3 Comparison of RSM and ANN method using R2, RMSE and APD value. Method
R2
RMSE
APD (%)
RSM ANN
0.9329 0.9021
9.5060 15.0182
20.6790 10.0124
Fig. 3. ANN structure for predicting QR code readability.
60
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
a. Relation diagram between reading distance and QR code readability
b. Relation diagram between belt moving speed and QR code readability
c. Response surface figure between reading distance, code size and QR code readability Fig. 4. Relation diagram between significant factors and QR code readability.
61
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64 Table 4 Variance analysis of regression equation for QR code readability. Source
Sum of squares
DOF
Mean square
F-value
P-value
Model X1: Reading distance X2: Code size X3: Coded characters X4: Belt moving speed X1 X2 X1 X3 X1 X4 X2 X3 X2 X4 X3 X4 X21 X22 X23 X24 Residual Lack of Fit Pure Error Cor Total
24172.18 668.18 437.88 18.56 7530.68 5814.06 1.56 14.06 39.06 351.56 14.06 32.86 192.97 42.33 2.73 1544.49 973.65 570.83 25716.67
14 1 1 1 1 1 1 1 1 1 1 1 1 1 1 15 10 5 29
1726.58 668.18 437.88 18.56 7530.68 5814.06 1.56 14.06 39.06 351.56 14.06 32.86 192.97 42.33 2.73 102.97 97.37 114.17
16.77 6.49 4.25 0.18 73.14 56.47 0.02 0.14 0.38 3.41 0.14 0.32 1.87 0.41 0.03
<0.0001 0.0223 0.0570 0.6772 <0.0001 <0.0001 0.9036 0.7169 0.5472 0.0844 0.7169 0.5805 0.1912 0.5311 0.8728
0.85
0.6132
selected randomly to train, 5 groups to validate, and the other 5 groups to test. Coefficient of determination (R2) is important to describe the data relationship. A higher value for R2 indicated a good fit for the predicted model (Rakshit et al., 2015). Exception of R2, it also needs a small error between actual data and predicted data. In this research, Root Mean Square Error (RMSE) (Eq. (3)), and Average Percentage Deviations (APD) (Eq. (4)) were used to check the fits for the two methods. The RMSE and APD are defined as follow (Astray et al., 2016):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi PN 2 i¼1 ðyp ye Þ RMSE ¼ N
APD ¼
PN yp ye I¼1 ye 100 N
ð3Þ
ð4Þ
where N is the number of experiment group, yp is the predicted value, ye is the experimental value. The R2, RMSE and APD for the RSM and ANN were listed in Table 3. The R2 and RMSE for the RSM method were calculated to be 0.9329 and 9.5060, whilst that of the ANN method was 0.9021 and 15.0182. In the view of the two evaluation values, it showed that RSM method predicts more accurately than the ANN method. However, the APD value of RSM with 20.6790 is higher than that of ANN with 10.0124, which indicated that ANN method has a good precision. Although not all the evaluation indexed of RSM method was superior to ANN method, the main parameters of RSM method was dominant. Therefore, it is feasible that employing the RSM method optimized QR code readability in movement state.
The relation diagram between reading distance and QR code readability by maintaining other factors constant at middle value are shown in Fig. 4a. It was obvious that the QR code readability increased with increasing in reading distance from 35 to 140 mm. However, further increase in reading distance more than 140 mm leaded to the decline on QR code readability. It demonstrated that too short distance or too long distance was not suitable for reading. Fig. 4b shows that relation between belt moving speed and QR code readability by maintaining other factors constant at middle value. The decline tendency of QR code readability was maintained with the bet moving speed increasing. When the speed was more than 3.5 m/min, the readability was less than 61%. This is due to the fact that the reading time of every barcode was reduced with the increased speed. Fig. 4c shows the strong interaction of reading distance and code size. The variation of QR code readability with reading distance had the tendency of ascend and then descend, as cited in the one factor analysis. With the increase of code size, QR code readability observed to be increased and then decreased. It also indicated that QR code had a good reading rate in the range of middle value combination of code size and reading distance. The relationship of code size and reading distance was further analyzed with Fig. 5. Barcode scanners have a limited reading angle and distance. In the readability range, the scope enlarged with the increased distance. Therefore, it is difficult that a big size was read with a short distance because of absent the whole barcode image.
4.2. Influence factors analysis Table 4 and Eq. (2) show that the influence degree of different factors were listed belt moving speed > reading distance > code size > coded characters and main factors of reading distance, belt moving speed have the significant effects on the response (P < 0.05). There was a very significant interaction between reading distance and code size, which was the important effect of QR code readability. The relation between main factors and response can be visualized by plots which were obtained from model graph tool of the Design Expert software (Fig. 4).
Fig. 5. Sketch diagram of reading range in different distance with a constant size code.
62
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
a. Intelligent conveyer belt actual picture
Main menu
Current
state:
Scanning barcodes Product order ID Traceable ID Production name Production specification Operation time
b. Control system Fig. 6. Intelligent conveyer belt actual picture and its control system.
The low readability in the case of code size with 60 mm and reading distance with 35 mm proved the situation. 4.3. Optimization readability parameters and validation In order to obtain optimized parameters for high readability, optimization method provided by Design Expert 8.06 was used. The lower and upper bound of reading distance, code size and coded characters were respectively given 35–210 mm, 10–60 mm and 20–100 bytes. Improving the belt moving speed can improve the
efficiency for the whole system, so maximized the speed was as a subordinate goal. A group of optimal solution with a theoretical readability of 97.03% was obtained with 141.45 mm as reading distance, 34.58 mm as code size, 100 bytes as coded characters and 2.98 m/min as belt moving speed. For round numbers, 20 tests were performed under the condition of 141 mm, 35 mm, 100 bytes and 3.00 m/min. The average value of QR code readability was 95% and the relative deviation was 2.1%. It shows that reading rate was high and efficiency was good with optimal readability parameter.
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
4.4. Continuous traceability between logistics and storage Logistics and storage is the main processing for agro-food supply chain. Because of disconnection information between the two processing, it is difficult to trace continuously. The intelligent conveyer belt focuses on the entry and out warehouse of the joint of logistics and storage. It have the function of goods conveying, barcode scanning, information checking and abnormal eliminating. A control system was developed and embedded in the control display unit (Fig. 6). With the system, a control flow was established as seen in Fig. 7. The intelligent conveyer belt can exchange data with ERP system. When some products need to out of the warehouse, a electronic product order was sent to the belt control system from the ERP. The control system analyzed the order and generated the different product name, specification and quantity. The workers can see the product list in the screen and find the product in the storage. The belt was deployed in the exit of the warehouse. It was operated easily between storage and logistic trucks. When the workers put the product with traceable barcode into the belt, the belt moved with the optimization speed. The barcode scanner was triggered and read the barcode in the moving processing if the product entered the reading range. A extracting procedure was performed that product information can be obtain according to the traceable barcode. If the extracted information meet with the product list, delivery products quantity added one and information was recorded. If the meet was not success, the product was eliminated. The second judge was carried out for confirming
63
the delivery products quantity. If the quantity reached the order requirement, the next product was continued to scan. When all the products of the order were achieved, the information was uploaded to ERP. The intelligent conveyer belt was applied in a apple marketing enterprise. The enterprise has been used ERP and the simple traceability system. Continuous traceability was tested with the optimization readability parameters on the intelligent conveyer belt platform. Three parameters which is traceability breadth, depth and precision are evaluated the effects between before and after using the intelligent conveyer belt. According to the definition by Golan et al. (2004), breadth describes the amount of information the traceability system records, the depth of a traceability system is how far back or forward the system tracks, and precision reflects the degree of assurance with which the tracing system can pinpoint a particular food product’s movement or characteristics. The comparison result was shown in Table 5. Difference with not using the belt, the information amount was increased of additional out warehouse information and logistic information. Meanwhile, the order information was related with the product information by using the intelligent conveyer belt. Before using the belt, the information can only trace to apples material source. The intelligent equipment provided an effective method to track back to the logistic direction, because the relationship of traceability code, order ID and logistic truck number was constructed. Processing batch was always used in the existing simple traceability system. The traceability system was improved through control system in the belt comminuting with ERP. Precision of the improved
Fig. 7. Control flow of the intelligent conveyer belt between logistics and storage.
64
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
Table 5 Comparison the traceability effects between before and after using the intelligent conveyer belt. Before using the intelligent conveyer belt
After using the intelligent conveyer belt
Breadth
Record the apples material source information; Record processing information
Depth
Forward to apples material source
Precision
Identification the trace unit with a processing batch apples
Record the apples material source information; Record processing information; Record out warehouse information; Record logistic information; Relating order information Forward to apples material source; Back to the logistic direction Identification the trace unit with a box apples
traceability system with the trace unit of a box apples was higher than the existing traceability system. Therefore, it is obvious that the breath was extended, the depth was deepened and the precision was improved after using the intelligent conveyer belt. 5. Conclusions and future works QR code as one type of 2D barcode is widely used in traceable product identification. Improving the QR code readability in moving processing is the basis for implementing the continuous traceability between storage and logistic. With RSM analysis, the main factors of reading distance, belt moving speed and the interaction between reading distance and code size have the significant effect on QR code readability. A optimal reading parameters of 141.45 mm as reading distance, 34.58 mm as code size, 100 bytes as coded characters and 2.98 m/min as belt moving speed were obtained. Using the intelligent conveyer belt with the optimal parameters, the continuous traceability can be implemented with the extended breadth, deepened depth and improved precision. The study performed a attempt to optimize the reading parameters in moving processing. Although RSM is an effective method to obtain optimal values, it has some space to improve optimization effects. In the future work, improving optimization effects will be studies combined RSM and ANN. Further, the obtained parameters in this study are suitable for slow speed moving processing. In fact, high speed moving is very popular. It is necessary to develop and test a conveying-reading synchronous equipment to implement QR code reading in high speed moving processing. Acknowledgments The authors would like to thank the referees for their suggestions, which improved the content and presentation of this paper. This work was funded by the National Natural Science Foundation of China (No. 31671593) and the National Key Technology R&D Program of China (No. 2013BAD19B04). References Ampatzidis, Y., Vougioukas, S., Bochtis, D., Tsatsarelis, C., 2009. A yield mapping system for hand-harvested fruits based on RFID and GPS location technologies: field testing. Precision Agric. 10 (1), 63–72.
Astray, G., Gullón, B., Labidi, J., Gullón, P., 2016. Comparison between developed models using response surface methodology (RSM) and artificial neural networks (ANNs) with the purpose to optimize oligosaccharide mixtures production from sugar beet pulp. Ind. Crops Prod. 92, 290–299. Bas, D., Boyaci, I.H., 2007. Modeling and optimization I: Usability of response surface methodology. J. Food Eng. 78, 836–845. Bertolini, M., Bevilacqua, M., Massini, R., 2006. FMECA approach to product traceability in the food industry. Food Control 17 (2), 137–145. Bingol, D., Hercan, M., Elevli, S., Kilic, E., 2012. Comparison of the results of response surface methodology and artificial neural network for the biosorption of lead using black cumin. Biores. Technol. 112, 111–115. Cunha, C.R., Peres, E., Morais, R., Oliveira, A.A., Matos, S.G., Fermandes, M.A., Ferreira, P.J.S.G., Reis, M.J.C.S., 2010. The use of mobile devices with multi-tag technologies for an overall contextualized vineyard management. Comput. Electron. Agric. 73 (2), 154–164. Donis-González, I.R., Guyer, D.E., Pease, A., 2012. Application of Response Surface Methodology to systematically optimize image quality in computer tomography: A case study using fresh chestnuts (Castanea spp.). Comput. Electron. Agric. 87, 94–107. Golan, E., Krissoff, B., Kuchler, F., Calvin, L., Nelson, K., Price, G., 2004. Traceability in the US food supply: Economic theory and industry studies. Agric. Econ. Rep. 3, 1–48. Hamid, H.A., Jenidi, Y., Thiememans, W., Somerfield, C., Gomes, R.L., 2016. Predicting the capability of carboxylated cellulose nanowhiskers for the remediation of copper from water using response surface methodology (RSM) and artificial neural network (ANN) model. Ind. Crops Prod. 93, 108–120. International Organization for Standardization, 2007. Traceability in the feed and food chain: general principles and basic requirements for system design and implementation. ISO 22005:2007. International Organization for Standardization, 2006. Information technology – automatic identification and data capture techniques – QR Code 2005 bar code symbology specification. ISO/IEC18004:2006. Newland Auto-ID Tech. 2016. http://www.nlscan.com/showPro.aspx?id=118&cid= 776248. Nourbakhsh, H., Emam-Djomeh, Z., Omid, M., Mirsaeedghazi, H., Moini, S., 2014. Prediction of red plum juice permeate flux during membrane processing with ANN optimized using RSM. Comput. Electron. Agric. 102 (1), 1–9. Qian, J.P., Yang, X.T., Wu, X.M., Xing, B., Wu, B.G., Li, M., 2015. Farm and environment information bidirectional acquisition system with individual tree identification using smartphones for orchard precision management. Comput. Electron. Agric. 116, 101–108. Qian, J.P., Yang, X.T., Wu, X.M., Zhao, L., Fan, B.L., Xing, B., 2012. A traceability system incorporating 2D barcode and RFID technology for wheat flour mills. Comput. Electron. Agric. 89 (11), 76–85. Rakshit, M., Sharma, A., Saha, J., Sarkar, P.K., 2015. Optimization of soaking condition of blackgram to minimize flatogenic sugar content in blackgram-based products. LWT – Food Sci. Technol. 63 (2), 814–820. Regattieri, A., Gamberi, M., Manzini, R., 2007. Traceability of food products: General framework and experimental evidence. J. Food Eng. 81 (2), 347–356. Ruiz-Garcia, L., Lunadei, L., 2011. The role of RFID in agriculture: Applications, limitations and challenges. Comput. Electron. Agric. 79 (1), 42–50. So-In, C., Poolsanguan, S., Rujirakul, K., 2014. A hybrid mobile environmental and population density management system for smart poultry farms. Comput. Electron. Agric. 109 (11), 287–301. Steinberger, G., Rothmund, M., Auernhammer, H., 2009. Mobile farm equipment as a data source in an agricultural service architecture. Comput. Electron. Agric. 65 (2), 238–246. Tarjan, L., Senk, I., Tegeltija, S., Stankovski, S., Ostojic, G., 2014. A readability analysis for QR code application in a traceability system. Comput. Electron. Agric. 109 (11), 1–11. Voulodimos, A.S., Patrikakis, C.Z., Sideridis, A.B., Ntafis, V.A., Xylouri, E.M., 2010. A complete farm management system based on animal identification using RFID technology. Comput. Electron. Agric. 70 (2), 380–388. Wang, C., Wang, H., Liu, Y., Huang, L., 2016a. Optimization of surface treatment for flotation separation of polyvinyl chloride and polyethylene terephthalate waste plastics using response surface methodology. J. Clean. Prod. 139 (15), 866–872. Wang, G., Deng, Y., Xu, X., He, X., Zhao, Y., Zou, Y., Liu, Z., Yue, J., 2016b. Optimization of air jet impingement drying of okara using response surface methodology. Food Control 59, 743–749. Yang, X.T., Qian, J.P., Li, J., Ji, Z.T., Fan, B.L., Xing, B., Li, W.Y., 2016. A real-time agrofood authentication and supervision system on a novel code for improving traceability credibility. Food Control 66, 17–26.
Contents lists available at ScienceDirect
Computers and Electronics in Agriculture journal homepage: www.elsevier.com/locate/compag
Original papers
Optimization of QR code readability in movement state using response surface methodology for implementing continuous chain traceability Jianping Qian a,b, Xiaowei Du a,b, Baoyan Zhang c, Beilei Fan a,b, Xinting Yang a,b,⇑ a
National Engineering Research Center for Information Technology in Agriculture, Beijing Academy of Agriculture and Forestry Science, Beijing 100097, China National Engineering Laboratory for Agri-product Quality Traceability, Beijing 100097, China c Tianjin Rural Affairs Committee Information Center, Tianjin 3300061, China b
a r t i c l e
i n f o
Article history: Received 6 May 2016 Received in revised form 9 May 2017 Accepted 11 May 2017
Keywords: QR code Traceability Response surface methodology Barcode readability Optimization
a b s t r a c t Logistics and storage is the main processing for agro-food supply chain. Because of disconnection information between the two processing, it is difficult to trace continuously. An intelligent conveyer belt provides an effective method to associate storage and logistics by QR code scanning and information recording. Improving the QR code readability in movement state is the core of implementing continuous chain traceability with this belt. In this paper, a intelligent conveyer belt including automatic conveyer unit, barcode scanning unit, fault remove unit and control display unit was designed. Four factors affected QR readability were selected and the value range was confirmed, which was reading distance, code size, coded characters and belt moving speed. Based on the belt, an Central Composite Inscribed (CCI) experiment of four factors with five levels was designed using Response Surface Methodology (RSM) to obtain the optimal reading parameters. The result shows that the main factors of reading distance, belt moving speed and the interaction between reading distance and code size have the significant effect on QR code readability. Under the optimization condition of 141.45 mm reading distance, 34.58 mm code size, 100 bytes coded characters and 2.98 m/min belt moving speed, the average value of QR code readability was 95%. With the optimization parameters, the intelligent conveyer belt was used in an apple marketing enterprise. The result shows that the continuous traceability between storage and logistic can be implemented with the extended breadth, deepened depth and improved precision. Ó 2017 Elsevier B.V. All rights reserved.
1. Introduction Traceability is an effective method to ensure food safety and quality and to reduce the costs associated with recalls (Regattieri et al., 2007; Yang et al., 2016). In the last three decades, some astounding events, such as the BSE crisis and the problems posed by foods ingredients from genetically modified (GM) crops, have strongly focused attention on the topic of agro-food traceability (Bertolini et al., 2006). The ISO 22005:2007 food traceability standard requires that each company knows both its suppliers and customers, based on the principle of one-up and one-down (International Organization for Standardization, 2007). To achieve traceability, traceable unit identification and information record are very important. Identification technologies such as barcode and Radio Frequency Identification (RFID) as
⇑ Corresponding author at: National Engineering Research Center for Information Technology in Agriculture, Beijing Academy of Agriculture and Forestry Science, Beijing 100097, China. E-mail address: [email protected] (X. Yang). http://dx.doi.org/10.1016/j.compag.2017.05.009 0168-1699/Ó 2017 Elsevier B.V. All rights reserved.
distinguishing traceable units method can be integrated into a supply chain system (Cunha et al., 2010; Ruiz-Garcia and Lunadei, 2011). Ampatzidis et al. (2009) evaluated using RFID and barcode technologies in manual fruit harvesting to improve traceability. A platform for livestock management was also formulated by deploying mobile computing, RFID technology and wireless/mobile networking (Voulodimos et al., 2010). With the development of mobile communication technology, the information collecting and uploading time using portable devices (mobile phone, PDA, tablet PC) has become an effective means for farming operation information collection (So-In et al., 2014; Qian et al., 2015). Steinberger et al. (2009) developed mobile farming information collection equipment that transmitted the information to the server through the internet. Under the framework of from farm to table, it including many supply chain nodes, such as product, storage, logistic and sale. In the inner of supply chain enterprise, it is convenient to trace continuously through traceable unit identification and information record. For example, Qian et al. (2012) describe a study with a primary goal to develop a Wheat Flour Milling Traceability System
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
(WFMTS), incorporating 2D barcode and RFID technologies, and to validate the system in a wheat flour mill enterprise in China. Actually, because of disarticulated traceable unit identification and absent information record, it is difficult to trace continuously between supply chain nodes. The existing studies and applications provide a good reference for constructing traceability system framework or implementing supply chain inner traceability. But implementing continuous traceability between supply chain nodes faces some challenges with the demand of efficiency: (1) Transformation information can be recorded synchronously with the traceable unit identification; (2) High intelligent level can improve efficiency and reduce error. A intelligent conveyer belt including automatic conveyer unit, barcode scanning unit, fault remove unit and control display unit provided a effective solution for continuous traceability between storage and logistics. Difference with the static state (Tarjan et al., 2014), optimization of barcode reading in movement state with the intelligent conveyer belt need to further study. Response Surface Methodology (RSM) is a compilation of mathematical and statistical methods suitable to optimize, develop and improve processes (Bingol et al., 2012). In this paper, RSM was used for obtaining the optimal reading parameters and the index of breath, depth, precision was evaluated for continuous traceability effects. 2. QR code and the intelligent conveyer belt 2.1. QR code QR code is a two-dimensional barcode defined by the industrial standard ISO/IEC18004:2006 (International Organization for Standardization, 2006). QR code structure is shown in Fig. 1. Each QR code is structured by dark (logical ‘‘1”) and light (logical ‘‘0”) modules. The modules are evenly distributed in a square net of fields, where the size of a field is the size of a single module. By the standard ISO/IEC18004 one module should be sized 4 4 px (pixels) with the print resolution of 300 dpi (dots per inch). This size ensures readability by the majority of optical devices. 2.2. Intelligent conveyer belt Common conveyer belts have the main function of convey goods and less for recording information. The modified intelligent conveyer belt developed by our research team consists of automatic conveyer unit, barcode scanning unit, fault remove unit and control display unit. The structure of the belt is shown in Fig. 2.
57
The speed of conveyer unit can be adjusted from 0 m/min to 3.5 m/min. NSL-FM420 scanning module (Newland Auto-ID Tech., 2016) by Newland Auto-ID Technology was used as barcode scanning unit. The module supports QR code scanning and identification of mobile barcode. In order to fitting to scan from different directions of the packages, a holder with six free degrees was designed to support the barcode scanning module. Fault remove module adopted a pneumatic mode, easy to eliminate large packaging. If the product on the conveyer belt was not accord with the product list of order, the product was removed. A industrial personal computer was used as control display unit. The unit can control the other units according to the business and display the device status and order list. The device was also integrated with the WiFi module, which can realize wireless data transmission from the remote system, such as (Enterprise Resource Planning (ERP) system.
3. Materials and method 3.1. Factors selection and value range A QR code consists of black modules (square dots) arranged in a square grid on a white background, which can be read by an imaging device (such as a camera and scanner) and processed using Reed–Solomon error correction until the image can be appropriately interpreted. The required data are then extracted from patterns that are present in both horizontal and vertical components of the image. The amount of data that can be stored in the QR code symbol depends on encoding modes (numeric, alphanumeric, byte/ binary, and kanji), version (1, . . . , 40, indicating the overall dimensions of the symbol), and error correction level (L, M, Q, H). According to the test result in static sate and fixed distance (15 cm) with different mobile phones by Tarjan et al. (2014), QR code readability is not directly influenced by the number of coded characters, or by the error correction level, but by the size of modules. Since the data storage is not so many and the error correction level is not very high for traceable encoding requirement, it is suitable with the fixed value of 7 version and M error correction level in this research. In order to test the readability with different distance, the factor of reading distance was selected. Code size was a test factor corresponding with module size under the set printer resolution. Although the coded characters not affects directly the reading rate under fixed distance, the factor also was considered under condition of variable distance and moving processing in this research. Difference with static state, moving speed was the important factor
Fig. 1. QR code structure (International Organization for Standardization, 2006).
58
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
Fig. 2. Structure diagram of the intelligent conveyer belt.
in motion process. Four factors were selected and a preliminary laboratory trial which objective was to obtain factors value range, was presented. The main factors and value range was listed on the below. (1) Reading distance Reading distance is the important factor, since every reader has different reading distance and it has a good reading rate in the special distance. For the selected scanning module, the technology handbook shows it has a reading range from 35 to 210 mm. (2) Code size Suitable size label is not only benefit to read but also reasonable to paste in the package. For actual agro-food traceability label, the size of 40 60 mm and 60 90 mm are usually applied. Therefore, the maximum value of QR code size sets to 60 60 mm. If the code size is less than 10 10 mm, it is difficult to read. (3) Coded characters QR code has bigger characters capacity than 1D barcode. For implementing traceability, some important information such as traceability ID, producer, date and check result, are encoded to the code. Other information can be queried through loading to the database with traceability ID. The encoded information is about 20–100 characters.
or in combination, and their interactions on response variables in a process allowing develop, improve and optimize such process (Bas and Boyaci, 2007). Moreover, RSM determines the effect of independent variables on the process and generates a mathematical model that accurately describes the process (Wang et al., 2016a). RSM can be helpful to quantitatively and routinely adjust parameters that influence QR code readability. Central Composite Design (CCD) and Box-Behnken Design are always used in RSM (DonisGonzález et al., 2012). For CCD, there are three types which is Central Composite Circumscribed (CCC), Central Composite Inscribed (CCI) and Central Composite Face-centered (CCF) according to different value (Wang et al., 2016b). Generally, RSM includes three steps. First step is to confirm a approximate range included optimum condition. Second step is to establish the relationship model between response and the set of independent factors. Third step is to optimize the process using the established model. (2) Experiment design According to the preliminary trial results, a CCI design of four numerical factors (reading distance, code size, coded characters and belt moving speed) with five levels. As a rule of CCI experiment, six repetitions at the center point, sixteen repetitions at the cube point and eight repetitions at the axial point were used in this study. With a a value of 0.5, the factors and its levels were shown in Table 1. (3) Data analysis
(4) Belt moving speed With the modified conveyer belt, moving speed can be adjusted. Through adjusting the belt speed with different gradient in the preliminary trial, it was found that the reading rate has a drastic decline in case of exceeding 3.5 m/min. On the other side, if the speed is less than 2.0 m/min, it affects the operating efficiency. The moving speed is set from 2.0 to 3.5 m/min. 3.2. Optimization experiment with RSM (1) RSM RSM, originally described by Box and Wilson, is a statistical tool that enables to assess the effect of the independent variables, alone
Every experiment point was repeated twenty times and the average of QR code readability acted as response value. The experimental results of the CCI were fitted with a second-order polynomial equation by a multiple regression technique. The quadratic model for predicting the optimal point was expressed as follow:
Y ¼ b0 þ
k X i¼1
bi vi þ
k k1 X k X X bii vi vi þ bij vi vj þ eij i¼1
ð1Þ
i¼1 j¼jþ1
where Y is the response value, b0 ,bi , bii , bij are regression coefficients for intercept, linear, quadratic and interaction coefficients, respectively, vi and vj are the independent variables and k is number of factors.
59
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64 Table 1 Factors levels coding. Levels
Factors
Upper cube point Upper axial point Center point (0) Lower axial point Lower cube point
(1) (0.5) (+0.5) (+1)
Reading distance (mm)
Code size (mm)
Coded characters (byte)
Belt moving speed (m/min)
35 78.75 122.5 166.25 210
10 22.5 35 47.5 60
20 40 60 80 100
2.0 2.375 2.75 3.125 3.5
The data were statistically analyzed using Design Expert 8.06 (Stat-Ease, Inc., Minneapolis, Minnesota) for optimizing reading parameters. Significant effects of independent variables on each response was determined by ANOVA. The multiple linear regression analysis of the experimental data yielded the second order polynomial models for predicting QR code readability.
QR code readability ¼ þ105:89060 þ 0:18347X 1 þ 1:94653X 2 1:01801X 3 þ 2:51080X 4 þ 8:71429E 003X 1 X 2 8:92857E 005X 1 X 3 þ 1:4286E 002X 1 X 4 1:5625E 003X 2 X 3 þ 0:25000X 2 X 4 3:1250E 002X 3 X 4
4. Results and discussion The experiments were performed according to the design matrix in Table 2. The statistic test factor, F-Value, was used to evaluate the significance of the model at the 95% confidence level. The results of quadratic model in the form of ANOVA are given in Table 3. The F-value of 16.77 implied the model is significant. There is only a 0.01% chance that F-value could occur due to noise. Also, values of P less than 0.05 indicated model terms were significant. The Lack of Fit F-value of 0.85 implied the lack of fit is not significant relative to the pure error. There is a 61.32% chance that a Lack of Fit F-value could occur due to noise. Non-significant lack of fit indicated the model have good fit. Through test data regression, the quadratic model in terms of actual factors was expressed as follows:
Table 2 Designed matrix by RSM and obtained value of response. Run order
Reading distance (mm)
Code size (mm)
Coded characters (byte)
Belt moving speed (m/min)
QR code readability (%)
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
122.5 210 122.5 35 122.5 210 166.25 35 210 210 210 122.5 122.5 122.5 210 122.5 35 122.5 122.5 210 122.5 122.5 35 35 35 35 35 122.5 78.75 210
35 60 35 60 47.5 10 35 10 10 60 10 35 35 35 60 35 60 35 22.5 10 35 35 10 10 60 10 60 35 35 60
80 100 60 20 60 20 60 100 100 20 100 60 60 60 20 60 20 60 60 20 60 40 20 100 100 20 100 60 60 100
2.75 2 2.75 3.5 2.75 3.5 2.75 2 3.5 2 2 2.375 3.125 2.75 3.5 2.75 2 2.75 2.75 2 2.75 2.75 2 3.5 3.5 3.5 2 2.75 2.75 3.5
95 90 95 0 90 25 85 100 20 90 75 95 80 90 55 75 35 70 70 55 95 90 100 40 10 35 30 80 85 45
1:83538E 003X 21 5:4483E 002X 22 þ 9:96743E 003X 23 7:20375X 24
ð2Þ
4.1. Comparison of RSM method and ANN method ANN have high learning ability and capability of identifying and modeling the complex non-linear relationships between the input and output of a system (Nourbakhsh et al., 2014; Hamid et al., 2016). In order to compare RSM method and ANN method, an ANN model with three layers was employed. As the structure shown in Fig. 3, there were four neurons in the input layer including reading distance, code size, coded characters and belt moving speed. The hidden layer contained 10 neurons. QR code readability was used as only one output neuron. From input layer to hidden layer, Levenberg-Marquardt backpropagation (Trainlm Function) was used in Matlab R2014a. In 30 groups dataset, 20 groups were
Table 3 Comparison of RSM and ANN method using R2, RMSE and APD value. Method
R2
RMSE
APD (%)
RSM ANN
0.9329 0.9021
9.5060 15.0182
20.6790 10.0124
Fig. 3. ANN structure for predicting QR code readability.
60
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
a. Relation diagram between reading distance and QR code readability
b. Relation diagram between belt moving speed and QR code readability
c. Response surface figure between reading distance, code size and QR code readability Fig. 4. Relation diagram between significant factors and QR code readability.
61
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64 Table 4 Variance analysis of regression equation for QR code readability. Source
Sum of squares
DOF
Mean square
F-value
P-value
Model X1: Reading distance X2: Code size X3: Coded characters X4: Belt moving speed X1 X2 X1 X3 X1 X4 X2 X3 X2 X4 X3 X4 X21 X22 X23 X24 Residual Lack of Fit Pure Error Cor Total
24172.18 668.18 437.88 18.56 7530.68 5814.06 1.56 14.06 39.06 351.56 14.06 32.86 192.97 42.33 2.73 1544.49 973.65 570.83 25716.67
14 1 1 1 1 1 1 1 1 1 1 1 1 1 1 15 10 5 29
1726.58 668.18 437.88 18.56 7530.68 5814.06 1.56 14.06 39.06 351.56 14.06 32.86 192.97 42.33 2.73 102.97 97.37 114.17
16.77 6.49 4.25 0.18 73.14 56.47 0.02 0.14 0.38 3.41 0.14 0.32 1.87 0.41 0.03
<0.0001 0.0223 0.0570 0.6772 <0.0001 <0.0001 0.9036 0.7169 0.5472 0.0844 0.7169 0.5805 0.1912 0.5311 0.8728
0.85
0.6132
selected randomly to train, 5 groups to validate, and the other 5 groups to test. Coefficient of determination (R2) is important to describe the data relationship. A higher value for R2 indicated a good fit for the predicted model (Rakshit et al., 2015). Exception of R2, it also needs a small error between actual data and predicted data. In this research, Root Mean Square Error (RMSE) (Eq. (3)), and Average Percentage Deviations (APD) (Eq. (4)) were used to check the fits for the two methods. The RMSE and APD are defined as follow (Astray et al., 2016):
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi PN 2 i¼1 ðyp ye Þ RMSE ¼ N
APD ¼
PN yp ye I¼1 ye 100 N
ð3Þ
ð4Þ
where N is the number of experiment group, yp is the predicted value, ye is the experimental value. The R2, RMSE and APD for the RSM and ANN were listed in Table 3. The R2 and RMSE for the RSM method were calculated to be 0.9329 and 9.5060, whilst that of the ANN method was 0.9021 and 15.0182. In the view of the two evaluation values, it showed that RSM method predicts more accurately than the ANN method. However, the APD value of RSM with 20.6790 is higher than that of ANN with 10.0124, which indicated that ANN method has a good precision. Although not all the evaluation indexed of RSM method was superior to ANN method, the main parameters of RSM method was dominant. Therefore, it is feasible that employing the RSM method optimized QR code readability in movement state.
The relation diagram between reading distance and QR code readability by maintaining other factors constant at middle value are shown in Fig. 4a. It was obvious that the QR code readability increased with increasing in reading distance from 35 to 140 mm. However, further increase in reading distance more than 140 mm leaded to the decline on QR code readability. It demonstrated that too short distance or too long distance was not suitable for reading. Fig. 4b shows that relation between belt moving speed and QR code readability by maintaining other factors constant at middle value. The decline tendency of QR code readability was maintained with the bet moving speed increasing. When the speed was more than 3.5 m/min, the readability was less than 61%. This is due to the fact that the reading time of every barcode was reduced with the increased speed. Fig. 4c shows the strong interaction of reading distance and code size. The variation of QR code readability with reading distance had the tendency of ascend and then descend, as cited in the one factor analysis. With the increase of code size, QR code readability observed to be increased and then decreased. It also indicated that QR code had a good reading rate in the range of middle value combination of code size and reading distance. The relationship of code size and reading distance was further analyzed with Fig. 5. Barcode scanners have a limited reading angle and distance. In the readability range, the scope enlarged with the increased distance. Therefore, it is difficult that a big size was read with a short distance because of absent the whole barcode image.
4.2. Influence factors analysis Table 4 and Eq. (2) show that the influence degree of different factors were listed belt moving speed > reading distance > code size > coded characters and main factors of reading distance, belt moving speed have the significant effects on the response (P < 0.05). There was a very significant interaction between reading distance and code size, which was the important effect of QR code readability. The relation between main factors and response can be visualized by plots which were obtained from model graph tool of the Design Expert software (Fig. 4).
Fig. 5. Sketch diagram of reading range in different distance with a constant size code.
62
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
a. Intelligent conveyer belt actual picture
Main menu
Current
state:
Scanning barcodes Product order ID Traceable ID Production name Production specification Operation time
b. Control system Fig. 6. Intelligent conveyer belt actual picture and its control system.
The low readability in the case of code size with 60 mm and reading distance with 35 mm proved the situation. 4.3. Optimization readability parameters and validation In order to obtain optimized parameters for high readability, optimization method provided by Design Expert 8.06 was used. The lower and upper bound of reading distance, code size and coded characters were respectively given 35–210 mm, 10–60 mm and 20–100 bytes. Improving the belt moving speed can improve the
efficiency for the whole system, so maximized the speed was as a subordinate goal. A group of optimal solution with a theoretical readability of 97.03% was obtained with 141.45 mm as reading distance, 34.58 mm as code size, 100 bytes as coded characters and 2.98 m/min as belt moving speed. For round numbers, 20 tests were performed under the condition of 141 mm, 35 mm, 100 bytes and 3.00 m/min. The average value of QR code readability was 95% and the relative deviation was 2.1%. It shows that reading rate was high and efficiency was good with optimal readability parameter.
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
4.4. Continuous traceability between logistics and storage Logistics and storage is the main processing for agro-food supply chain. Because of disconnection information between the two processing, it is difficult to trace continuously. The intelligent conveyer belt focuses on the entry and out warehouse of the joint of logistics and storage. It have the function of goods conveying, barcode scanning, information checking and abnormal eliminating. A control system was developed and embedded in the control display unit (Fig. 6). With the system, a control flow was established as seen in Fig. 7. The intelligent conveyer belt can exchange data with ERP system. When some products need to out of the warehouse, a electronic product order was sent to the belt control system from the ERP. The control system analyzed the order and generated the different product name, specification and quantity. The workers can see the product list in the screen and find the product in the storage. The belt was deployed in the exit of the warehouse. It was operated easily between storage and logistic trucks. When the workers put the product with traceable barcode into the belt, the belt moved with the optimization speed. The barcode scanner was triggered and read the barcode in the moving processing if the product entered the reading range. A extracting procedure was performed that product information can be obtain according to the traceable barcode. If the extracted information meet with the product list, delivery products quantity added one and information was recorded. If the meet was not success, the product was eliminated. The second judge was carried out for confirming
63
the delivery products quantity. If the quantity reached the order requirement, the next product was continued to scan. When all the products of the order were achieved, the information was uploaded to ERP. The intelligent conveyer belt was applied in a apple marketing enterprise. The enterprise has been used ERP and the simple traceability system. Continuous traceability was tested with the optimization readability parameters on the intelligent conveyer belt platform. Three parameters which is traceability breadth, depth and precision are evaluated the effects between before and after using the intelligent conveyer belt. According to the definition by Golan et al. (2004), breadth describes the amount of information the traceability system records, the depth of a traceability system is how far back or forward the system tracks, and precision reflects the degree of assurance with which the tracing system can pinpoint a particular food product’s movement or characteristics. The comparison result was shown in Table 5. Difference with not using the belt, the information amount was increased of additional out warehouse information and logistic information. Meanwhile, the order information was related with the product information by using the intelligent conveyer belt. Before using the belt, the information can only trace to apples material source. The intelligent equipment provided an effective method to track back to the logistic direction, because the relationship of traceability code, order ID and logistic truck number was constructed. Processing batch was always used in the existing simple traceability system. The traceability system was improved through control system in the belt comminuting with ERP. Precision of the improved
Fig. 7. Control flow of the intelligent conveyer belt between logistics and storage.
64
J. Qian et al. / Computers and Electronics in Agriculture 139 (2017) 56–64
Table 5 Comparison the traceability effects between before and after using the intelligent conveyer belt. Before using the intelligent conveyer belt
After using the intelligent conveyer belt
Breadth
Record the apples material source information; Record processing information
Depth
Forward to apples material source
Precision
Identification the trace unit with a processing batch apples
Record the apples material source information; Record processing information; Record out warehouse information; Record logistic information; Relating order information Forward to apples material source; Back to the logistic direction Identification the trace unit with a box apples
traceability system with the trace unit of a box apples was higher than the existing traceability system. Therefore, it is obvious that the breath was extended, the depth was deepened and the precision was improved after using the intelligent conveyer belt. 5. Conclusions and future works QR code as one type of 2D barcode is widely used in traceable product identification. Improving the QR code readability in moving processing is the basis for implementing the continuous traceability between storage and logistic. With RSM analysis, the main factors of reading distance, belt moving speed and the interaction between reading distance and code size have the significant effect on QR code readability. A optimal reading parameters of 141.45 mm as reading distance, 34.58 mm as code size, 100 bytes as coded characters and 2.98 m/min as belt moving speed were obtained. Using the intelligent conveyer belt with the optimal parameters, the continuous traceability can be implemented with the extended breadth, deepened depth and improved precision. The study performed a attempt to optimize the reading parameters in moving processing. Although RSM is an effective method to obtain optimal values, it has some space to improve optimization effects. In the future work, improving optimization effects will be studies combined RSM and ANN. Further, the obtained parameters in this study are suitable for slow speed moving processing. In fact, high speed moving is very popular. It is necessary to develop and test a conveying-reading synchronous equipment to implement QR code reading in high speed moving processing. Acknowledgments The authors would like to thank the referees for their suggestions, which improved the content and presentation of this paper. This work was funded by the National Natural Science Foundation of China (No. 31671593) and the National Key Technology R&D Program of China (No. 2013BAD19B04). References Ampatzidis, Y., Vougioukas, S., Bochtis, D., Tsatsarelis, C., 2009. A yield mapping system for hand-harvested fruits based on RFID and GPS location technologies: field testing. Precision Agric. 10 (1), 63–72.
Astray, G., Gullón, B., Labidi, J., Gullón, P., 2016. Comparison between developed models using response surface methodology (RSM) and artificial neural networks (ANNs) with the purpose to optimize oligosaccharide mixtures production from sugar beet pulp. Ind. Crops Prod. 92, 290–299. Bas, D., Boyaci, I.H., 2007. Modeling and optimization I: Usability of response surface methodology. J. Food Eng. 78, 836–845. Bertolini, M., Bevilacqua, M., Massini, R., 2006. FMECA approach to product traceability in the food industry. Food Control 17 (2), 137–145. Bingol, D., Hercan, M., Elevli, S., Kilic, E., 2012. Comparison of the results of response surface methodology and artificial neural network for the biosorption of lead using black cumin. Biores. Technol. 112, 111–115. Cunha, C.R., Peres, E., Morais, R., Oliveira, A.A., Matos, S.G., Fermandes, M.A., Ferreira, P.J.S.G., Reis, M.J.C.S., 2010. The use of mobile devices with multi-tag technologies for an overall contextualized vineyard management. Comput. Electron. Agric. 73 (2), 154–164. Donis-González, I.R., Guyer, D.E., Pease, A., 2012. Application of Response Surface Methodology to systematically optimize image quality in computer tomography: A case study using fresh chestnuts (Castanea spp.). Comput. Electron. Agric. 87, 94–107. Golan, E., Krissoff, B., Kuchler, F., Calvin, L., Nelson, K., Price, G., 2004. Traceability in the US food supply: Economic theory and industry studies. Agric. Econ. Rep. 3, 1–48. Hamid, H.A., Jenidi, Y., Thiememans, W., Somerfield, C., Gomes, R.L., 2016. Predicting the capability of carboxylated cellulose nanowhiskers for the remediation of copper from water using response surface methodology (RSM) and artificial neural network (ANN) model. Ind. Crops Prod. 93, 108–120. International Organization for Standardization, 2007. Traceability in the feed and food chain: general principles and basic requirements for system design and implementation. ISO 22005:2007. International Organization for Standardization, 2006. Information technology – automatic identification and data capture techniques – QR Code 2005 bar code symbology specification. ISO/IEC18004:2006. Newland Auto-ID Tech. 2016. http://www.nlscan.com/showPro.aspx?id=118&cid= 776248. Nourbakhsh, H., Emam-Djomeh, Z., Omid, M., Mirsaeedghazi, H., Moini, S., 2014. Prediction of red plum juice permeate flux during membrane processing with ANN optimized using RSM. Comput. Electron. Agric. 102 (1), 1–9. Qian, J.P., Yang, X.T., Wu, X.M., Xing, B., Wu, B.G., Li, M., 2015. Farm and environment information bidirectional acquisition system with individual tree identification using smartphones for orchard precision management. Comput. Electron. Agric. 116, 101–108. Qian, J.P., Yang, X.T., Wu, X.M., Zhao, L., Fan, B.L., Xing, B., 2012. A traceability system incorporating 2D barcode and RFID technology for wheat flour mills. Comput. Electron. Agric. 89 (11), 76–85. Rakshit, M., Sharma, A., Saha, J., Sarkar, P.K., 2015. Optimization of soaking condition of blackgram to minimize flatogenic sugar content in blackgram-based products. LWT – Food Sci. Technol. 63 (2), 814–820. Regattieri, A., Gamberi, M., Manzini, R., 2007. Traceability of food products: General framework and experimental evidence. J. Food Eng. 81 (2), 347–356. Ruiz-Garcia, L., Lunadei, L., 2011. The role of RFID in agriculture: Applications, limitations and challenges. Comput. Electron. Agric. 79 (1), 42–50. So-In, C., Poolsanguan, S., Rujirakul, K., 2014. A hybrid mobile environmental and population density management system for smart poultry farms. Comput. Electron. Agric. 109 (11), 287–301. Steinberger, G., Rothmund, M., Auernhammer, H., 2009. Mobile farm equipment as a data source in an agricultural service architecture. Comput. Electron. Agric. 65 (2), 238–246. Tarjan, L., Senk, I., Tegeltija, S., Stankovski, S., Ostojic, G., 2014. A readability analysis for QR code application in a traceability system. Comput. Electron. Agric. 109 (11), 1–11. Voulodimos, A.S., Patrikakis, C.Z., Sideridis, A.B., Ntafis, V.A., Xylouri, E.M., 2010. A complete farm management system based on animal identification using RFID technology. Comput. Electron. Agric. 70 (2), 380–388. Wang, C., Wang, H., Liu, Y., Huang, L., 2016a. Optimization of surface treatment for flotation separation of polyvinyl chloride and polyethylene terephthalate waste plastics using response surface methodology. J. Clean. Prod. 139 (15), 866–872. Wang, G., Deng, Y., Xu, X., He, X., Zhao, Y., Zou, Y., Liu, Z., Yue, J., 2016b. Optimization of air jet impingement drying of okara using response surface methodology. Food Control 59, 743–749. Yang, X.T., Qian, J.P., Li, J., Ji, Z.T., Fan, B.L., Xing, B., Li, W.Y., 2016. A real-time agrofood authentication and supervision system on a novel code for improving traceability credibility. Food Control 66, 17–26.