A pantograph state detection device and method based on image processing
Through the pantograph state detection device based on image processing, the pantograph's horn loss, running posture and wear status are analyzed, and the problems of low detection efficiency and poor safety in the prior art are solved, and more efficient and reliable infection are achieved. Pantograph status detection.
Patent Information
- Application Number
- CN202210388628.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-04-14
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-04-14
AI Technical Summary
In the prior art, the pantograph state detection efficiency is low and the safety is poor, and the main focus is on skateboard wear. The overall pantograph state analysis is less, which affects the detection accuracy.
The pantograph state detection device based on image processing, including a field trigger module, an image acquisition module and an image processing module, is used to analyze whether the pantograph's horns are missing, the running posture and the wear status of the skateboard through image processing technology, and provide a systematic state. Test.
It improves the accuracy and stability of inspection, reduces manual maintenance time, optimizes the maintenance process, and improves the efficiency and reliability of pantograph maintenance.
Smart Images

Figure CN114877814B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of urban rail transit safety engineering, and in particular to a pantograph state detection device and method based on image processing. Background Art
[0002] With the rapid development of urbanization, urban transportation is facing tremendous pressure and challenges. Urban rail transit has become the first choice to relieve traffic pressure in big cities with its advantages of high safety, high speed, high punctuality, large passenger flow, and strong comfort. The networked operation of rail transit has put forward higher requirements on the safety and stability of the pantograph network system. The pantograph is a key electrical equipment used to obtain power source for the train, and its status has a significant impact on the safe operation of the train. Therefore, the status monitoring of the pantograph is of great significance to reduce the probability of pantograph failure.
[0003] The detection of pantograph status is an important indicator for daily inspection of urban rail train vehicles. At present, the main method for detecting the pantograph status is still manual detection, but this method requires manual climbing to the top to detect the pantograph, which has low detection efficiency, poor safety, and the detection results are greatly affected by human factors. With the rapid development of image processing technology and machine vision measurement technology, train pantograph status detection based on image processing has become an effective method. However, in actual applications, due to factors such as the quality of flash lighting, the complex background environment of the pantograph, and the stains on the slide, the processing of the pantograph image is difficult, which in turn affects the low detection accuracy. At the same time, the existing technology focuses more on the detection of slide wear, and less on the analysis of the overall pantograph status. Summary of the invention
[0004] The object of the present invention is to provide a simple, efficient, accurate and reliable pantograph state detection device and method to realize intelligent detection of the pantograph state.
[0005] The technical solution to achieve the purpose of the present invention is: a pantograph state detection device based on image processing, including a field trigger module, an image acquisition module and an image processing module, wherein:
[0006] The on-site trigger module is used to locate the position of the pantograph slide and provide a trigger signal for the camera;
[0007] The image acquisition module acquires the image of the pantograph of the running train and transmits it to the image processing module;
[0008] The image processing module receives the real-time pantograph image, analyzes the real-time pantograph image using image processing technology, analyzes whether the pantograph horn is missing, the pantograph operating posture and the wear status of the slide plate, obtains and stores the real-time status information of the pantograph.
[0009] A pantograph state detection method based on image processing, the method adopts the pantograph state detection device based on image processing, and the specific steps are as follows:
[0010] S1: Extract the full pantograph image and detect whether the horns are missing;
[0011] S2: Extract the left and right horn boundary points and analyze the operating posture of the pantograph;
[0012] S3: Extract the pantograph slide picture, use the row grayscale and positioning algorithm and image processing technology to obtain the slide wear curve, and analyze the slide wear state;
[0013] S4: Evaluate the status of the pantograph based on the results of the ram horn detection, the pantograph operation posture and the slide plate wear status.
[0014] Compared with the prior art, the present invention has the following significant advantages:
[0015] (1) The present invention analyzes the pantograph status in a more systematic way, not only analyzing the state of the slide plate, but also analyzing the missing horn and the operating posture of the pantograph;
[0016] (2) The present invention adds a skateboard positioning link before extracting the skateboard edge, which can effectively reduce the interference of the background environment and further improve the detection accuracy;
[0017] (3) The pantograph status detection device of the present invention has good detection accuracy and stability, can reduce manual maintenance time, optimize the maintenance process, and further improve the efficiency and reliability of pantograph maintenance. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 It is a module diagram of the pantograph status detection device in the present invention.
[0019] Figure 2 It is a schematic diagram of the design of the installation distance of the magnetic steel sensor in the present invention.
[0020] Figure 3 It is a schematic diagram of the pantograph slide plate and the positioning axle in the present invention.
[0021] Figure 4 It is a schematic diagram of the installation of the image acquisition module in the present invention.
[0022] Figure 5 It is a flow chart of the pantograph status detection method in the present invention.
[0023] Figure 6 This is a diagram of the pantograph operating posture analysis results in the present invention.
[0024] Figure 7This is a precise positioning diagram of the pantograph slide plate in the present invention.
[0025] Figure 8 It is a curve diagram of the remaining wear thickness of the pantograph slide plate in the present invention.
[0026] Fig. 9 This is a diagram showing the wear variation of the pantograph slide in the present invention. DETAILED DESCRIPTION
[0027] The present invention provides a pantograph state detection device based on image processing, comprising a field trigger module, an image acquisition module and an image processing module, wherein:
[0028] The on-site trigger module is used to accurately locate the position of the pantograph slide and provide a trigger signal for the camera;
[0029] The image acquisition module is responsible for collecting the pantograph images of the running train and transmitting them to the image processing module;
[0030] The image processing module receives the real-time pantograph image, analyzes the real-time pantograph image using image processing technology, analyzes whether the pantograph horn is missing, the pantograph operating posture and the wear status of the slide plate, obtains and stores the real-time status information of the pantograph.
[0031] Furthermore, the field trigger module mainly includes a PLC and a magnetic steel sensor, wherein:
[0032] PLC is installed in the trackside control box for real-time control of on-site equipment;
[0033] The magnetic steel sensor is installed on the inner side of the rail to accurately locate the pantograph slide. One magnetic steel is used to locate one slide. When the magnetic steel sensor detects the positioning axle, the camera is triggered to take a photo. The installation distance of the magnetic steel sensor is calculated as follows:
[0034] Assuming that the distance between the target skateboard and the positioning axle is L1, the camera shooting distance is L2, and the camera lens position is taken as the starting point, along the train's forward direction, if the target skateboard is in front of the positioning axle, the installation distance of the magnetic steel sensor is L2-L1; if the skateboard is behind the positioning axle, the installation distance of the magnetic steel sensor is L2+L1. Considering the position interference problem of the actual site sleepers, it is necessary to add alternative positioning axles to calculate the installation distance.
[0035] Furthermore, the image acquisition module includes a front and a rear camera module, each group includes a centerline camera and two wear cameras, which are installed above the contact network and collect images of the pantograph in a bird's-eye view manner, wherein the centerline camera takes images of the entire bow, and the wear camera takes images of the skateboard; the front and rear camera modules, the front camera module takes images of the pantograph and the skateboard when the pantograph enters the detection area, and the rear camera module takes images of the pantograph and the skateboard when the pantograph leaves the detection area, and the front and rear camera modules cooperate with each other to complete the image acquisition of the pantograph and the assembled skateboard.
[0036] The present invention also provides a pantograph state detection method based on image processing, and the specific steps are as follows:
[0037] S1: Extract the full pantograph image and detect whether the horns are missing;
[0038] S2: Extract the left and right horn boundary points and analyze the operating posture of the pantograph;
[0039] S3: Extract the pantograph slide picture, use the row grayscale and positioning algorithm and image processing technology to obtain the slide wear curve, and analyze the slide wear state;
[0040] S4: Evaluate the status of the pantograph based on the results of the ram horn detection, the pantograph operation posture and the slide plate wear status.
[0041] Furthermore, the specific process of detecting whether the sheep's horns are missing in step S1 is as follows:
[0042] S11: Based on the camera triggering scheme and the ATO automatic driving mode of the train, the upper and lower thresholds are set according to experience to achieve the positioning of the pantograph;
[0043] S12: For the pantograph image after positioning, Gaussian filtering, Canny edge detection and Hough transform are performed in sequence. The edge straight line of the horn is detected according to the edge angle and position information of the horn. The missing horn is detected by analyzing whether the edge straight line of the horn exists. Assume that the left and right horn edge straight lines are denoted as l left , l right , straight line l left The endpoints are (u left1 ,v left1 )、(u left2 ,v left2 ), with an angle of θ left , straight line l right The endpoints are (u right1 ,v right1 )、(u right2 ,v right2 ), with an angle of θ right , the screening conditions for the horn edge straight line are as follows:
[0044]
[0045]
[0046] Furthermore, the step S2 extracts the left and right horn boundary points, and the specific process is as follows:
[0047] S21: Based on step S12, locate the sheep horns and extract the straight line l left , l right The center point of the sheep's horns is selected, and a rectangular window with fixed width and height is set with this point as the reference point to realize the positioning of the sheep's horns;
[0048] S22: After the horns are located, the boundary points of the left and right horns are detected using a corner point detection method based on a sliding window, as follows:
[0049] Extract the left horn location area Qy left The coordinates of the lower left pixel point, starting from the lower left pixel point, using a 3×3 sliding window, the current point is located at the lower left corner of the window, and scan Qy one by one from bottom to top vertically and from left to right horizontally left Pixel points, and count the number of edge points n in the sliding window. When n>3 and (i,j), (i+1,j), (i+2,j) are all edge points, stop edge point scanning and consider point (i,j) as the boundary point of the left horn, recorded as (u left3 ,v left3 );
[0050] Extract the right horn location area Qy right The coordinates of the lower right corner pixel point, starting from the lower right corner pixel point, using a 3×3 sliding window, the current point is located at the lower right corner of the window, and scan Qy one by one from bottom to top vertically and from right to left horizontally. right Pixel points, and count the number of edge points n in the sliding window. When n>3 and (i,j), (i-1,j), (i-2,j) are all edge points, stop edge point scanning and consider point (i,j) as the boundary point of the right horn, recorded as (u right3 ,v right3 ).
[0051] Furthermore, the analysis of the operating posture of the pantograph in step S2 is as follows:
[0052] Based on the left and right horn boundary points extracted in step S22, the pantograph centerline offset value, pantograph vertical height difference and pantograph horizontality are used to analyze the pantograph operation posture after camera calibration. The specific index definitions are as follows:
[0053] Centerline offset value:
[0054] Vertical height difference: P y =||y left -y right |-y_base|
[0055] Levelness:
[0056] In the formula, (x left ,y left )、(x right ,y right ) are respectively (u left3 ,v left3 )、(u right3 ,v right3 ) The coordinate values after camera calibration, x_base, y_base and k_base are respectively the centerline position of the normal posture of the pantograph, the height difference between the left and right horns and the slope of the pantograph head connection line;
[0057] When the centerline offset value, vertical height difference, and horizontality exceed the set thresholds, the pantograph posture may be abnormal.
[0058] Furthermore, the wear state of the slide plate is analyzed in step S3, and the specific process is as follows:
[0059] S31: Preliminary positioning of the target slide is achieved based on the empirical value, and then the slide positioning technology based on the row grayscale sum is used to achieve accurate positioning of the slide;
[0060] S32: extracting the wear area of the slide plate and realizing the upper and lower edges of the target slide plate by using edge detection technology;
[0061] S33: using camera calibration technology to obtain the remaining thickness curve of the slide plate in the wear area;
[0062] S34: Analyze the wear state of the slide plate according to the slide plate remaining thickness curve.
[0063] Furthermore, the step S31 uses the slide positioning technology based on the row grayscale sum to achieve accurate positioning of the slide, and the specific process is as follows:
[0064] S311: After the initial positioning of the slide, calculate the row grayscale and Y1 i , to determine whether there is a trough point in the row grayscale and curve. If there is no trough point, the slider positioning thresholds are h1 and h2. up =H-h_cut,h1 down =H, where H is the height of the image after cropping, and h_cut is a threshold designed based on the new slider height and the size of a single pixel;
[0065] S312: The curve has a trough point, and the peak function is used to filter out the points that meet the condition Y1 i ≤Y t1 The trough points are sorted in ascending order of row numbers, recorded as h1_p1,h1_p2,…,h1_p m , where Y t1 is the designed screening threshold, m is the number of trough points, and the reverse search satisfies the formula abs(h1_p k -H)>εt t1 m≥1, the first trough point of k∈{1,2,…,m} is taken as the target trough point h1_p k , where H is the distance threshold between the trough point and the image boundary, ε t1 is the deviation threshold, then the upper and lower boundaries of the slide are initially h up _tmp=h1_p k -h_cut,h down _tmp=h1_p k ;
[0066] S313: The upper and lower boundaries of the slide are precisely located. After the initial positioning of the slide, the local areas containing the boundaries are selected to calculate the row grayscale sum. The upper and lower boundaries are precisely extracted as follows:
[0067] The upper boundary of the slide, the grayscale sum of the row is Y2 i , filter to meet the condition Y2 i ≤Y t2 The data points {(x21,y21),(x22,y22)…,(x2 n ,y2 n )}, where n is the number of data points that meet the conditions, Y t2 To design the screening threshold, the upper boundary of the slide is relocated according to the following formula;
[0068]
[0069] The lower boundary of the slider, the grayscale sum of the row is Y3 i , filter to meet the condition Y3 i ≤Y t3 The data points {(x31,y31),(x32,y32)…,(x3 n ,y3 n )}, where n is the number of data points that meet the conditions, Y t3 To design the screening threshold, the lower boundary of the slide is relocated according to the following formula.
[0070]
[0071] Furthermore, the wear state of the slide plate is analyzed in step S34, as follows:
[0072] According to the remaining thickness curve of the slide plate, the minimum remaining thickness W is adopted. M and wear variation W V Analyze the state of the skateboard, and the state judgment rules are as follows:
[0073] W M =min(W(x)) <G M ,x∈[0,L s ]
[0074]
[0075] Where W(x) represents the wear value of the slide at position x, L s Indicates the length of the wear area of the slide, G M is the wear alarm threshold, G V is the alarm threshold of the change in the wear of the slide plate; x i 、x i+1 Respectively represent the i-th and i+1-th positions;
[0076] When the minimum remaining thickness of the slide plate W M Less than the wear threshold G M When , the slide plate is in an over-wear state;
[0077] When abs(W V ) is greater than the threshold G V When the wear curve of the slide plate rises or falls beyond the threshold, if the wear change W V First greater than the threshold G V , then within a certain width distance is less than G V , there is groove wear.
[0078] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.
[0079] Example
[0080] Combination Figure 1 A pantograph status detection device based on image processing is characterized in that it includes a field trigger module, an image acquisition module and an image processing module; the field trigger module is used to accurately locate the position of the pantograph slide and provide a trigger signal for the camera; the image acquisition module is responsible for collecting pictures of the pantograph of a running train and transmitting them to the image processing module; the image processing module receives and analyzes the real-time pantograph image, obtains and stores the real-time status information of the pantograph.
[0081] The on-site trigger module mainly includes a PLC and a magnetic steel sensor. The PLC is installed in the trackside control box for real-time control of on-site equipment. The magnetic steel sensor is installed on the inner side of the rail for accurate positioning of the pantograph slide. A scheme of positioning one slide with one magnetic steel is adopted. When the magnetic steel sensor detects the positioning axle, the camera is triggered to take a photo. The installation distance of the magnetic steel sensor is calculated as follows:
[0082] like Figure 2-3 As shown, taking the train skateboard 1 as an example, the positioning axle is the 7th axle, the distance between the train skateboard 1 and the positioning axle is 1150mm, the camera shooting distance is 3800mm, and the camera lens position is taken as the starting point. Along the forward direction of the train, the skateboard 1 is in front of the positioning axle, and the installation distance of the magnetic steel sensor is 2650mm. Similarly, the installation distances of the positioning magnetic steel sensors of the remaining skateboards can be calculated;
[0083] The test train is equipped with 2 pantographs, each pantograph is equipped with 2 slide plates, and the train has a total of 4 slide plates, numbered slide plate 1, slide plate 2, slide plate 3 and slide plate 4. Slide plates 1 and 3 use the 7th axle as the positioning axle, and the corresponding installation distances of magnetic steel sensors are 2650mm and 5250mm respectively. Slide plates 2 and 4 use the 17th axle as the positioning axle, and the corresponding installation distances of magnetic steel sensors are 2950mm and 4950mm respectively.
[0084] Taking into account the position interference problem of the actual on-site sleepers, it is necessary to add alternative positioning axles to calculate the installation distance. Slide plates 1 and 3 use the 8th axle as the alternative positioning axle, and the corresponding magnetic steel sensor installation distances are 4950mm and 2950mm respectively. Slide plates 2 and 4 use the 18th axle as the alternative positioning axle, and the corresponding magnetic steel sensor installation distances are 5250mm and 2650mm respectively.
[0085] The image acquisition module includes a front and back camera module, such as Figure 4 As shown, each group includes a centerline camera and two wear cameras, which are installed above the contact network and collect pantograph images by overhead shooting. The centerline camera takes pictures of the entire pantograph, and the wear camera takes pictures of the slide plate. The positive camera module takes pictures of the pantograph and the slide plate when the pantograph enters the detection area, and the negative camera module takes pictures of the pantograph and the slide plate when the pantograph leaves the detection area. The positive and negative camera modules cooperate with each other to complete the picture collection of the pantograph and its assembled slide plate.
[0086] The image processing module uses image processing technology to analyze the real-time pantograph image, and analyzes and stores whether the pantograph horn is missing, the pantograph operation posture and the wear status of the slide plate.
[0087] Combination Figure 5, a pantograph state detection method based on image processing, characterized in that the specific steps are as follows:
[0088] S1: Extract the full pantograph image and detect whether the horns are missing;
[0089] S2: Extract the left and right horn boundary points and analyze the operating posture of the pantograph;
[0090] S3: Extract the pantograph slide picture, use the row grayscale and positioning algorithm and image processing technology to obtain the slide wear curve, and analyze the slide wear state;
[0091] S4: Obtain the horn detection results, pantograph operation posture results and slide plate wear status results of the above steps, and judge the status of the pantograph.
[0092] The specific process of the sheep horn missing detection in step S1 is as follows:
[0093] S11: Based on the camera triggering scheme and the ATO automatic driving mode of the train, the upper and lower thresholds are set according to experience to realize the positioning of the pantograph, and the thresholds are set to row_up = 380 and row_down = 680;
[0094] S12: For the pantograph image after positioning, Gaussian filtering, Canny edge detection and Hough transform are performed in sequence. The edge straight line of the horn is detected according to the edge angle and position information of the horn. The missing horn is detected by analyzing whether the edge straight line of the horn exists. Assume that the left and right horn edge straight lines are denoted as l left , l right , straight line l left The endpoints are (u left1 ,v left1 )、(u left2 ,v left2 ), with an angle of θ left , straight line l right The endpoints are (u right1 ,v right1 )、(u right2 ,v right2 ), with an angle of θ right , the screening conditions for the horn edge straight line are as follows:
[0095]
[0096]
[0097] The specific process of extracting the horn boundary points in step S2 is as follows:
[0098] S21: Based on step S12, locate the sheep horns and extract the straight line lleft , l right The center point of the image is selected, and a rectangular window with fixed width and height is set with this point as the reference point to realize the positioning of the horns. The width and height of the rectangular window are set to 100 pixels.
[0099] S22: After the horns are located, the boundary points of the left and right horns are detected using a corner point detection method based on a sliding window, as follows:
[0100] Extract the left horn location area Qy left The coordinates of the pixel point in the lower left corner are taken as the starting point. A 3×3 sliding window is used. The current point is located in the lower left corner of the window. Scan Qy one by one from bottom to top vertically and from left to right horizontally. left Pixel points, and count the number of edge points n in the sliding window. When n>3 and (i,j), (i+1,j), (i+2,j) are all edge points, stop edge point scanning and consider point (i,j) as the boundary point of the left horn, recorded as (u left3 ,v left3 );
[0101] Extract the right horn location area Qy right The coordinates of the pixel point in the lower right corner. Take this point as the starting point and use a 3×3 sliding window. The current point is located in the lower right corner of the window. Scan Qy one by one from bottom to top vertically and from right to left horizontally. right Pixel points, and count the number of edge points n in the sliding window. When n>3 and (i,j), (i-1,j), (i-2,j) are all edge points, stop edge point scanning and consider point (i,j) as the boundary point of the right horn, recorded as (u right3 ,v right3 );
[0102] The extraction of the boundary points of the horns is as follows Figure 6 As shown, Figure 6 The red ☆ in the figure indicates the detected left and right horn boundary points, and their pixel coordinates are (253,223) and (1329,221) respectively.
[0103] Combination Figure 6 , the pantograph operation posture analysis described in step S2, based on the left and right horn boundary points extracted in step S22, the coordinate values after camera calibration are (870.5587, -399.3508) and (-642.8516, -409.6961) respectively, and the pantograph centerline offset value, pantograph vertical height difference and pantograph horizontality are used to analyze the pantograph operation posture, the centerline offset value Px = 3.4515mm, the vertical height difference Py = 8.6058mm, and the horizontality Pa = 0.3257°;
[0104] Perform statistical analysis on the normal pantograph posture, obtain the centerline position of the normal pantograph posture, the height difference between the left and right horns, and the slope of the pantograph head connection line, and set x_base = 117.3051 mm, y_base = 1.7391 mm, and k_base = 0.0012;
[0105] The alarm thresholds of the centerline offset value, vertical height difference and horizontality are set to 20mm, 20mm and 5° respectively, and the test pantograph posture is normal.
[0106] Combination Figure 7 to Figure 9 , the wear status of the skateboard is analyzed by taking the four skateboard images taken by the left camera in the positive camera group as an example, which are named skateboard 1, skateboard 2, skateboard 3 and skateboard 4 respectively.
[0107] The specific process of analyzing the wear state of the slide plate in step S3 is as follows:
[0108] S31: Preliminary positioning of the target slide is achieved based on the empirical value, and then the slide positioning technology based on the row grayscale sum is used to achieve accurate positioning of the slide. The empirical value is set to row_up=350, row_down=630, column_left=1000, column_right=2050;
[0109] S32: extracting the wear area of the slide plate and realizing the upper and lower edges of the target slide plate by using edge detection technology;
[0110] S33: using camera calibration technology to obtain the remaining thickness curve of the slide plate in the wear area;
[0111] S34: Analyze the wear state of the slide plate according to the slide plate remaining thickness curve.
[0112] The skateboard positioning technology based on the row grayscale sum in step S31 is specifically analyzed by taking the skateboard 1 picture as an example. The skateboard positioning result is as follows: Figure 7 The specific analysis process is as follows:
[0113] S311: After the initial positioning of the slide, calculate the row grayscale and Y1 i , to determine whether there is a trough point in the row grayscale and curve. If there is no trough point, the slider positioning thresholds are h1 and h2. up =H-h_cut,h1 down =H, where H=280 is the height of the image after cropping, and h_cut=110 is a threshold designed based on the new slider height and the size of a single pixel;
[0114] S312: The curve has a trough point, and the peak function is used to filter out the points that meet the condition Y1 i ≤Y t1The trough points are sorted in ascending order of row numbers, recorded as h1_p1,h1_p2,…,h1_p m , where Y t1 =150000 is the designed screening threshold, m is the number of trough points, and the reverse search satisfies the formula abs(h1_p k -H)>ε t1 m≥1, the first trough point of k∈{1,2,…,m} is taken as the target trough point h1_p k , where H is the distance threshold between the trough point and the image boundary, ε t1 =10 is the deviation threshold, then the upper and lower boundaries of the slide are initially h up _tmp=112,h down _tmp=222;
[0115] S313: The upper and lower boundaries of the slide are precisely located. After the initial positioning of the slide, the local areas containing the boundaries are selected to calculate the row grayscale sum. The upper and lower boundaries are precisely extracted as follows:
[0116] The upper boundary of the slide, the grayscale sum of the row is Y2 i , filter to meet the condition Y2 i ≤Y t2 The data points {(x21,y21),(x22,y22)…,(x2 n ,y2 n )}, where n is the number of data points that meet the conditions, Y t2 =15000 is the designed screening threshold, and the upper boundary of the slide is relocated according to the following formula, h up =14;
[0117]
[0118] The lower boundary of the slider, the grayscale sum of the row is Y3 i , filter to meet the condition Y3 i ≤Y t3 The data points {(x31,y31),(x32,y32)…,(x3 n ,y3 n )}, where n is the number of data points that meet the conditions, Y t3 =25000 is the designed screening threshold, and the lower boundary of the slide is relocated according to the following formula, h down =108.
[0119]
[0120] Step S34: The wear state of the slide is analyzed, and the minimum remaining thickness W is adopted according to the remaining thickness curve of the slide. M and wear variation W VThe state of the slide plate is analyzed, and the remaining thickness curves of the four slide plates used for the test are as follows: Figure 8 As shown in the figure, the minimum remaining thickness of the four slides are 24.61mm, 27.42mm, 24.77mm and 23.31mm respectively. The alarm threshold G is set. M =20mm, all four slide plates are larger than G M , then the test slide is not in an over-wear state;
[0121] The wear variation curve of the slide plate is as follows: Fig. 9 As shown, set the wear alarm threshold G V =1mm, the wear variation of the test slide is less than G V , then there is no groove wear on the test slide.
[0122] The results of the ram horn detection, the pantograph operating posture and the wear status of the slide plate are collected to judge the status of the pantograph.
[0123] The other structures of the pantograph state detection device and method based on image processing described in the present invention can be found in the prior art and will not be described in detail here.
[0124] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Therefore, any modification, equivalent change and modification made to the above embodiment according to the technical essence of the present invention without departing from the technical solution of the present invention shall still fall within the scope of the technical solution of the present invention.
Claims
1. A pantograph state detection device based on image processing, characterized in that: It includes a field trigger module, an image acquisition module and an image processing module, wherein: The on-site trigger module is used to locate the position of the pantograph slide and provide a trigger signal for the camera; The image acquisition module acquires the image of the pantograph of the running train and transmits it to the image processing module; The image processing module receives the real-time pantograph image, analyzes the real-time pantograph image using image processing technology, analyzes whether the pantograph horn is missing, the pantograph operation posture and the wear state of the slide plate, obtains the real-time state information of the pantograph and stores it; The detection method of the detection device is as follows: S1: Extract the full pantograph image and detect whether the horns are missing; S2: Extract the left and right horn boundary points and analyze the operating posture of the pantograph; S3: Extract the pantograph slide picture, use the row grayscale and positioning algorithm and image processing technology to obtain the slide wear curve, and analyze the slide wear state; S4: judging the status of the pantograph according to the results of the horn detection, the pantograph operation posture and the wear status of the slide plate; The specific process of analyzing the wear state of the slide plate in step S3 is as follows: S31: Preliminary positioning of the target slide is achieved based on the empirical value, and then the slide positioning technology based on the row grayscale sum is used to achieve accurate positioning of the slide; S32: extracting the wear area of the slide plate and realizing the upper and lower edges of the target slide plate by using edge detection technology; S33: using camera calibration technology to obtain the remaining thickness curve of the slide plate in the wear area; S34: analyzing the wear state of the slide plate according to the remaining thickness curve of the slide plate; Step S31 uses the slide positioning technology based on the row grayscale sum to achieve accurate positioning of the slide, and the specific process is as follows: S311: After the initial positioning of the slide, calculate the row grayscale and Y1 i , to determine whether there is a trough point in the row grayscale and curve. If there is no trough point, the slider positioning thresholds are h1 and h2. up =H-h_cut,h1 down =H, where H is the height of the image after cropping, and h_cut is a threshold designed based on the new slider height and the size of a single pixel; S312: The curve has a trough point, and the peak function is used to filter out the points that meet the condition Y1 i ≤Y t1 The trough points are sorted in ascending order of row numbers, recorded as h1_p1,h1_p2,…,h1_p m , where Y t1 is the designed screening threshold, m is the number of trough points, and the reverse search satisfies the formula abs(h1_p k -H)>ε t1 m≥1, the first trough point of k∈{1,2,…,m} is taken as the target trough point h1_p k , where H is the distance threshold between the trough point and the image boundary, ε t1 is the deviation threshold, then the upper and lower boundaries of the slide are initially h up _tmp=h1_p k -h_cut,h down _tmp=h1_p k ; S313: The upper and lower boundaries of the slide are precisely located. After the initial positioning of the slide, the local areas containing the boundaries are selected to calculate the row grayscale sum. The upper and lower boundaries are precisely extracted as follows: The upper boundary of the slide, the grayscale sum of the row is Y2 i , filter to meet the condition Y2 i ≤Y t2 The data points {(x21,y21),(x22,y22)…,(x2 n ,y2 n )}, where n is the number of data points that meet the conditions, Y t2 To design the screening threshold, the upper boundary of the slide is relocated according to the following formula; The lower boundary of the slider, the grayscale sum of the row is Y3 i , filter to meet the condition Y3 i ≤Y t3 The data points {(x31,y31),(x32,y32)…,(x3 n ,y3 n )}, where n is the number of data points that meet the conditions, Y t3 To design the screening threshold, the lower boundary of the slide is relocated according to the following formula: Step S34 analyzes the wear state of the slide plate, specifically as follows: According to the remaining thickness curve of the slide plate, the minimum remaining thickness W is adopted. M and wear variation W V Analyze the state of the skateboard, and the state judgment rules are as follows: W M =min(W(x))<G M ,x∈[0,L s ] Where W(x) represents the wear value of the slide at position x, L s Indicates the length of the wear area of the slide, G M is the wear alarm threshold, G V is the alarm threshold of the wear change of the slide plate; x i 、x i+1 Respectively represent the i-th and i+1-th positions; When the minimum remaining thickness of the slide plate W M Less than the wear threshold G M When , the slide plate is in an over-wear state; When abs(W V ) is greater than the threshold G V When the wear curve of the slide plate rises or falls beyond the threshold, if the wear change W V First greater than the threshold G V , then within a certain width distance is less than G V , there is groove wear.
2. The pantograph state detection device based on image processing according to claim 1, characterized in that: The field trigger module includes a PLC and a magnetic steel sensor, wherein: The PLC is installed in the trackside control box and is used for real-time control of on-site equipment; The magnetic steel sensor is installed on the inner side of the rail to locate the pantograph slide. One magnetic steel is used to locate one slide. When the magnetic steel sensor detects the positioning axle, the camera is triggered to take a photo. The installation distance of the magnetic steel sensor is calculated as follows: Assuming that the distance between the target skateboard and the positioning axle is L1, and the camera shooting distance is L2, taking the camera lens position as the starting point, along the train's forward direction, if the target skateboard is in front of the positioning axle, the installation distance of the magnetic steel sensor is L2-L1; if the skateboard is behind the positioning axle, the installation distance of the magnetic steel sensor is L2+L1.
3. The pantograph state detection device based on image processing according to claim 1, characterized in that: The image acquisition module includes a front camera module and a rear camera module, each of which includes a centerline camera and two wear cameras, which are installed above the contact network and collect images of the pantograph in a bird's-eye view manner, wherein the centerline camera takes images of the entire pantograph, and the wear camera takes images of the skateboard; the front camera module takes images of the pantograph and the skateboard when the pantograph enters the detection area, and the rear camera module takes images of the pantograph and the skateboard when the pantograph leaves the detection area. The front and rear camera modules cooperate with each other to complete the image acquisition of the pantograph and the assembled skateboard.
4. A pantograph state detection method based on image processing, characterized in that: The method adopts the pantograph state detection device based on image processing as claimed in any one of claims 1 to 3, and the specific steps are as follows: S1: Extract the full pantograph image and detect whether the horns are missing; S2: Extract the left and right horn boundary points and analyze the operating posture of the pantograph; S3: Extract the pantograph slide picture, use the row grayscale and positioning algorithm and image processing technology to obtain the slide wear curve, and analyze the slide wear state; S4: judging the status of the pantograph according to the results of the horn detection, the pantograph operation posture and the wear status of the slide plate; The specific process of analyzing the wear state of the slide plate in step S3 is as follows: S31: Preliminary positioning of the target slide is achieved based on the empirical value, and then the slide positioning technology based on the row grayscale sum is used to achieve accurate positioning of the slide; S32: extracting the wear area of the slide plate and realizing the upper and lower edges of the target slide plate by using edge detection technology; S33: using camera calibration technology to obtain the remaining thickness curve of the slide plate in the wear area; S34: analyzing the wear state of the slide plate according to the remaining thickness curve of the slide plate; Step S31 uses the slide positioning technology based on the row grayscale sum to achieve accurate positioning of the slide, and the specific process is as follows: S311: After the initial positioning of the slide, calculate the row grayscale and Y1 i , to determine whether there is a trough point in the row grayscale and curve. If there is no trough point, the slider positioning thresholds are h1 and h2. up =H-h_cut,h1 down =H, where H is the height of the image after cropping, and h_cut is a threshold designed based on the new slider height and the size of a single pixel; S312: The curve has a trough point, and the peak function is used to filter out the points that meet the condition Y1 i ≤Y t1 The trough points are sorted in ascending order of row numbers, recorded as h1_p1,h1_p2,…,h1_p m , where Y t1 is the designed screening threshold, m is the number of trough points, and the reverse search satisfies the formula abs(h1_p k -H)>ε t1 m≥1, the first trough point of k∈{1,2,…,m} is taken as the target trough point h1_p k , where H is the distance threshold between the trough point and the image boundary, ε t1 is the deviation threshold, then the upper and lower boundaries of the slide are initially h up _tmp=h1_p k -h_cut,h down _tmp=h1_p k ; S313: The upper and lower boundaries of the slide are precisely located. After the initial positioning of the slide, the local areas containing the boundaries are selected to calculate the row grayscale sum. The upper and lower boundaries are precisely extracted as follows: The upper boundary of the slide, the grayscale sum of the row is Y2 i , filter to meet the condition Y2 i ≤Y t2 The data points {(x21,y21),(x22,y22)…,(x2 n ,y2 n )}, where n is the number of data points that meet the conditions, Y t2 To design the screening threshold, the upper boundary of the slide is relocated according to the following formula; The lower boundary of the slider, the grayscale sum of the row is Y3 i , filter to meet the condition Y3 i ≤Y t3 The data points {(x31,y31),(x32,y32)…,(x3 n ,y3 n )}, where n is the number of data points that meet the conditions, Y t3 To design the screening threshold, the lower boundary of the slide is relocated according to the following formula: Step S34 analyzes the wear state of the slide plate, specifically as follows: According to the remaining thickness curve of the slide plate, the minimum remaining thickness W is adopted. M and wear variation W V Analyze the state of the skateboard, and the state judgment rules are as follows: W M =min(W(x))<G M ,x∈[0,L s ] Where W(x) represents the wear value of the slide at position x, L s Indicates the length of the wear area of the slide, G M is the wear alarm threshold, G V is the alarm threshold of the wear change of the slide plate; x i 、x i+1 Respectively represent the i-th and i+1-th positions; When the minimum remaining thickness of the slide plate W M Less than the wear threshold G M When , the slide plate is in an over-wear state; When abs(W V ) is greater than the threshold G V When the wear curve of the slide plate rises or falls beyond the threshold, if the wear change W V First greater than the threshold G V , then within a certain width distance is less than G V , there is groove wear.
5. The pantograph state detection method based on image processing according to claim 4, characterized in that: Step S1 is to detect whether the sheep's horns are missing, and the specific process is as follows: S11: Based on the camera triggering scheme and the ATO automatic driving mode of the train, the upper and lower thresholds are set according to experience to achieve the positioning of the pantograph; S12: For the pantograph image after positioning, Gaussian filtering, Canny edge detection and Hough transform are performed in sequence. The edge straight line of the horn is detected according to the edge angle and position information of the horn. The missing horn is detected by analyzing whether the edge straight line of the horn exists. Assume that the left and right horn edge straight lines are denoted as l left , l right , straight line l left The endpoints are (u left1 ,v left1 )、(u left2 ,v left2 ), with an angle of θ left , straight line l right The endpoints are (u right1 ,v right1 )、(u right2 ,v right2 ), with an angle of θ right , the screening conditions for the horn edge straight line are as follows:
6. The pantograph state detection method based on image processing according to claim 5, characterized in that: The specific process of extracting the left and right horn boundary points in step S2 is as follows: S21: Based on step S12, locate the sheep horns and extract the straight line l left , l right The center point of the sheep's horns is selected, and a rectangular window with a fixed width and height is set with the center point as the reference point to realize the positioning of the sheep's horns; S22: After the horns are located, the boundary points of the left and right horns are detected using a corner point detection method based on a sliding window, as follows: Extract the left horn location area Qy left The coordinates of the lower left pixel point, starting from the lower left pixel point, using a 3×3 sliding window, the current point is located at the lower left corner of the window, and scan Qy one by one from bottom to top vertically and from left to right horizontally left Pixel points, and count the number of edge points n in the sliding window. When n>3 and (i,j), (i+1,j), (i+2,j) are all edge points, stop edge point scanning and consider point (i,j) as the boundary point of the left horn, recorded as (u left3 ,v left3 ); Extract the right horn location area Qy right The coordinates of the lower right corner pixel point, starting from the lower right corner pixel point, using a 3×3 sliding window, the current point is located at the lower right corner of the window, and scan Qy one by one from bottom to top vertically and from right to left horizontally. right Pixel points, and count the number of edge points n in the sliding window. When n>3 and (i,j), (i-1,j), (i-2,j) are all edge points, stop edge point scanning and consider point (i,j) as the boundary point of the right horn, recorded as (u right3 ,v right3 ).
7. The pantograph state detection method based on image processing according to claim 6, characterized in that: The analysis of the pantograph's operating posture in step S2 is as follows: Based on the left and right horn boundary points extracted in step S22, the pantograph centerline offset value, pantograph vertical height difference and pantograph horizontality are used to analyze the pantograph operation posture after camera calibration. The specific index definitions are as follows: Centerline offset value: Vertical height difference: P y =||y left -y right |-y_base| Levelness: In the formula, (x left ,y left )、(x right ,y right ) are respectively (u left3 ,v left3 )、(u right3 ,v right3 ) The coordinate values after camera calibration, x_base, y_base and k_base are respectively the centerline position of the normal posture of the pantograph, the height difference between the left and right horns and the slope of the pantograph head connection line; When the centerline offset value, vertical height difference, and horizontality exceed the set thresholds, the pantograph posture may be abnormal.
