Track line induction board vehicle-mounted dynamic detection method

Abstract

The invention discloses a track line induction board vehicle-mounted dynamic detection method, which comprises the following steps that a 2D image acquisition module acquires image data of an induction board, a bracket and a bolt area to obtain an original induction board image; the 3D point cloud acquisition module acquires three-dimensional point cloud data of the induction plate and left and right steel rail areas; encoder pulse signals and radio frequency tag signals deployed on a line are collected, synchronization with the mileage of the line is achieved, and meanwhile trigger pulses are output to an induction plate data collection unit to collect data at equal intervals; performing intelligent detection and analysis on the acquired image data of the induction plate to obtain defect information including induction plate cracks, bracket cracks, bolt missing and bolt loosening; and processing the acquired point cloud data of the induction plates to obtain the area, height, longitudinal displacement and malocclusion information of each induction plate. The technical problems that an existing detection method is low in detection efficiency, insufficient in detection precision and incomplete in detection content can be solved.

Description

Technical Field

[0001] The present application relates to the field of railway engineering detection technology, and in particular to a vehicle-mounted dynamic detection method for track line induction plates based on computer vision technology. Background Art

[0002] Currently, many urban rail transit lines in my country use linear motor induction plates. During operation, timely detection and maintenance of induction plate anomalies are crucial for driving safety. Currently, the surface condition and geometric parameters of these induction plates rely on manual inspections and measurements, which is inefficient and consumes significant human resources and time. Consequently, devices with induction plate detection capabilities have been developed. Among them, invention applications CN111123282A and CN101513883A utilize two-point laser or ultrasonic distance sensors to measure induction plate height; while utility model patent CN215706365U proposes an induction plate foreign object detection device based on 3D image recognition. However, these devices are mounted on a trolley and can only measure induction plate height. They fail to eliminate errors caused by the trolley's vibration during operation, resulting in low detection accuracy and efficiency.

[0003] In the prior art, the following documents are mainly related to this application: Document 1 is a Chinese invention application filed by Beijing Jiaotong University on March 12, 2018, and published on November 2, 2021, with publication number CN108573213A. The application discloses a system and method for automatic detection of defective states of rail fasteners. The system includes: a first positioning module; a second positioning module; a segmentation module; a processing module; and a detection module. The system and method provided in this aspect can realize automatic positioning judgment of fasteners in different states, solving the problem of missed detection and objective accuracy of detection results that are difficult to guarantee with traditional manual methods, and at the same time provides new methods and new ideas for the design of automatic detection equipment for abnormal fastener states. The application can accurately and effectively identify abnormal defective fasteners in rail lines, significantly improve detection efficiency, and provide a good foundation for meeting the safe and efficient online detection of rail transit lines. The system can realize online detection with high detection speed, can adapt to the detection needs of different time periods under sufficient light source, and has strong system reliability and high accuracy.

[0004] Document 2 is a Chinese invention application filed by Suzhou Lichuang Zhiheng Electronic Technology Co., Ltd. on August 8, 2022, and published on September 16, 2022, with publication number CN115063416A. The application discloses a method and system for detecting the status of a rail fastener. The method for detecting the status of a rail fastener includes obtaining an image of the rail fastener to be identified, and inputting it into a first rail environment segmentation model and a second rail environment segmentation model to obtain the center point coordinate information and contour point coordinate information of the nut, rail and fastener; obtaining a local image of the rail fastener to be identified based on the center point and contour point coordinate information of the fastener, and inputting it into the fastener status detection model to obtain an initial predicted state; obtaining state auxiliary judgment information based on the center point and contour point coordinate information of the nut, rail and fastener, and further obtaining the state category of the rail fastener image to be identified. The detection method provided in the application obtains the position and shape of the fastener through the center point and contour point coordinate information of the nut, rail and fastener, obtains state auxiliary judgment information, and combines the initial predicted state of the fastener to achieve high-precision detection of the rail fastener status.

[0005] Document 3 is a Chinese invention application filed by Southwest Jiaotong University on March 14, 2019, and published on June 28, 2019, with publication number CN109948690A. This application discloses a high-speed rail scene perception method based on deep learning and structural information, including the following steps: Step 1: Acquire track images, divide them into training and test sets, and annotate the images in the training set to form a dataset; Step 2: Construct an SSD network model and construct a loss function; Step 3: Using the dataset formed in Step 1, iteratively train the network obtained in Step 2 to obtain a training model; Step 4: Input the video to be detected and perceived frame by frame into the training model obtained in Step 3, extract features, and obtain the location and category information of fasteners and shoulder guards. This information is used to distinguish between turnouts and ordinary track; Step 5: Cluster the location information of the positioning results in Step 4 to complete the perception of rails and sleepers. This application can detect and semantically segment track components in the turnout area with high accuracy and high speed.

[0006] Due to the large variety of sensor plate defects, none of the aforementioned references 1 to 3 involve detecting abnormal conditions such as the sensor plate's apparent state, displacement, misaligned threads, and missing or loose bolts. This makes it impossible to achieve systematic detection coverage of the sensor plate's state, and there is also a lack of a corresponding complete intelligent detection algorithm. Summary of the Invention

[0007] In view of this, the purpose of this application is to provide a vehicle-mounted dynamic detection method for a track line induction plate to solve the technical problems of low detection efficiency, insufficient detection accuracy, and incomplete detection content of existing detection methods.

[0008] In order to achieve the above-mentioned invention objectives, the present application specifically provides a technical implementation scheme of a vehicle-mounted dynamic detection method of a track line induction plate, comprising the following steps: S1) The 2D image acquisition module collects image data I of the sensor plate, bracket, and bolt area to obtain an original sensor plate image; S2) The 3D point cloud acquisition module collects 3D point cloud data P (x, y, z, i) of the sensor plate and the left and right rail areas, where x is the X-axis coordinate, y is the Y-axis coordinate, z is the Z-axis coordinate, and i is the reflection intensity of the point cloud; S3) collecting encoder pulse signals and radio frequency tag signals deployed on the line to achieve synchronization with the line mileage, and at the same time controlling the 2D image acquisition module and the 3D point cloud acquisition module to collect data at equal intervals; S4) Surface defect detection of sensor plates: Intelligent detection and analysis of collected sensor plate image data to obtain defect information including sensor plate cracks, bracket cracks, missing bolts, and loose bolts; Sensor plate geometric parameter detection: Process the collected sensor plate point cloud data to obtain the area, height, longitudinal displacement and tooth misalignment information of each sensor plate.

[0009] Furthermore, the 3D point cloud acquisition module includes three sets of camera components, and is characterized in that the step S2) includes: S21) first obtains point cloud data and performs filtering preprocessing, searching for point clouds belonging to the same target between two frames of point clouds taken by adjacent camera components at the same mileage; S22) The rotation matrix R and translation matrix T between the two frames of point clouds are then calculated through point cloud feature registration.

[0010] Furthermore, the point cloud feature registration in step S22) includes the following process: S221) Perform preliminary coarse registration on the two point cloud data of PA and PB to obtain an initial rotation matrix R ’ and the translation matrix T ’ , perform corresponding transformation on PB to obtain PB ’ =PB*R ’ +T ’ ; Among them, PA is the point cloud data of the left camera assembly, and PB is the point cloud data of the camera assembly adjacent to the left camera assembly; S222) Sampling the point cloud PA to obtain a set of sampling points; S223) searching for the point closest to the sampling point in the point cloud PB1 to obtain a set of matching points; S224) Calculate the updated rotation matrix R based on the matching points ’ and the translation matrix T ’ , for PB’ Perform the corresponding transformation to obtain PB ’ =PB*R ’ +T ’ ; S225) Repeat steps S222) to S224) and use the mean square error loss as the objective function to calculate the registration error. When the registration error is less than the set threshold T or the number of iterations is equal to the maximum number of iterations, the registration ends and the obtained rotation matrix R and translation matrix T are saved to achieve point cloud feature registration.

[0011] Furthermore, a photoelectric encoder is installed in the wheel of the track inspection train, and a radio frequency tag reader is installed under the bottom of the track inspection train. Step S3) includes the following process: S31) During the running of the track inspection train, the processing board collects the equally spaced pulse waveforms emitted by the photoelectric encoder, and multiplies and divides the pulse waveforms into input sources for the sampling triggers of the 2D image acquisition module and the 3D point cloud acquisition module, and calculates the preliminary mileage based on this; S32) Scan the RFID tag signal at the set sampling frequency. If the RFID tag signal is sensed, the actual mileage information represented by the tag is queried in the tag ID and actual mileage comparison table, and then the actual mileage information is mapped to each pulse signal to achieve synchronization with the line mileage.

[0012] Furthermore, in step S3), the line mileage is corrected according to the following formula: Assume that the real mileages of two adjacent RFID tags are M1 and M2 respectively. n , the corresponding trigger pulse counts are N1, N n , then between N1 and N n The real mileage between i =M1+(N i -N1)*(M n -M1) / (N n -N1), i=1, 2, 3...n.

[0013] Furthermore, the intelligent detection analysis in step S4) is combined with the target detection network and the classification network to make a judgment; the defect information includes the defect category, the mileage, the coordinates of the rectangular box in the image (x, y, w, h), as well as the actual physical size and the defect severity level.

[0014] Furthermore, the detection of surface defects of the induction plate in step S4) includes the following process: S41) performing intelligent detection and analysis on the collected sensor plate image data to obtain defect information including sensor plate cracks, bracket cracks, bolt missing, and bolt loosening; S42) First, the target detection network model is used to detect cracks in the sensing plate, cracks in the bracket, missing bolts, and normal bolts. Then, the normal bolt sub-image is sent to the classification network model to determine whether it is loose.

[0015] Furthermore, the object detection network model annotates the original sensor plate images with object detection boxes, dividing them into training and validation sets. The annotated object detection boxes are labeled with categories such as sensor plate cracks, bracket cracks, missing bolts, and normal bolts, with category IDs represented by 0, 1, 2, and 3, respectively. The entire dataset is divided into 80% of the samples as the training set and 20% as the validation set.

[0016] Furthermore, in the step S42), a target detection network model is constructed based on yolov10, CSPDarknet is used as the backbone network for feature extraction, a path aggregation network is used as the neck network, large convolution kernels and partitioned attention are used to enhance the contextual learning ability of the model, and the overall efficiency of the model is improved through spatial-channel decoupling downsampling and rank-guided modules.

[0017] Furthermore, the input size of the target detection network model is 1280, and the output detection frame of the target detection network model includes the target category, target frame and confidence. The training set is loaded into the target detection network model for iterative training, and the maximum number of iterations epoch=400 is set. After training for 10 epochs, the Maosic enhancement operation is turned off, and only conventional image enhancement operations including horizontal flipping, vertical flipping and chromaticity transformation are retained. The validation set is tested once every epoch iteration, and the average accuracy mAP information of the validation set is recorded. At the same time, the model parameters with the highest average accuracy mAP during the entire iterative training process are saved to the local disk, and the file is named best.pt. The model parameters obtained by the latest round of iterative training are saved as the last.pt file. Each image I to be detected is input into the network model loaded with the parameter best.pt for forward reasoning to obtain several target detection frames. The target categories 0, 1, 2, and 3 correspond to the induction plate crack, bracket crack, missing bolt, and normal bolt, respectively. The target detection frame with the target category 3 is used to intercept the ROI area in the image I to obtain the bolt sub-image S i , i=1, 2, 3...n, n is the number of target detection boxes with target category 3, S iThey are respectively sent into the classification network model to determine whether the bolt is loose. The classification network model divides the induction plate bolt sub-image data set into two categories: normal bolts and loose bolts, and the corresponding labels are 0 and 1 respectively. 80% of the normal bolts and loose bolt images are used as the classification training set, and the remaining 20% is used as the classification verification set. The classification network model loads the induction plate bolt sub-image data set for iterative training. The maximum number of iterations is set to 300, the learning rate is set to 0.001, and image enhancement operations such as random rotation, horizontal flipping, vertical flipping, and chromaticity transformation are adopted. The verification set is tested once every 1 epoch iteration, the classification accuracy ACC information is recorded, and the model parameters with the maximum accuracy ACC during the entire iterative training process are saved as the best.param file, and the model parameters obtained from the latest round of iterative training are saved as the last.param file. Bolt sub-image S i After being input into the classification model loaded with best.param and inferred, the confidence conf0 belonging to category 0 and the confidence conf1 belonging to category 1 are obtained. If conf0 < conf1, this bolt is loose; otherwise, it is a normal bolt.

[0018] Furthermore, in the point cloud processing in step S4), according to the distribution characteristics of the induction plate, the front, back, left, and right edge positions of each induction plate are extracted through edge detection, so as to segment the point cloud of each induction plate. The height information of each position of the induction plate is further obtained from the coordinate z of the point cloud. The longitudinal displacement information of the induction plate is obtained by calculating the offset of the gap between adjacent two induction plates exceeding the standard value range. The misalignment information of the induction plate includes the lateral difference, height, and angle deviation between adjacent induction plates. The vertices of the induction plate are obtained through the boundary intersection points of the induction plate, and the lateral and height deviations of the induction plate are obtained by comparing the coordinate x deviation and coordinate z deviation between the corresponding vertices of adjacent induction plates. Then, the angle deviation is solved by fitting the normal vector of the induction plate plane calculated.

[0019] Furthermore, the detection of the geometric parameters of the induction plate in step S4) includes the following process: S411) First, the induction plate is segmented by block, and the upper, lower, left, and right edge positions of the induction plate are extracted through gradient edge detection, so as to segment the point cloud Psb of each induction plate n , n = 1, 2, 3... n; S412) Perform bilateral filtering on the point cloud Psb to eliminate noise and obtain Psb n ’ , the induction plate point cloud Psb​​​​​​S413) If the maximum value Max > the allowable ultra-high value, it is determined that the induction plate is abnormally ultra-high. On the contrary, it is judged whether the minimum value Min < the allowable ultra-low value holds. If it holds, it is determined that the induction plate is abnormally too low. Otherwise, the height of the induction plate is normal; S414) The longitudinal displacement information of the induction plate is obtained by calculating the offset of the gap between two adjacent induction plates exceeding the standard value range. d1 is the longitudinal displacement between the first induction plate Psb1 ’ and the second induction plate Psb2 ’ The longitudinal displacement between them, d2 is the second induction plate Psb2 ’ and the third induction plate Psb3 ’ The longitudinal displacement between them; S415) If the allowable minimum displacement Offset Min < d1 < the allowable maximum displacement Offset Max , it is judged that the longitudinal displacement of the first induction plate is normal. Otherwise, it is judged that the longitudinal displacement of the first induction plate is abnormal.

[0020] Furthermore, the misalignment information of the induction plate includes the lateral, height and angle deviations between adjacent induction plates. The detection of the geometric parameters of the induction plate in step S4) includes the following processes: S421) The vertices of the induction plate are obtained through the boundary intersection points of the induction plate. The vertices of the first induction plate are A1, B1, C1, D1, and the vertices of the second induction plate are A2, B2, C2, D2, and so on. Each vertex includes x, y, z coordinates; S422) c1 is the lateral deviation between the first induction plate and the second induction plate, which is obtained by taking the absolute value of the difference between the X-axis coordinate D1(x) of D1 and the X-axis coordinate C2(x) of C2, that is, |D1(x) - C2(x)|. c2 is the lateral deviation between the second induction plate and the third induction plate. If the lateral deviation is greater than the lateral deviation threshold Tx, it is considered misaligned. Otherwise, it is determined that the lateral deviation is normal; The height deviation is obtained by taking the absolute value of the difference between the z coordinates of the corresponding vertices of adjacent induction plates. The height deviations between the first induction plate and the second induction plate are |D1(z) - C2(z)| and |B1(z) - A2(z)|. If the height deviation is greater than the height deviation threshold Tz, it is determined that there is misalignment. Otherwise, it is determined that the height deviation is normal.

[0021] Furthermore, step S423) includes: The point cloud of the first sensor plate is framed by the vertices A1, B1, C1, and D1 of the first sensor plate. Then, the plane equation of the sensor plate surface is iteratively calculated using a plane fitting algorithm based on random sampling consistency. Its normal vector is n1 (A, B, C), and the normal vector n2 (A, B, C) of the second sensor plate is calculated. The angle deviation is then solved by the vector inner product based on the fitted normal vector of the sensor plate plane. , if the angle deviation If the angle deviation is greater than the threshold, it is judged as malocclusion, otherwise it is normal.

[0022] Furthermore, the method includes: S5) After the detection result is obtained by calculation in step S4), the detection result information and the corresponding original data are transmitted to the ground user terminal, and the operator is notified to specify the operation plan.

[0023] By implementing the technical solution of the on-board dynamic detection method of the track line sensor plate provided by the present application, the following beneficial effects are achieved: (1) The vehicle-mounted dynamic detection method for the track line induction plate of this application automatically processes the three-dimensional point cloud data of the induction plate and analyzes the geometric parameter information of the induction plate, such as the height anomaly, longitudinal displacement, and misalignment of the induction plate, to provide a reliable basis for the operation and maintenance of the induction plate. It realizes the vehicle-mounted dynamic detection of the rail transit induction plate, greatly improves the detection efficiency and accuracy, and comprehensively covers the detection of the appearance and geometric parameters of the induction plate; (2) The on-board dynamic detection method for track line induction plates in this application uses a target detection network and a classification network to intelligently analyze surface cracks, missing bolts, and loose defects on the induction plates, thereby achieving end-to-end detection of surface defects on the induction plates and further improving detection efficiency and accuracy. (3) The on-board dynamic detection method of the track line sensor plate of this application uses a 2D camera component to collect the surface texture features of the sensor plate body and the bracket, and a 3D camera component to collect the rail contour and the three-dimensional point cloud information of the sensor plate. It can be carried on the train to collect and process the sensor plate data at high speed, realize automatic and efficient detection of the surface defects and geometric parameters of the sensor plate, and guide relevant management and operating personnel to maintain the sensor plate. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] To more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other embodiments can be derived from these drawings without inventive effort.

[0025] Figure 1This is a structural block diagram of a specific embodiment of the track line induction plate vehicle-mounted dynamic detection system based on the method of this application; Figure 2 This is a schematic diagram of the distribution of 3D point cloud modules of a specific embodiment of the vehicle-mounted dynamic detection method of the track line sensor plate of the present application; Figure 3 This is a flowchart of the joint calibration of a 3D camera assembly in accordance with a specific embodiment of the vehicle-mounted dynamic detection method for a track line sensor plate of the present application; Figure 4 This is a block diagram of the structure of the mileage positioning and synchronization unit in a specific embodiment of the track line sensor board vehicle-mounted dynamic detection system based on the method of this application; Figure 5 This is a flow chart of the detection of surface defects of the induction plate of a specific embodiment of the vehicle-mounted dynamic detection method of the track line induction plate of the present application; Figure 6 This is a schematic diagram of geometric parameter items of a specific embodiment of the track line induction plate vehicle-mounted dynamic detection method of the present application.

[0026] In the figure: 1-sensor plate data acquisition unit, 2-mileage positioning and synchronization unit, 3-intelligent analysis and processing unit, 4-communication unit, 5-power supply unit, 6-storage unit, 7-sensor plate, 8-bracket, 9-bolt, 10-rail, 11-2D image acquisition module, 12-3D point cloud acquisition module, 121-left 3D camera assembly, 122-center 3D camera assembly, 123-right 3D camera assembly, 21-photoelectric encoder, 22-processing board, 23-RFID tag reader. DETAILED DESCRIPTION

[0027] For the purpose of reference and clarity, the technical terms, abbreviations or abbreviations used below are recorded as follows: ICP: Iterative Closest Point, iterative closest point algorithm; yolov10: an object detection network. CSPDarknet: a feature extraction module; PAN: Path Aggregation Network, path aggregation network; ResNet-18: a lightweight classification network model; 2D: 2 dimension, two-dimensional; 3D:3 dimension, three-dimensional; FPGA: Field Programmable Gate Array, field programmable gate array; CPU: Central Processing Unit; GPU: Graphics Processing Unit; JPG: An image compression format.

[0028] To make the purpose, technical solutions, and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0029] As attached Figure 1 To the attached Figure 6 As shown, a specific embodiment of the vehicle-mounted dynamic detection method of the track line induction plate of the present application is given. The present application is further explained below in conjunction with the drawings and specific embodiments.

[0030] In order to achieve full-factor collection of various defect characteristics of the sensor plate and realize automatic and efficient detection of surface defects and geometric dimension anomalies of the sensor plate, the specific embodiment of the present application provides an efficient and robust on-board dynamic detection method for track line sensor plates, which can adapt to the complex and changeable railway line environment and maintain high detection efficiency while ensuring the detection effect.

[0031] Example 1 An embodiment of the on-board dynamic detection method of a track line induction plate of the present application specifically comprises the following steps: S1) The 2D image acquisition module 11 acquires image data I of the sensor plate 7, bracket 8 and bolt 9 to obtain an original sensor plate image; S2) The 3D point cloud acquisition module 12 acquires three-dimensional point cloud data P (x, y, z, i) of the sensor plate and the left and right rails 10, where x is the X-axis coordinate, y is the Y-axis coordinate, z is the Z-axis coordinate, and i is the reflection intensity of the point cloud; S3) Collect encoder pulse signals and radio frequency tag signals deployed on the line to synchronize with the line mileage, and control the 2D image acquisition module 11 and the 3D point cloud acquisition module 12 to collect data at equal intervals; S4) Surface defect detection of sensor plates: Intelligent detection and analysis of collected sensor plate image data to obtain defect information including sensor plate cracks, bracket cracks, missing bolts, and loose bolts; Sensor plate geometric parameter detection: Process the collected sensor plate point cloud data to obtain the area, height, longitudinal displacement and tooth misalignment information of each sensor plate 7; Steps S1) to S3) are executed in no particular order.

[0032] The 3D point cloud acquisition module 12 further includes three camera components, and step S2) further includes: S21) first obtains point cloud data and performs filtering preprocessing, searching for point clouds belonging to the same target between two frames of point clouds taken by adjacent camera components at the same mileage; S22) The rotation matrix R and translation matrix T between the two frames of point clouds are then calculated through point cloud feature registration.

[0033] As attached Figure 3 As shown, the point cloud feature registration in step S22) further includes the following process: S221) Perform preliminary coarse registration on the two point cloud data of PA and PB to obtain an initial rotation matrix R ’ and the translation matrix T ’ , perform corresponding transformation on PB to obtain PB ’ =PB*R ’ +T ’ ; Among them, PA is the point cloud data of the left camera assembly, and PB is the point cloud data of the camera assembly adjacent to the left camera assembly; S222) Sampling the point cloud PA to obtain a set of sampling points; S223) searching for the point closest to the sampling point in the point cloud PB1 to obtain a set of matching points; S224) Calculate the updated rotation matrix R based on the matching points ’ and the translation matrix T ’ , for PB ’ Perform the corresponding transformation to obtain PB ’ =PB*R ’ +T ’ ; S225) Repeat steps S222) to S224) and use the mean square error loss as the objective function to calculate the registration error. When the registration error is less than the set threshold T or the number of iterations is equal to the maximum number of iterations, the registration ends and the obtained rotation matrix R and translation matrix T are saved to achieve point cloud feature registration.

[0034] A photoelectric encoder 21 is installed in the wheel of the track inspection train, and a radio frequency tag reader 23 is installed under the bottom of the track inspection train. Step S3) further includes the following process: S31) During the track inspection train's travel, the processing board 22 collects the equally spaced pulse waveforms emitted by the photoelectric encoder 21, multiplies and divides the pulse waveforms into input sources for the sampling triggers of the 2D image acquisition module 11 and the 3D point cloud acquisition module 12, and calculates the preliminary mileage accordingly. S32) Scan the RFID tag signal at the set sampling frequency. If the RFID tag signal is sensed, the actual mileage information represented by the tag is queried in the tag ID and actual mileage comparison table, and then the actual mileage information is mapped to each pulse signal to achieve synchronization with the line mileage.

[0035] In step S3), the line mileage is further corrected according to the following formula: Assume that the real mileages of two adjacent RFID tags are M1 and M2 respectively. n , the corresponding trigger pulse counts are N1, N n , then between N1 and N n The real mileage between i =M1+(N i -N1)*(M n -M1) / (N n -N1), i=1, 2, 3……n.

[0036] The intelligent detection analysis in step S4) combines the target detection network and the classification network to make a judgment. The defect information further includes the defect category, the mileage, the coordinates of the rectangular box in the image (x, y, w, h), the actual physical size, and the severity level of the defect.

[0037] As attached Figure 5 As shown, the surface defect detection of the induction plate in step S4) further includes the following processes: S41) performing intelligent detection and analysis on the collected sensor plate image data to obtain defect information including sensor plate cracks, bracket cracks, bolt missing, and bolt loosening; S42) First, the target detection network model is used to detect cracks in the sensing plate, cracks in the bracket, missing bolts, and normal bolts. Then, the normal bolt sub-image is sent to the classification network model to determine whether it is loose.

[0038] The object detection network model annotates the original sensor plate images with object detection boxes and divides them into training and validation sets. The object detection box labels include sensor plate cracks, bracket cracks, missing bolts, and normal bolts, with class IDs represented by 0, 1, 2, and 3, respectively. The entire dataset is divided into 80% of the samples as the training set and 20% as the validation set.

[0039] In step S42), a target detection network model is constructed based on yolov10. The feature extraction part uses CSPDarknet as the backbone network, the path aggregation network as the neck network, and the context learning ability of the model is enhanced by large convolution kernels and partitioned attention. The overall efficiency of the model is improved through spatial-channel decoupling downsampling and rank-guided modules.

[0040] The input size of the object detection network model is 1280, and the output detection boxes of the object detection network model include the object category, the object box, and the confidence level. Load the training set and send it into the object detection network model for iterative training. Set the maximum number of iterations epoch = 400. After training for 10 epochs, turn off the Maosic augmentation operation and only retain the conventional image augmentation operations including horizontal flipping, vertical flipping, and chromaticity transformation. Perform a test on the validation set after each 1 epoch iteration, and record the mean average precision (mAP) information of the validation set. At the same time, save the model parameters with the highest mAP during the entire iterative training process to the local disk, and name the file best.pt. Save the model parameters obtained from the latest round of iterative training as the last.pt file. Input each image I to be detected into the network model loaded with the parameters best.pt for forward inference to obtain several object detection boxes. The object categories 0, 1, 2, and 3 correspond to the crack of the induction plate, the crack of the bracket, the missing bolt, and the normal bolt respectively. Use the object detection box with the object category 3 to intercept the ROI region in the image I to obtain the bolt sub-image S i , i = 1, 2, 3... n, where n is the number of object detection boxes with the object category 3, and S i are respectively sent into the classification network model to determine whether the bolt is loose or not. The classification network model divides the induction plate bolt sub-image dataset into two categories: normal bolts and loose bolts, and the corresponding labels are 0 and 1 respectively. Use 80% of the normal bolts and loose bolt images as the classification training set, and the remaining 20% as the classification validation set. The classification network model loads the induction plate bolt sub-image dataset for iterative training. Set the maximum number of iterations to 300, the learning rate to 0.001, and adopt the image augmentation operations of random rotation, horizontal flipping, vertical flipping, and chromaticity transformation. Perform a test on the validation set after each 1 epoch iteration, record the classification accuracy (ACC) information, and save the model parameters with the maximum ACC during the entire iterative training process as the best.param file. Save the model parameters obtained from the latest round of iterative training as the last.param file. The bolt sub-image S i is input into the classification model loaded with best.param. After inference, obtain the confidence level conf0 belonging to category 0 and the confidence level conf1 belonging to category 1. If conf0 < conf1, then this bolt is loose; otherwise, it is a normal bolt.

[0041] In the point cloud processing in step S4), according to the distribution characteristics of the induction plates 7, the front, rear, left, and right edge positions of each induction plate 7 are extracted through edge detection, so as to segment the point cloud of each induction plate 7. The height information of each position of the induction plate 7 is further obtained from the z coordinate of the point cloud. The longitudinal displacement information of the induction plate 7 is obtained by calculating the offset of the gap between two adjacent induction plates 7 exceeding the standard value range. The misalignment information of the induction plates includes the lateral difference, height, and angle deviation between adjacent induction plates 7. The vertices of the induction plate 7 are obtained through the boundary intersection points of the induction plate 7. The lateral and height deviations of the induction plate 7 are obtained by comparing the coordinate x deviation and coordinate z deviation between the corresponding vertices of adjacent induction plates 7. Then, the angle deviation is solved through the normal vector of the induction plate plane calculated by fitting.

[0042] The detection of the geometric parameters of the induction plate in step S4) further includes the following process: S411) First, the induction plate is segmented by block, and the upper, lower, left, and right edge positions of the induction plate 7 are extracted through gradient edge detection, so as to segment the point cloud Psb of each induction plate 7 n , n = 1, 2, 3... n; S412) Perform bilateral filtering on the point cloud Psb to eliminate noise and obtain Psb n ’ , the point cloud Psb of the induction plate ’ The z coordinate of each point in is the height value of each position of the induction plate. Extract the maximum value Max and minimum value Min of the z coordinate in Psb n ’ ; S413) If the maximum value Max > the allowable ultra-high value, it is determined that the induction plate is abnormally ultra-high. On the contrary, it is judged whether the minimum value Min < the allowable ultra-low value holds. If it holds, it is determined that the induction plate is abnormally too low. Otherwise, the height of the induction plate is normal; S414) The longitudinal displacement information of the induction plate 7 is obtained by calculating the offset of the gap between two adjacent induction plates 7 exceeding the standard value range. d1 is the longitudinal displacement between the first induction plate Psb1 ’ and the second induction plate Psb2 ’ , d2 is the longitudinal displacement between the second induction plate Psb2 ’ and the third induction plate Psb3 ’ ; S415) If the allowable minimum displacement Offset Min < d1 < the allowable maximum displacement Offset Max , it is judged that the longitudinal displacement of the first induction plate is normal. Otherwise, it is judged that the longitudinal displacement of the first induction plate is abnormal.

[0043] The sensor plate misalignment information includes the lateral, height and angular deviations between adjacent sensor plates 7. The sensor plate geometric parameter detection in step S4) further includes the following processes: S421) Obtaining the vertices of the sensing plate through the boundary intersections of the sensing plate 7. The vertices of the first sensing plate are A1, B1, C1, and D1. The vertices of the second sensing plate are A2, B2, C2, and D2. And so on. Each vertex includes x, y, and z coordinates. S422) c1 is the lateral deviation between the first and second sensing plates, obtained by taking the absolute value of the difference between the X-axis coordinate D1(x) of D1 and the X-axis coordinate C2(x) of C2, i.e., |D1(x)-C2(x)|. c2 is the lateral deviation between the second and third sensing plates. If the lateral deviation is greater than the lateral deviation threshold Tx, it is considered to be a misaligned tooth. Otherwise, it is determined to be normal. S423) The height deviation is the absolute value of the difference in coordinates z between the corresponding vertices of adjacent sensing plates 7. The height deviations of the first and second sensing plates are |D1(z)-C2(z)| and |B1(z)-A2(z)|. If the height deviation is greater than the height deviation threshold Tz, it is determined to be a malocclusion. Otherwise, it is determined to be normal.

[0044] Step S423) includes: The point cloud of the first sensor plate is framed by the vertices A1, B1, C1, and D1 of the first sensor plate. Then, the plane equation of the surface of the sensor plate 7, Ax+By+Cz+D=0, is iteratively calculated using a plane fitting algorithm based on random sampling consistency. Its normal vector is n1(A, B, C), and the normal vector n2(A, B, C) of the second sensor plate is calculated. The angle deviation is then solved by the vector inner product based on the fitted normal vector of the sensor plate plane. , if the angle deviation If the angle deviation is greater than the threshold, it is judged as malocclusion, otherwise it is normal.

[0045] The onboard dynamic detection method of the track line induction plate further comprises: S5) After the test results are calculated in step S4), the test result information and the corresponding original data are transmitted to the ground user terminal, and the operator is notified to specify the operation plan.

[0046] Example 2 As attached Figure 1 As shown, an embodiment of the track line induction plate vehicle-mounted dynamic detection system 10 based on the method described in Example 1 is installed on a track detection train and specifically includes: The sensor board data acquisition unit 1 includes a 2D image acquisition module 11 and a 3D point cloud acquisition module 12; The 2D image acquisition module 11 acquires image data I of the sensor plate 7, bracket 8 and bolt 9 to obtain an original sensor plate image; The 3D point cloud acquisition module 12 acquires three-dimensional point cloud data P (x, y, z, i) of the sensing plate and the left and right rails 10. Here, x is the X-axis coordinate, which refers to the positive direction of the X-axis from left to right along the direction of the sleepers when facing the extension direction of the rails 10; y is the Y-axis coordinate, which refers to the positive direction of the Y-axis along the direction of the rails 10; z is the Z-axis coordinate, which is defined by the X and Y axes using the right-hand screw rule and indicates the height of the sensing plate 7; and i is the reflection intensity of the point cloud. The mileage positioning and synchronization unit 2 collects encoder pulse signals and radio frequency tag signals deployed on the line to achieve synchronization with the line mileage, and at the same time outputs trigger pulses to the sensor board data acquisition unit 1 to collect data at equal intervals; The intelligent analysis and processing unit 3 includes a sensor plate surface defect detection module and a sensor plate geometric parameter detection module. The sensor plate surface defect detection module feeds the collected sensor plate image data into the intelligent detection model for analysis, obtaining defect information including sensor plate cracks, bracket cracks, missing bolts, and loose bolts. The sensor plate geometric parameter detection module feeds the collected sensor plate point cloud data into the point cloud processing model to obtain the area, height, longitudinal displacement, and misalignment information of each sensor plate 7.

[0047] The onboard dynamic detection method for track line sensor panels also includes a communication unit 4, which includes a wired or wireless communication module. After the intelligent analysis and processing unit 3 calculates the detection results, it transmits the detection results and corresponding raw data to the ground user terminal via the communication unit 4, and notifies the operator to specify the operation plan. The dynamic detection system also includes a storage unit 6 for storing the collected raw data in a designated location. The dynamic detection system also includes a power supply unit, which provides power to the entire system, including AC / DC conversion and voltage stabilization.

[0048] The 2D image acquisition module includes two sets of 2D camera assemblies, each consisting of a high-speed linear array camera, a high-resolution lens, a laser, and peripheral circuitry. These assemblies are responsible for collecting image data I of the sensor plate (body) 7, bracket 8, and bolt 9. The 3D point cloud acquisition module 12 further includes three sets of camera assemblies, each of which includes a high-speed 3D area array camera, a high-resolution lens, a linear laser light source, and peripheral circuitry. The 2D image acquisition module can be tilted toward the center axis of the sensor plate by adjusting its installation angle to ensure clear visibility of the sensor plate bracket area. Each acquisition module utilizes a dedicated laser light source for fill lighting, ensuring the system can function properly in various lighting environments.

[0049] like Figure 2As shown, there is a certain range of common overlapping field of view between the two adjacent (3D) camera components of the 3D point cloud acquisition module 12, which is used to calibrate the three 3D point cloud acquisition modules 12 to the same coordinate system. The calibration process is shown in the attached Figure 3 As shown, the 3D point cloud acquisition module 12 first acquires point cloud data and performs filtering preprocessing, searching for point clouds belonging to the same target between two point cloud frames taken at the same mileage by adjacent camera components. It then calculates the transformation matrix (including the rotation matrix R and translation matrix T) between the two point cloud frames using a point cloud feature registration algorithm (e.g., ICP).

[0050] The 3D point cloud acquisition module 12 performs preliminary coarse registration on the two point cloud data of PA and PB to obtain an initial rotation matrix R ’ and the translation matrix T ’ , perform corresponding transformation on PB to obtain PB ’ =PB*R ’ +T ’ , PA is the point cloud data of the left camera component, and PB is the point cloud data of the camera component adjacent to the left camera component. Sample the point cloud PA to obtain a set of sampling points, find the points closest to the sampling points in the point cloud PB1, obtain a set of matching points, and then calculate the updated rotation matrix R based on the matching points ’ and the translation matrix T ’ , for PB ’ Perform the corresponding transformation to obtain PB ’ =PB*R ’ +T ’ . ’ Perform cyclic update iterations and use mean square error loss (MSE) as the objective function to calculate the registration error. When the registration error is less than the set threshold T or the number of iterations is equal to the maximum number of iterations, the registration ends and the obtained rotation matrix R and translation matrix T are saved to achieve point cloud feature registration.

[0051] As attached Figure 4As shown, the mileage positioning and synchronization unit 2 further includes a photoelectric encoder 21, a processing board 22, and a radio frequency tag reader 23. The photoelectric encoder 21 is installed in the wheel of the track inspection train. During the operation of the track inspection train, the processing board 22 collects the equally spaced pulse waveforms emitted by the photoelectric encoder 21, and after multiplying and dividing the pulse waveforms, uses them as the input source for the sampling trigger of the 2D image acquisition module 11 and the 3D point cloud acquisition module 12. The sampling interval of a trigger pulse cycle is set to 0.5mm, and the distance after 1000 pulses is 0.5m, and the preliminary mileage is calculated accordingly. Since the wheels inevitably wear or other conditions cause changes in wheel diameter during operation, resulting in the sampling interval not being strictly equal to 0.5mm, it is necessary to use radio frequency tags deployed on the line to calibrate and synchronize the mileage. In this embodiment, the RFID tag reader 23 is installed under the bottom of the track inspection train and scans the RFID tag signal at a set sampling frequency (e.g., 1000 Hz). If the RFID tag signal is sensed, the actual mileage information represented by the tag is queried in the tag ID and actual mileage comparison table, and then the actual mileage information is mapped to each pulse signal to achieve synchronization with the line mileage.

[0052] The mileage positioning and synchronization unit 2 further corrects the line mileage according to the following formula: Assume that the actual mileages of two adjacent radio frequency tags are M1 and M2 respectively. n , the corresponding trigger pulse counts are N1, N n (The mileage calculated based on the equal pulse interval of 0.5mm is N1*0.5mm, N n *0.5mm), then between N1 and N n The real mileage between i =M1+(N i -N1)*(M n -M1) / (N n -N1), i=1, 2, 3……n.

[0053] In order to ensure storage efficiency, the present application stores the collected raw data on a storage unit 6, which can specifically be a solid-state hard drive. Since the amount of point cloud data is large, a 16-bit depth map is used to store the raw data. In order to further reduce the amount of data, a data compression algorithm is used to compress and store the raw data. Since three-dimensional data is sensitive to distance information, lossy compression methods can cause data distortion and cause detection errors. Therefore, a lossless compression method is required. First, differential encoding is performed, and then positive and negative number remapping is performed. Run-length encoding and variable-length encoding are combined to reduce the number of bits of the entire image. The two-dimensional raw image data can be compressed using a general lossy compression algorithm. The system uses the JPG compression algorithm to compress 2,000 images and store them in one large file to ensure data storage and copying efficiency.

[0054] The intelligent analysis and processing unit 3 is composed of an industrial-grade computer and corresponding detection software. To improve detection efficiency, this application uses a CPU (Central Processing Unit) for data processing and a GPU (Graphic Processing Unit) for calculation acceleration. The sensor plate surface defect detection module sends the collected sensor plate image data to the intelligent detection model for analysis to obtain defect information including sensor plate cracks, bracket cracks, missing bolts, and loose bolts. The intelligent detection model combines the target detection network model and the classification network model for judgment. As shown in the attached figure, Figure 5 As shown in the figure, the intelligent detection model first uses the object detection network model to detect cracks in the sensor plate, cracks in the bracket, missing bolts, and normal bolts. The normal bolt sub-image is then fed into the classification network model to determine whether it is loose. Defect information further includes the defect category, mileage, coordinates of the rectangular box in the image (x and y coordinates of the upper left vertex, width w and height h of the rectangular box), actual physical dimensions, and defect severity.

[0055] The object detection network model annotates the original sensor plate images with object detection boxes and divides them into training and validation sets. The object detection box labels include sensor plate cracks, bracket cracks, missing bolts, and normal bolts, with class IDs represented by 0, 1, 2, and 3, respectively. The entire dataset is divided into 80% of the samples as the training set and 20% as the validation set.

[0056] The intelligent analysis and processing unit 3 builds a target detection network model based on yolov10. The feature extraction part uses CSPDarknet as the backbone network and the path aggregation network (PAN) as the neck network (Neck). It uses large convolution kernels and partitioned attention to enhance the context learning ability of the model, and improves the overall efficiency of the model through spatial-channel decoupling downsampling and rank-guided modules.

[0057] The input size of the object detection network model is 1280 (an integer multiple of 32). The output detection boxes of the object detection network model include the object category, the object box (the x and y coordinates x1, y1 of the upper left vertex and the x and y coordinates x2, y2 of the lower right vertex), and the confidence. The training set is loaded and sent into the object detection network model for iterative training. The maximum number of iterations epoch = 400 is set. After training for 10 epochs, the Maosic augmentation operation is turned off, and only the conventional image augmentation operations including horizontal flipping, vertical flipping, and chromaticity transformation are retained. After each epoch iteration (traversing all training samples), the validation set is tested once, and the mean Average Precision (mAP) information of the validation set is recorded. At the same time, the model parameters with the highest mAP during the entire iterative training process are saved to the local disk, and the file is named best.pt. The model parameters obtained from the latest round of iterative training are saved as the last.pt file. Each image I to be detected is input into the network model loaded with the parameters best.pt for forward inference, and several object detection boxes are obtained. The object categories 0, 1, 2, and 3 correspond to the crack of the induction plate, the crack of the bracket, the missing bolt, and the normal bolt respectively. The Region Of Interesting (ROI) area is intercepted in the image I using the object detection box with the object category 3 to obtain the bolt sub-image S i , i = 1, 2, 3... n, where n is the number of object detection boxes with the object category 3, and S i are respectively sent into the classification network model to determine whether the bolt is loose. The classification network model divides the induction plate bolt sub-image data set into two categories: normal bolts and loose bolts, and the corresponding labels are 0 and 1 respectively. 80% of the normal bolts and loose bolt images are used as the classification training set, and the remaining 20% are used as the classification validation set. The classification network model loads the induction plate bolt sub-image data set for iterative training. The maximum number of iterations is set to 300, the learning rate is set to 0.001, and the image augmentation operations of random rotation (the angle is limited within 5°), horizontal flipping, vertical flipping, and chromaticity transformation are adopted. After each epoch iteration (traversing all training samples), the validation set is tested once, the classification accuracy ACC information is recorded, and the model parameters with the maximum ACC during the entire iterative training process are saved as the best.param file. The model parameters obtained from the latest round of iterative training are saved as the last.param file. The bolt sub-image S i is input into the classification model loaded with best.param, and the confidence conf0 belonging to category 0 and the confidence conf1 belonging to category 1 are obtained after inference. If conf0 < conf1, this bolt is loose; otherwise, it is a normal bolt.

[0058] Based on the distribution characteristics of the induction plate 7, the point cloud processing model extracts the front, back, left, and right edge positions of each induction plate 7 through edge detection, thereby segmenting the point cloud of each induction plate 7. The height information of each position of the induction plate 7 is further obtained from the z coordinate of the point cloud. The longitudinal displacement information of the induction plate 7 is calculated by calculating the offset of the gap between two adjacent induction plates 7 exceeding the standard value range. The induction plate misalignment information includes the lateral difference, height, and angle deviation between adjacent induction plates 7. The vertices of the induction plate 7 are obtained through the boundary intersection points of the induction plate 7. The lateral and height deviations of the induction plate 7 are obtained by comparing the coordinate x deviation and coordinate z deviation between the corresponding vertices of adjacent induction plates 7. Then, the angle deviation is solved by fitting the normal vector of the induction plate plane calculated.

[0059] As shown in the appendix Figure 6 shown, the induction plate geometric parameter detection module first divides the induction plate into blocks. When further dividing the induction plate (body) 7, the significant height difference between the induction plate (body) 7 and its surroundings is fully utilized. The upper, lower, left, and right edge positions of the induction plate 7 can be extracted through gradient edge detection, thereby segmenting the point cloud Psb of each induction plate 7 n , n = 1, 2, 3... n. Bilateral filtering is performed on the point cloud Psb to eliminate noise and obtain Psb n ’ , the point cloud Psb of the induction plate ’ The z coordinate of each point in it is the height value of each position of the induction plate. The maximum value Max and minimum value Min of the z coordinate in Psb are extracted n ’ . If the maximum value Max > the allowable ultra-high value, it is determined that the induction plate is abnormally ultra-high. Otherwise, it is judged whether the minimum value Min < the allowable ultra-low value holds. If it holds, it is determined that the induction plate is abnormally too low. Otherwise, the height of the induction plate is normal. The longitudinal displacement information of the induction plate 7 is calculated by calculating the offset of the gap between two adjacent induction plates 7 exceeding the standard value range. d1 is the longitudinal displacement between the first induction plate Psb1 ’ and the second induction plate Psb2 ’ , d2 is the longitudinal displacement between the second induction plate Psb2 ’ and the third induction plate Psb3 ’ . If the allowable minimum displacement Offset Min < d1 < the allowable maximum displacement Offset Max , it is judged that the longitudinal displacement of the first induction plate is normal. Otherwise, it is judged that the longitudinal displacement of the first induction plate is abnormal.

[0060] The sensor plate misalignment information includes the lateral, height, and angular deviations between adjacent sensor plates 7. The sensor plate geometric parameter detection module obtains the sensor plate vertices through the boundary intersections of the sensor plates 7. The vertices of the first sensor plate are A1, B1, C1, and D1, and the vertices of the second sensor plate are A2, B2, C2, and D2, and so on. Each vertex includes x, y, and z coordinates. c1 is the lateral deviation between the first and second sensor plates. It is obtained by taking the absolute value of the difference between the X-axis coordinate D1(x) of D1 and the X-axis coordinate C2(x), that is, |D1(x)-C2(x)|. c2 is the lateral deviation between the second and third sensor plates. If the lateral deviation is greater than the lateral deviation threshold Tx, it is considered misaligned. Otherwise, it is judged to be normal. The height deviation is the absolute value of the difference between the coordinates z of the corresponding vertices of adjacent sensing plates 7. The height deviations of the first and second sensing plates are |D1(z)-C2(z)| and |B1(z)-A2(z)|. If the height deviation is greater than the height deviation threshold Tz, it is determined to be a malocclusion; otherwise, it is determined to be normal.

[0061] The sensor plate geometric parameter detection module uses the vertices A1, B1, C1, and D1 of the first sensor plate to define the point cloud of the first sensor plate. Then, using a plane fitting algorithm based on random sampling consensus (RANSAC), iteratively calculates the plane equation of the sensor plate 7 surface, Ax+By+Cz+D=0, with its normal vector n1(A, B, C). It also calculates the normal vector n2(A, B, C) of the second sensor plate. The angle deviation is then solved using the inner product of the vectors from the fitted sensor plate plane normal vector. , if the angle deviation If the angle deviation is greater than the threshold, it is judged as malocclusion, otherwise it is normal.

[0062] After the intelligent analysis and processing unit 3 calculates the test results, it must be transmitted to the ground user terminal (i.e., the ground management platform) via the communication unit 4, and the operator is notified to specify the work plan. The communication unit 4 further transmits the test results and corresponding raw data to the ground user terminal via 4G / 5G wireless communication using HTTP / HTTPS, TCP / IP, and UDP protocols. The entire system is powered by the power supply unit 5. First, the power supply unit 5 receives an AC 220V power supply from the vehicle terminal, converts it to the DC 110V and DC 24V voltages required by the system through the AC / DC conversion module, and supplies the system's various units with these voltages.

[0063] In the description of this application, it should be noted that when an element is referred to as being "fixed to" or "disposed on" another element, it may be directly disposed on the other element or indirectly disposed on the other element. When an element is referred to as being "connected to" another element, it may be directly connected to the other element or indirectly connected to the other element.

[0064] It should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present application.

[0065] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of the technical features indicated. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of the features. Throughout the description of this application, "plurality" or "several" means two or more, unless otherwise specifically defined.

[0066] It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification for people familiar with this technology to understand and read, and are not used to limit the conditions under which this application can be implemented. Therefore, they have no technical significance. Any structural modification, change in proportional relationship or adjustment of size should still fall within the scope of the technical content disclosed in this application without affecting the efficacy and purpose that can be achieved by this application.

[0067] By implementing the technical solution of the on-board dynamic detection method of the track line sensor plate described in the specific embodiments of this application, the following technical effects can be achieved: (1) The on-board dynamic detection method for the track line induction plate described in the specific embodiment of this application automatically processes the three-dimensional point cloud data of the induction plate and analyzes the induction plate geometric parameter information such as height anomaly, longitudinal displacement, and misalignment, providing a reliable basis for the operation and maintenance of the induction plate, realizing the on-board dynamic detection of the rail transit induction plate, greatly improving the detection efficiency and accuracy, and comprehensively covering the detection of the appearance and geometric parameters of the induction plate; (2) The on-board dynamic detection method for track line induction plates described in the specific embodiments of this application uses a target detection network and a classification network to intelligently analyze surface cracks, missing bolts, and loose defects on the induction plates, thereby achieving end-to-end detection of surface defects on the induction plates and further improving detection efficiency and accuracy. (3) The on-board dynamic detection method for track line sensor plates described in the specific embodiments of this application utilizes a 2D camera component to collect the surface texture features of the sensor plate body and the bracket, and a 3D camera component to collect the rail contour and the three-dimensional point cloud information of the sensor plate. The method can be carried on a train to collect and process sensor plate data at high speed, thereby realizing automatic and efficient detection of sensor plate surface defects and sensor plate geometric parameters, and guiding relevant management and operating personnel to perform maintenance on the sensor plates.

[0068] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0069] The above description is only a preferred embodiment of the present application and does not constitute any formal limitation to the present application. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any technician familiar with the art can use the above-disclosed methods and technical contents to make many possible changes and modifications to the technical solution of the present application, or modify it into an equivalent embodiment with equivalent changes, without departing from the spirit and technical solution of the present application. Therefore, any simple modification, equivalent replacement, equivalent change and modification made to the above embodiments based on the technical essence of the present application without departing from the content of the technical solution of the present application still falls within the scope of protection of the technical solution of the present application.

Claims

1. A vehicle-mounted dynamic detection method for a track line induction plate, characterized in that: The following steps are involved: S1) The 2D image acquisition module (11) acquires image data I of the sensing plate (7), the bracket (8) and the bolt (9) to obtain an original sensing plate image; S2) The 3D point cloud acquisition module (12) acquires the three-dimensional point cloud data P (x, y, z, i) of the sensor plate and the left and right rails (10), where x is the X-axis coordinate, y is the Y-axis coordinate, z is the Z-axis coordinate, and i is the reflection intensity of the point cloud; S3) collecting encoder pulse signals and radio frequency tag signals deployed on the line to achieve synchronization with the line mileage, and at the same time controlling the 2D image acquisition module (11) and the 3D point cloud acquisition module (12) to collect data at equal intervals; S4) Surface defect detection of sensor plates: Intelligent detection and analysis of collected sensor plate image data to obtain defect information including sensor plate cracks, bracket cracks, missing bolts, and loose bolts; Sensor plate geometric parameter detection: The collected sensor plate point cloud data is processed to obtain the area, height, longitudinal displacement and tooth misalignment information of each sensor plate (7).

2. According to the on-board dynamic detection method of the track line induction plate of claim 1, the 3D point cloud acquisition module (12) includes three sets of camera components, characterized in that: The step S2) further comprises: S21) first obtains point cloud data and performs filtering preprocessing, searching for point clouds belonging to the same target between two frames of point clouds taken by adjacent camera components at the same mileage; S22) The rotation matrix R and translation matrix T between the two frames of point clouds are then calculated through point cloud feature registration.

3. The on-board dynamic detection method of the track line induction plate according to claim 2 is characterized in that: The point cloud feature registration in step S22) is further The following processes are included: S221) Perform preliminary coarse registration on the two point cloud data of PA and PB to obtain an initial rotation matrix R ’ and the translation matrix T ’ , perform corresponding transformation on PB to obtain PB ’ =PB*R ’ +T ’ ; Among them, PA is the point cloud data of the left camera assembly, and PB is the point cloud data of the camera assembly adjacent to the left camera assembly; S222) Sampling the point cloud PA to obtain a set of sampling points; S223) searching for the point closest to the sampling point in the point cloud PB1 to obtain a set of matching points; S224) Calculate the updated rotation matrix R based on the matching points ’ and the translation matrix T ’ , for PB ’ Perform the corresponding transformation to obtain PB ’ =PB*R ’ +T ’ ; S225) Repeat steps S222) to S224) and use the mean square error loss as the objective function to calculate the registration error. When the registration error is less than the set threshold T or the number of iterations is equal to the maximum number of iterations, the registration ends and the obtained rotation matrix R and translation matrix T are saved to achieve point cloud feature registration.

4. The onboard dynamic detection method of the track line induction plate according to any one of claims 1 to 3, wherein a photoelectric encoder (21) is installed in the wheel of the track detection train, and a radio frequency tag reader (23) is installed under the bottom of the track detection train, characterized in that: The step S3) further The following processes are included: S31) During the running of the track inspection train, the processing board (22) collects the equally spaced pulse waveforms emitted by the photoelectric encoder (21), and multiplies and divides the pulse waveforms into input sources for the sampling triggers of the 2D image acquisition module (11) and the 3D point cloud acquisition module (12), and calculates the preliminary mileage accordingly; S32) Scan the RFID tag signal at the set sampling frequency. If the RFID tag signal is sensed, the actual mileage information represented by the tag is queried in the tag ID and actual mileage comparison table, and then the actual mileage information is mapped to each pulse signal to achieve synchronization with the line mileage.

5. The on-board dynamic detection method of the track line induction plate according to claim 4 is characterized in that: In step S3), the line mileage is further corrected according to the following formula: Assume that the real mileages of two adjacent RFID tags are M1 and M2 respectively. n , the corresponding trigger pulse counts are N1, N n , then between N1 and N n The real mileage between i =M1+(N i -N1)*(M n -M1) / (N n -N1), i=1, 2, 3……n.

6. The on-board dynamic detection method for a track line induction plate according to claim 1, 2, 3 or 5, characterized in that: The intelligent detection analysis in step S4) is combined with the target detection network and the classification network to make a judgment; the defect information includes the defect category, the mileage, the coordinates of the rectangular box in the image (x, y, w, h), the actual physical size, and the defect severity level.

7. The on-board dynamic detection method of the track line induction plate according to claim 6 is characterized in that: The detection of surface defects of the induction plate in step S4) further includes the following process: S41) performing intelligent detection and analysis on the collected sensor plate image data to obtain defect information including sensor plate cracks, bracket cracks, bolt missing, and bolt loosening; S42) First, the target detection network model is used to detect cracks in the sensing plate, cracks in the bracket, missing bolts, and normal bolts. Then, the normal bolt sub-image is sent to the classification network model to determine whether it is loose.

8. The on-board dynamic detection method for a track line induction plate according to claim 7, characterized in that: The target detection network model annotates the original sensor plate image with target detection frames and divides it into a training set and a validation set. The label categories of the target detection frame annotation include sensor plate cracks, bracket cracks, bolt missing, and normal bolts, with 0, 1, 2, and 3 representing the category ID respectively. The entire dataset is divided into 80% of the samples as the training set and 20% of the samples as the validation set.

9. The on-board dynamic detection method for a track line induction plate according to claim 7 or 8, characterized in that: In the step S42), a target detection network model is constructed based on yolov10. The feature extraction part uses CSPDarknet as the backbone network, the path aggregation network as the neck network, and the context learning ability of the model is enhanced by large convolution kernels and partitioned attention. The overall efficiency of the model is improved through spatial-channel decoupling downsampling and rank-guided modules.

10. The on-board dynamic detection method for a track line induction plate according to claim 9, characterized in that: The input size of the target detection network model is 1280, and the output detection frame of the target detection network model includes the target category, target frame and confidence; the training set is loaded into the target detection network model for iterative training, and the maximum number of iterations epoch=400 is set. After training 10 epochs, the Maosic enhancement operation is turned off, and only conventional image enhancement operations including horizontal flipping, vertical flipping and chromaticity conversion are retained. A verification set test is performed every time an epoch iteration is completed, and the average accuracy mAP information of the verification set is recorded. At the same time, the model parameters with the highest average accuracy mAP during the entire iterative training process are saved to the local disk, and the file is named best.pt. The model parameters obtained by the latest round of iterative training are saved as the last.pt file; each image I to be detected is input into the network model loaded with the parameter best.pt for forward reasoning to obtain several target detection frames, and the target categories 0, 1, 2, and 3 correspond to the induction plate crack, bracket crack, bolt missing, and normal bolt, respectively; Use the target detection box with the target category of 3 to intercept the ROI region in the image I to obtain the bolt sub-image S i , i = 1, 2, 3... n, where n is the number of target detection boxes with the target category of 3, and S i are respectively sent into the classification network model to determine whether the bolt is loose; the classification network model divides the induction plate bolt sub-image data set into two categories: normal bolts and loose bolts, and the corresponding labels are 0 and 1 respectively. 80% of the normal bolts and loose bolt images are used as the classification training set, and the remaining 20% are used as the classification verification set; the classification network model loads the induction plate bolt sub-image data set for iterative training. The maximum number of iterations is set to 300, the learning rate is set to 0.001, and random rotation, horizontal flipping, vertical flipping, and chromaticity transformation image enhancement operations are adopted. The test of the verification set is carried out once every 1 epoch iteration, and the classification accuracy ACC information is recorded. The model parameters with the maximum accuracy ACC in the entire iterative training process are saved as the best.param file, and the model parameters obtained from the latest round of iterative training are saved as the last.param file; the bolt sub-image S i is input into the classification model loaded with best.param for inference to obtain the confidence conf0 belonging to category 0 and the confidence conf1 belonging to category 1. If conf0 < conf1, this bolt is loose; otherwise, it is a normal bolt.

11. The on-board dynamic detection method for a track line induction plate according to claim 1, 2, 3, 5, 7, 8 or 10, characterized in that: The point cloud processing in step S4) extracts the front, rear, left and right edge positions of each sensing plate (7) by edge detection according to the distribution characteristics of the sensing plate (7), thereby segmenting the point cloud of each sensing plate (7), and further obtaining the height information of each position of the sensing plate (7) from the coordinate z of the point cloud; the longitudinal displacement information of the sensing plate (7) is obtained by calculating the offset of the gap between two adjacent sensing plates (7) exceeding the standard value interval; the sensing plate misalignment information includes the lateral difference, height and angle deviation between adjacent sensing plates (7), the vertex of the sensing plate (7) is obtained by the boundary intersection of the sensing plate (7), the coordinate x deviation and coordinate z deviation between the corresponding vertices of the adjacent sensing plates (7) are compared to obtain the lateral and height deviations of the sensing plate (7), and then the angle deviation is solved by fitting the calculated sensing plate plane normal vector.

12. The on-board dynamic detection method of the track line induction plate according to claim 11, characterized in that: The detection of the geometric parameters of the sensing plate in step S4) further includes the following process: S411) First, the sensing plate is segmented into blocks, and the upper, lower, left, and right edge positions of the sensing plate (7) are extracted by gradient edge detection, thereby segmenting the point cloud Psb of each sensing plate (7) n , n=1, 2, 3...n; S412) Perform bilateral filtering on the point cloud Psb to eliminate noise and obtain Psb n ’ , sensor board point cloud Psb ’ The z coordinate of each point is the height value of each position of the sensor plate, and the Psb n ’ The maximum value Max and the minimum value Min of the z coordinate; S413) If the maximum value Max > the allowed super-high value, the sensor plate is judged to be abnormally super-high. Otherwise, it is determined whether the minimum value Min < the allowed super-low value. If so, the sensor plate is judged to be abnormally low. Otherwise, the sensor plate height is normal. S414) The longitudinal displacement information of the sensing plate (7) is obtained by calculating the offset of the gap between two adjacent sensing plates (7) exceeding the standard value range, d1 is the first sensing plate Psb1 ’ With the second sensor board Psb2 ’ The longitudinal displacement between the two plates, d2 is the second sensing plate Psb2 ’ With the third sensor board Psb3 ’ The longitudinal displacement between S415) If the minimum allowable displacement Offset is satisfied Min <d1 < the maximum allowable displacement Offset Max , it is determined that the longitudinal displacement of the first induction plate is normal; otherwise, it is determined that the longitudinal displacement of the first induction plate is abnormal.

13. The on-board dynamic detection method of the track line induction plate according to claim 12, characterized in that: The sensor plate misalignment information includes the lateral, height and angle deviations between adjacent sensor plates (7). The sensor plate geometric parameter detection in step S4) further includes the following process: S421) Obtain the vertices of the sensing plate through the boundary intersection of the sensing plate (7), the vertices of the first sensing plate are A1, B1, C1, D1, the vertices of the second sensing plate are A2, B2, C2, D2, and so on, each vertex includes x, y, z coordinates; S422) c1 is the lateral deviation between the first and second sensing plates, obtained by taking the absolute value of the difference between the X-axis coordinate D1(x) of D1 and the X-axis coordinate C2(x) of C2, i.e., |D1(x)-C2(x)|. c2 is the lateral deviation between the second and third sensing plates. If the lateral deviation is greater than the lateral deviation threshold Tx, it is considered to be a misaligned tooth. Otherwise, it is determined to be normal. S423) The height deviation is obtained by taking the absolute value of the difference between the coordinates z of the corresponding vertices of the adjacent sensing plates (7). The height deviations of the first sensing plate and the second sensing plate are |D1(z)-C2(z)| and |B1(z)-A2(z)|. If the height deviation is greater than the height deviation threshold Tz, it is determined to be a malocclusion. Otherwise, it is determined to be normal.

14. The on-board dynamic detection method of the track line induction plate according to claim 13, characterized in that: The step S423) further includes: The point cloud of the first sensing plate is framed by the vertices A1, B1, C1, and D1 of the first sensing plate. Then, the plane equation of the sensing plate (7) surface Ax+By+Cz+D=0 is iteratively calculated by the plane fitting algorithm based on random sampling consistency. Its normal vector is n1(A, B, C), and the normal vector n2(A, B, C) of the second sensing plate is calculated. Then, the angle deviation is solved by the vector inner product based on the fitted normal vector of the sensing plate plane. , if the angle deviation If the angle deviation is greater than the threshold, it is judged as malocclusion, otherwise it is normal.

15. The on-board dynamic detection method for a track line induction plate according to claim 1, 2, 3, 5, 7, 8, 10, 12, 13 or 14, characterized in that: The method further comprises: S5) After the detection result is obtained by calculation in step S4), the detection result information and the corresponding original data are transmitted to the ground user terminal, and the operator is notified to specify the operation plan.

Images