Track catenary parameter high-precision measurement method and system of multi-dimensional coupling algorithm

Abstract

The application relates to a track contact network parameter high-precision measurement method and system of a multi-dimensional coupling algorithm, solves the problems of insufficient precision of manual measurement, difficulty in continuous collection, hindering of operation by a bad environment, and safety risks of workers, and the method comprises the following steps: first, collecting a steel rail image, identifying rail head feature points and grading, generating a rail width reference after temperature compensation after solving an initial rail width, and establishing a space reference coordinate system in combination with inertial measurement data; then, calibrating parameters, comparing attitude angles, updating the reference when the attitude angles exceed a threshold, and otherwise compensating and filtering sensing data; subsequently, tracking a measured point of the contact network and adjusting a holder to be aligned, and finally, fusing multi-source data to solve and output three-dimensional coordinates, a guide height and a pull-out value. The application has the following effects: high-precision and automatic measurement of contact network geometric parameters is realized, safety risks and efficiency bottlenecks of manual measurement are significantly reduced, and reliable data support is provided for intelligent operation and maintenance of a railway.

Description

Technical Field

[0001] This invention relates to the field of rail transit monitoring, and in particular to a high-precision measurement method and system for track contact network parameters using a multi-dimensional coupling algorithm. Background Technology

[0002] As a core power supply component of electrified railways, the accuracy of geometric parameters such as conductor height and pull-out value directly determines the power supply stability and operational safety of trains. Therefore, these parameters need to be measured frequently and with high precision to meet the real-time and reliability requirements of railway operation and maintenance for equipment condition monitoring. With the development of electrified railways towards high speed and intelligence, traditional measurement methods are no longer suitable for the high-precision detection requirements under complex operating conditions, and more efficient multi-dimensional collaborative measurement technologies are urgently needed.

[0003] Currently, the measurement of track contact network parameters mainly relies on manual measurement techniques. This involves staff carrying equipment such as laser rangefinders and total stations to operate on-site, manually aligning the contact network measurement points and recording the data.

[0004] Regarding the aforementioned technologies, the inventors have discovered the following drawbacks: manual alignment is susceptible to visual errors and operational experience, leading to measurement deviations exceeding the requirements of high-precision operation and maintenance. Furthermore, due to limitations in manpower efficiency, only a sampling inspection mode can be adopted (such as selecting a few points every few hundred meters), making it impossible to achieve continuous data collection across the entire line. This can easily result in the omission of local parameter anomalies. Moreover, harsh environments such as rain, snow, backlight, and dust can directly hinder manual operation, causing inspection interruptions or a sharp drop in efficiency. Additionally, when workers are operating beside the track or in elevated areas, they also face safety risks such as train collisions, making it difficult to guarantee operational safety and the continuity of inspection. Summary of the Invention

[0005] To achieve high-precision, automated measurement of catenary geometric parameters, significantly reduce the safety risks and efficiency bottlenecks of manual measurement, and provide reliable data support for intelligent railway operation and maintenance, this application provides a high-precision measurement method and system for track catenary parameters using a multi-dimensional coupling algorithm.

[0006] Firstly, this application provides a high-precision measurement method for track contact network parameters using a multi-dimensional coupling algorithm, employing the following technical solution:

[0007] A high-precision measurement method for track contact network parameters using a multi-dimensional coupling algorithm, executed by a measuring device including a machine vision module, an inertial measurement unit, a laser ranging module, and a pan-tilt unit, includes the following steps:

[0008] The machine vision module acquires rail images, identifies rail head contour feature points and performs quality grading; based on the high-quality feature point set, the contour is fitted to calculate the initial rail width value, and the rail width reference parameters are generated by compensating with the rail thermal expansion coefficient and real-time rail temperature data; the inertial measurement unit is simultaneously started to acquire the initial attitude data of the measuring device, and a spatial reference coordinate system is established with the rail width reference as the scale reference and the initial attitude as the orientation reference.

[0009] The control unit rotates according to a preset angle sequence, and the laser ranging module collects corresponding angle distance data. An angle-distance mapping table is generated through a spatiotemporal correlation algorithm, forming a calibration parameter table that is bound to the track width reference and calibrated in the spatial reference coordinate system.

[0010] In the spatial reference coordinate system, the attitude angle of the measuring device relative to the track is calculated and compared with the absolute attitude angle measured by the inertial measurement unit to obtain the difference value. If the difference value exceeds the preset threshold, the track width reference is refitted and the spatial reference coordinate system and calibration parameter table are updated based on the track width reference and high-quality feature point historical data. Otherwise, ambient temperature and vibration data are collected, and preset temperature compensation coefficients and vibration noise reduction algorithms are called to compensate and filter the output data of the laser ranging module and the inertial measurement unit online to obtain compensated laser data and filtered inertial data.

[0011] The machine vision module identifies and tracks the points to be measured on the contact wire, and the trajectory is calculated by combining the motion prediction algorithm and the spatial reference coordinate system. The gimbal is adjusted to keep the laser measurement beam continuously aligned with the points to be measured, and visual tracking data is output.

[0012] By integrating compensated laser data, filtered inertial data, and visual tracking data, and using a multi-source fusion algorithm, the three-dimensional coordinates, guide height, and pull-out value parameters of the measured points on the contact wire are output.

[0013] By adopting the above technical solutions, the system utilizes multiple modules such as machine vision, inertial measurement, laser ranging, and gimbal to work together, combined with temperature and vibration compensation and attitude calibration to reduce environmental interference. It dynamically updates spatial reference and calibration parameters to ensure attitude accuracy, while ensuring that the laser is continuously aligned with the contact wire measurement point. Finally, through multi-source data fusion and calculation, it accurately outputs the three-dimensional coordinates, guide height, and pull-out value of the measurement point, effectively achieving high-precision measurement of track contact wire parameters and improving data reliability and measurement stability.

[0014] Secondly, this application provides a high-precision measurement system for track contact network parameters based on a multi-dimensional coupling algorithm, employing the following technical solution:

[0015] A high-precision measurement system for track contact network parameters using a multi-dimensional coupling algorithm includes a memory, a processor, and a program stored in the memory and executable on the processor. When the program is loaded and executed by the processor, it implements the high-precision measurement method for track contact network parameters using a multi-dimensional coupling algorithm as described in the first aspect. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a high-precision measurement method for track contact network parameters using a multi-dimensional coupling algorithm, according to an embodiment of this application.

[0017] Figure 2 This is a schematic diagram illustrating the process of acquiring rail images, identifying rail head contour feature points, and performing quality grading through a machine vision module, according to another embodiment of this application. Detailed Implementation

[0018] The present application will be further described in detail below with reference to the accompanying drawings.

[0019] Reference Figure 1 This application discloses a high-precision measurement method for track contact network parameters using a multi-dimensional coupling algorithm. The method is executed by a measuring device, which includes a machine vision module, an inertial measurement unit, a laser ranging module, and a gimbal. The method includes the following steps:

[0020] Step S100: The rail image is acquired through the machine vision module, the rail head contour feature points are identified and the quality is graded; the initial rail width value is calculated by fitting the contour based on the high quality feature point set, and the rail width reference parameters are generated by compensating with the rail thermal expansion coefficient and real-time rail temperature data; the inertial measurement unit is started simultaneously to acquire the initial attitude data of the measuring device, and a spatial reference coordinate system is established with the rail width reference as the scale reference and the initial attitude as the orientation reference.

[0021] The system includes: a machine vision module (using optical equipment and image processing technology to acquire and analyze image information, used to identify and measure the shape, size, and other features of objects); rail head contour feature points (geometric feature points on the edge of the rail head, used to determine the shape and size of the rail); quality grading (classifying the acquired feature points according to indicators such as clarity and completeness, distinguishing between high-quality and low-quality feature points); rail width reference parameters (parameters corrected for environmental factors such as temperature, based on the actual width of the rail, used as a reference for subsequent measurements); an inertial measurement unit (IMU) (a sensor that measures the three-axis attitude angles and acceleration of an object, used to determine the spatial attitude of the measuring device); and a spatial reference coordinate system (a coordinate system established with the rail width reference as the scale reference and the initial attitude as the orientation reference, used to unify the spatial position and attitude of the measuring device).

[0022] The necessary process is described below:

[0023] 1. Image Acquisition and Preprocessing: After the machine vision module acquires images of the rails, it performs preprocessing operations such as grayscale conversion and filtering to improve image quality and facilitate subsequent feature point extraction.

[0024] 2. Feature Point Extraction and Quality Grading: Edge detection algorithms (such as the Canny algorithm) are used to extract feature points of the railhead contour from the preprocessed image. Based on the edge strength, continuity, and other indicators of the feature points, the feature points are graded according to preset grading rules, and high-quality feature points are selected for subsequent calculations.

[0025] 3. Rail width reference parameter generation: Based on the selected high-quality feature point set, the rail head profile is fitted using the least squares method to calculate the initial rail width value. This is combined with the rail's thermal expansion coefficient (e.g., α = 1.2 × 10⁻⁻⁻⁴). 5 The initial rail width value is corrected using the temperature compensation formula (e.g., ΔL=α×L0×ΔT, where ΔL is the change in rail width, L0 is the initial rail width, and ΔT is the change in rail temperature) and the real-time rail temperature data (obtained through a temperature sensor) to generate the rail width reference parameters.

[0026] 4. Initial Attitude Data Acquisition and Coordinate System Establishment: The inertial measurement unit (IMU) acquires the initial attitude data of the measurement device, including three-axis attitude angles (such as pitch, roll, and yaw) and acceleration information. Using the track width reference parameter as the scale reference and the initial attitude data as the azimuth reference, a spatial reference coordinate system is established through coordinate transformation algorithms (such as Euler angle transformation) to provide a unified reference framework for subsequent measurements.

[0027] Step S200: Control the gimbal to rotate according to a preset angle sequence, and link the laser ranging module to collect the corresponding angle distance data. Generate an angle-distance mapping table through a spatiotemporal correlation algorithm to form a calibration parameter table that is bound to the track width reference and calibrated in the spatial reference coordinate system.

[0028] The components include: a gimbal: a mechanical device that precisely controls rotation angles, typically equipped with a stepper motor and motor driver, capable of rotating according to a preset angle sequence; a laser ranging module: a device that uses laser technology to measure distance, calculating distance by emitting laser pulses and measuring the round-trip time of the reflected light; a spatiotemporal correlation algorithm: an algorithm used to correlate and calibrate data collected at different times with corresponding spatial locations, ensuring data accuracy and consistency; an angle-distance mapping table: a table recording distance data measured by the laser ranging module at different rotation angles, used for subsequent spatial positioning and calibration; and a calibration parameter table: a table recording calibrated parameters, which are bound to the track width reference and calibrated in a spatial reference coordinate system for subsequent high-precision measurements.

[0029] The necessary procedures are as follows:

[0030] In step S200, the controllable pan-tilt unit is first controlled to rotate according to a preset angle sequence (e.g., rotating once every 10 degrees, from 0 degrees to 360 degrees). At each preset angle, the laser ranging module is activated to collect distance data of the target point (e.g., a specific point on the contact wire). The laser ranging module calculates the distance by emitting laser pulses and measuring the round-trip time of the reflected light, ensuring accurate distance data is obtained at each angular position. For example, at 0 degrees, the laser ranging module measures a distance of 1000 mm; at 10 degrees, the measured distance is 1005 mm. These data are recorded in real time, providing a basis for subsequent calibration.

[0031] Subsequently, a spatiotemporal correlation algorithm is applied to correlate and calibrate the distance data at different angles. This algorithm comprehensively considers both time (measurement time) and space (pan-tilt angle), transforming the distance data at each angle into calibration parameters through a specific spatiotemporal correlation function f(θ,d,t). Specifically, assuming the distance measured at angle θ is d and the measurement time is t, the algorithm calculates the calibrated distance value based on these parameters. For example, at an angle of 0 degrees, if the measured distance is 1000mm and the measurement time is t0, then the calibration parameter is f(0,1000,t0). At an angle of 10 degrees, if the measured distance is 1005mm and the measurement time is t1, then the calibration parameter is f(10,1005,t1). These calibration parameters are then recorded to form a calibration parameter table.

[0032] Finally, the generated calibration parameter table is bound to the track width reference parameters and calibrated in the spatial reference coordinate system. Using a coordinate transformation algorithm (such as Euler angle transformation), the data in the calibration parameter table is converted to the spatial reference coordinate system to ensure consistency with the track width reference.

[0033] In step S300, the attitude angle of the measuring device relative to the track is calculated in the spatial reference coordinate system, and the difference value is compared with the absolute attitude angle measured by the inertial measurement unit. If the difference value exceeds the preset threshold, the track width reference is refitted and the spatial reference coordinate system and calibration parameter table are updated based on the track width reference and high-quality feature point historical data. Otherwise, ambient temperature and vibration data are collected, and the preset temperature compensation coefficient and vibration noise reduction algorithm are called to compensate and filter the output data of the laser ranging module and the inertial measurement unit online to obtain compensated laser data and filtered inertial data.

[0034] Among them, attitude angles describe the orientation of an object in space, typically including pitch, roll, and yaw. These angles represent the rotation angles of the object on three mutually perpendicular axes. High-quality feature point historical data: Data on high-quality railhead profile feature points identified and recorded in step S100, used for subsequent rail width reference fitting and calibration. Preset temperature compensation coefficient: A coefficient used to correct the measuring equipment based on temperature changes, compensating for the impact of temperature variations on the measurement results. Vibration noise reduction algorithm: An algorithm used to reduce the interference of vibration on measurement data and improve measurement accuracy.

[0035] The process is described below:

[0036] 1. In step S300, the attitude angle of the measuring device relative to the track is first calculated in the spatial reference coordinate system. This process utilizes the track head contour feature points identified by the machine vision module to calculate the relative attitude angle of the measuring device relative to the track. The specific calculation formula is as follows:

[0037] Pitch angle: Calculated by the angle between the centerline of the railhead profile and the horizontal line. The specific formula is as follows: .

[0038] Roll angle: Calculated by the angle between the rail head profile edge line and the vertical line. The specific formula is as follows: .

[0039] in, It represents the offset of the rail head profile centerline in the horizontal direction, that is, the distance between the rail head profile centerline and the horizontal line in the horizontal direction. This indicates the vertical offset of the rail head profile centerline, that is, the vertical distance between the rail head profile centerline and the horizontal line. This indicates the offset of the rail head profile edge line in the vertical direction, that is, the distance between the rail head profile edge line and the vertical line in the vertical direction.

[0040] Yaw angle: Assuming the measuring device's direction of motion is consistent with the orbital direction, the yaw angle can be approximated as 0, or measured using other auxiliary sensors (such as a magnetometer). The specific formula is as follows: .

[0041] 2. Acquire the absolute attitude angles of the inertial measurement unit (IMU):

[0042] IMU Initialization: Start the preset inertial measurement unit (IMU) and perform initial calibration to ensure the accuracy of its measurement data.

[0043] Data Acquisition: The IMU determines the absolute attitude angles of the measuring device in space by measuring acceleration and angular velocity. The specific steps are as follows: The absolute attitude angles output by the IMU include pitch, roll, and yaw angles. These data can be acquired by the IMU's sensors and transmitted to the processing unit.

[0044] For example, the data format output by the IMU might be as follows:

[0045] Absolute pitch angle = Absolute roll angle = Absolute yaw angle = .

[0046] 3. Calculate the attitude angle difference value:

[0047] Difference Calculation: The attitude angle of the measuring device relative to the track is compared with the absolute attitude angle acquired by the IMU to calculate the attitude angle difference. The specific steps are as follows:

[0048] Pitch angle difference value: .

[0049] Roll angle difference value: .

[0050] Yaw angle difference: .

[0051] If any of these discrepancies exceeds a preset threshold, it indicates a significant change in the attitude of the measuring device, requiring recalibration. In this case, based on the track width datum and high-quality feature points recorded in historical data, the track width datum is refitted, and the spatial datum coordinate system and calibration parameter table are updated to ensure measurement accuracy. The specific steps are as follows:

[0052] Selecting high-quality feature points: Feature points with a confidence level higher than 90% and that have repeatedly appeared in the most recent 10 measurements are selected from historical data to ensure data stability and reliability. Refitting the rail head profile: Using these high-quality feature points, the rail head profile is refitted, and new rail width reference parameters are calculated.

[0053] Update the spatial reference coordinate system: Re-establish the spatial reference coordinate system based on the new track width reference parameters to ensure the accuracy and consistency of the coordinate system.

[0054] Update the calibration parameter table: Recalculate the parameters in the calibration parameter table according to the new spatial reference coordinate system to ensure the accuracy and reliability of the calibration parameters.

[0055] Conversely, if the difference is within the preset threshold range, it indicates that the attitude of the measuring device is relatively stable and recalibration is unnecessary. At this point, the system will collect ambient temperature and vibration data, which can affect measurement accuracy. To compensate for the influence of these environmental factors, the system calls a preset temperature compensation coefficient and vibration noise reduction algorithm. The specific steps are as follows:

[0056] 1. Collect ambient temperature and vibration data: Use temperature and vibration sensors to collect ambient temperature and vibration data in real time.

[0057] 2. Temperature Compensation: The output data of the laser ranging module and the inertial measurement unit are compensated based on temperature changes. The specific formula is as follows: ;in, It is the distance after compensation. It is the original measured distance. It is the temperature compensation coefficient. It is the change in temperature.

[0058] 3. Vibration noise reduction: The output data of the inertial measurement unit is filtered using a vibration noise reduction algorithm. For example, a Kalman filter can be used.

[0059] In step S400, the machine vision module identifies and tracks the contact wire measurement point, calculates the trajectory by combining the motion prediction algorithm and the spatial reference coordinate system, adjusts the pan-tilt unit to keep the laser measurement beam continuously aligned with the measurement point, and outputs visual tracking data.

[0060] Among them, the motion prediction algorithm is an algorithm used to predict the future position of the measured point to help adjust the gimbal attitude. The spatial reference coordinate system, as disclosed in step S100, is a coordinate system used to unify the spatial position and attitude of the measuring device. The laser measurement beam is a laser beam used to measure distance and needs to be continuously aligned with the measured point. The visual tracking data is the measured point position data output by the machine vision module, used for subsequent processing.

[0061] For details, please refer to steps S410 to S450.

[0062] Step S500: The compensated laser data, filtered inertial data, and visual tracking data are fused together and solved by a multi-source fusion algorithm to output the three-dimensional coordinates, guide height, and pull-out value parameters of the contact wire measured point.

[0063] Multi-source fusion algorithm: An algorithm used to integrate and process data from multiple sensors to improve the accuracy and reliability of measurement results.

[0064] The specific process described above can be found in steps S510 to S540.

[0065] Reference Figure 2The process involves acquiring rail images using a machine vision module, identifying rail head contour feature points, and performing quality grading, including:

[0066] Step S110: The lighting conditions are adaptively adjusted by the multi-source dynamic lighting device in the machine vision module, and the rail image is preprocessed by a preset denoising algorithm and distortion correction matrix to obtain a high-quality rail head area image.

[0067] The system includes a multi-source dynamic lighting device: a hardware system composed of multiple adjustable LED arrays (e.g., two side light sources on each side of the rail head and one top light source) and a light sensor. It can adjust the brightness and angle of each light source in real time according to the ambient light intensity and the reflective properties of the rail surface. Denoising algorithms are digital image processing methods used to eliminate random noise (such as sensor thermal noise and environmental interference noise) in images. Common methods include Gaussian filtering, median filtering, and non-local mean denoising. Distortion correction matrix is ​​a mathematical matrix describing the optical distortion (radial distortion and tangential distortion) of the camera lens. It is obtained through camera calibration and used to restore distorted images to their true geometric shape. High-quality rail head area image refers to an image with clear rail head contour edges, moderate contrast, no obvious noise or geometric distortion, and complete coverage of the rail head area (the rail head area accounts for no less than 60% of the image).

[0068] The specific process is described below:

[0069] 1. Dynamic supplemental lighting adjustment from multiple light sources:

[0070] When the device is started, the light sensor (sampling frequency 10Hz) collects the ambient light intensity (e.g. 500 lux-10000 lux) and the average brightness value of the rail image in real time (which is transmitted back in real time through the machine vision module).

[0071] Based on a preset brightness threshold (such as a target average brightness of 80-120 grayscale values), the control system adjusts the power of each LED light source using PWM (Pulse Width Modulation) technology.

[0072] If the ambient light is too strong (such as direct sunlight at noon, causing image overexposure), reduce the power of the top light source (e.g., from 100% to 30%) and adjust the angle of the side light source (tilt 45° → 60°) to reduce specular reflection.

[0073] If the ambient light is too weak (e.g., inside a tunnel, resulting in underexposed images), increase the power of the top light source (e.g., from 30% to 80%) and simultaneously turn on the side light source to supplement the light, ensuring that the grayscale difference between the edge of the track head and the background is ≥30.

[0074] 2. Preset noise reduction algorithm processing:

[0075] The original image (RGB format, 1920×1080 pixels) is first converted to grayscale (to preserve brightness information and simplify calculations).

[0076] Denoising algorithms for different noise types:

[0077] Gaussian filtering (convolution kernel 5×5, σ=1.2) is used to eliminate Gaussian noise introduced by the sensor (which manifests as uniformly distributed fine white dots).

[0078] Median filtering (3×3 window) is used to eliminate salt-and-pepper noise (manifested as isolated black and white dots) caused by environmental interference.

[0079] For high-noise scenes (such as rainy or snowy weather), nonlocal mean denoising (similar window 7×7, weight coefficient 0.1) is superimposed to further smooth the noise by matching the pixel values ​​of similar regions in the image, while preserving the details of the track head edge.

[0080] 3. Application of distortion correction matrix:

[0081] The distortion correction matrix was obtained in advance using the Zhang Zhengyou calibration method: using a checkerboard calibration board (6×9 squares, 20mm side length), 20 calibration images at different angles were taken within the working distance range of the measuring device (e.g., 500mm~800mm). The camera intrinsic parameters (focal length f_x=1200 pixels, f_y=1200 pixels, principal point coordinates u0=960, v0=540) and distortion coefficients (radial distortion k1=-0.03, k2=0.01, tangential distortion p1=0.002, p2=-0.001) were calculated using the least squares method to form the distortion correction matrix.

[0082] After denoising, the correction matrix is ​​applied using OpenCV's undistort function to eliminate the "bending" of the track head edges caused by lens distortion (such as straight edges becoming curved), restoring the true geometric shape of the track head.

[0083] 4. After preprocessing the image of the rail head region, the image is initially cropped by simple threshold segmentation (based on the gray difference between the rail head and the track bed, such as the gray value of the rail head being 60~150 and the track bed being <40), retaining only the ROI (region of interest) containing the rail head, thus reducing the computational load of subsequent model inference (the ROI region size is compressed to 800×400 pixels).

[0084] The final high-quality output image of the railhead area must meet the following requirements: contour edge gradient value ≥20 (clearly distinguishable), noise pixel ratio <0.5%, and geometric distortion error <0.1 pixel.

[0085] Step S120: The preprocessed image is inferred in real time by the pre-deployed improved U-Net deep learning model. Through the attention mechanism and depth-separable convolutional structure, the key feature points of the rail head contour are identified, and the confidence scores of each feature point are output synchronously.

[0086] The improved U-Net deep learning model is an improvement upon the classic U-Net architecture, enhanced by introducing an attention mechanism and a depthwise separable convolutional structure to improve its ability to recognize feature points on the railhead contour. The attention mechanism is a technique used to enhance the model's focus on key regions, improving the accuracy and efficiency of feature extraction. The depthwise separable convolutional structure is an optimized convolutional structure that separates standard convolution operations into depthwise convolution and pointwise convolution, reducing computational cost and the number of parameters, thus improving model efficiency. Key feature points are points of significant importance on the railhead contour, such as edge points and inflection points, which are crucial for subsequent calculation of rail width reference parameters.

[0087] The acquisition method and process are described below:

[0088] 1. Data Preparation: Collect a large amount of rail image data, including images under different lighting conditions, angles, and environments. Annotate these images, marking key feature points of the rail head contour, and generate labeled data for training.

[0089] 2. Model Improvement:

[0090] Introducing an attention mechanism: An attention mechanism is introduced into the key layers of the U-Net model to enhance the model's attention to the key regions of the railhead contour.

[0091] By employing channel attention and spatial attention mechanisms, the accuracy and efficiency of feature extraction can be improved.

[0092] The channel attention mechanism extracts channel features through global average pooling and global max pooling, and then generates channel weights through a shared fully connected layer and activation function to enhance the features of important channels.

[0093] The spatial attention mechanism extracts spatial features through global average pooling and global max pooling, and then generates spatial weights through shared convolutional layers and activation functions to enhance the features of important regions.

[0094] Depthwise separable convolutional structure: Using a depthwise separable convolutional structure in the U-Net model optimizes convolution operations and reduces computational cost and the number of parameters.

[0095] Depthwise separable convolution decomposes standard convolution into depthwise convolution and pointwise convolution, performing convolution operations on each input channel separately, and then merging the results through pointwise convolution.

[0096] 3. Model training:

[0097] Data preprocessing: The collected rail images are preprocessed, including normalization, cropping, and flipping, to enhance the model's generalization ability.

[0098] Training process: The improved U-Net model is trained using labeled data.

[0099] Choose an appropriate loss function, such as the cross-entropy loss function or the Dice loss function, to measure the difference between the model output and the true label.

[0100] Use optimization algorithms, such as Adam or SGD, to tune the model's parameters to minimize the loss function.

[0101] During training, data augmentation techniques (such as random cropping, rotation, flipping, etc.) are used to increase the diversity of training data and prevent overfitting.

[0102] Validation and tuning: Evaluate the model's performance on the validation set and adjust the model's hyperparameters, such as learning rate and batch size, based on the validation results to improve the model's accuracy and generalization ability.

[0103] 4. Real-time reasoning:

[0104] The trained and improved U-Net model is deployed to real-world applications to perform real-time inference on input rail images.

[0105] The model extracts image features step by step through convolutional layers, pooling layers, and upsampling layers, generates segmentation results, and identifies key feature points of the railhead contour.

[0106] 5. Post-processing: Post-process the segmentation results output by the model to extract the precise locations of key feature points. Thresholding segmentation and morphological operations (such as dilation and erosion) are used to remove noise and extract key feature points. For example, binarization thresholding is used to convert the segmentation results into a binary image, and then morphological operations are used to remove noise and extract key feature points.

[0107] 6. Output key feature points: Output the location information of the identified key feature points.

[0108] Step S130: Calculate the local gradient intensity variance of the feature points using the Sobel operator, and simultaneously calculate the Euclidean distance by comparing it with the preset standard profile of the rail head.

[0109] The Sobel operator is a gradient calculation operator used for image edge detection, capable of calculating the gradient intensity and direction of each pixel in an image. Local gradient intensity variance is a statistical measure of the change in gradient intensity around a feature point, used to evaluate the stability and reliability of the feature point. Euclidean distance measures the straight-line distance between two points in multidimensional space; here, it is used to compare the similarity between a feature point and a preset standard railhead profile. The preset standard railhead profile is a predefined standard shape of the railhead profile used for comparison with the actually detected feature points.

[0110] The process is described below:

[0111] 1. Calculate the local gradient strength variance: Use the Sobel operator to calculate the gradient for each feature point and its surrounding region, obtaining the gradient strength of each pixel. Calculate the gradient strength variance of the region surrounding the feature point to evaluate the gradient variation in that region. The smaller the gradient strength variance, the more stable the gradient variation around the feature point, and the higher the quality of the feature point.

[0112] 2. Calculate the Euclidean distance: Compare the position of each feature point with the preset standard profile of the rail head and calculate the Euclidean distance between them. The smaller the Euclidean distance, the higher the matching degree between the feature point and the standard profile, and the higher the quality of the feature point.

[0113] Step S140: Classify quality levels based on multi-indicator collaborative judgment rules: feature points with confidence scores not lower than a preset threshold, gradient strength variance not exceeding a preset threshold, and Euclidean distance not exceeding a preset threshold are classified as high-quality levels, and the rest are classified as low-quality levels, forming a high-quality and low-quality feature point label set.

[0114] In step S140, the multi-indicator collaborative judgment rule refers to comprehensively evaluating the quality of feature points using multiple key indicators. These indicators include confidence score, gradient strength variance, and Euclidean distance. Each indicator has its specific threshold; a feature point is only classified as high quality when it simultaneously meets all these threshold conditions. This comprehensive evaluation method ensures the comprehensiveness and reliability of feature point quality assessment, avoiding misjudgments that may arise from a single indicator.

[0115] In step S140, the system classifies the quality level of feature points by comprehensively evaluating their confidence scores, gradient strength variance, and Euclidean distance to the standard contour. First, the system checks whether the confidence score of each feature point reaches a preset high confidence threshold, such as 0.9, to ensure the model's recognition of feature points has high reliability. Next, the system evaluates whether the gradient strength variance of the feature point is lower than a preset low variance threshold, such as 0.05, to ensure stable gradient changes around the feature point, thereby improving the stability of the feature point. Finally, the system calculates the Euclidean distance between the feature point and the preset standard contour of the railhead and determines whether it is less than a preset distance threshold, such as 10 pixels, to ensure a high degree of matching between the feature point and the standard contour. Only feature points that simultaneously meet all three conditions are classified as high-quality; the rest are marked as low-quality, ultimately forming a label set of high- and low-quality feature points.

[0116] The initial rail width value is calculated by fitting the contour based on a high-quality feature point set, and the rail width reference parameters are generated by compensating with the rail thermal expansion coefficient and real-time rail temperature data, including:

[0117] Step S1A0: Extract a high-quality feature point set from the high- and low-quality feature point label set, call the pre-calibrated camera intrinsic and extrinsic parameter matrices, and convert the feature point pixel coordinates into three-dimensional coordinates in the physical coordinate system.

[0118] The camera intrinsic parameter matrix describes the camera's internal parameters, including focal length and optical center coordinates, and is used to convert pixel coordinates to coordinates in the camera coordinate system. The camera extrinsic parameter matrix describes the camera's position and orientation in the world coordinate system and is used to convert coordinates in the camera coordinate system to coordinates in the world coordinate system. The physical coordinate system is the actual three-dimensional spatial coordinate system used to represent the position of feature points in actual space.

[0119] The specific steps are as follows:

[0120] 1. Extracting a high-quality feature point set: Select all feature points labeled as high-quality from the high- and low-quality feature point label sets to form a high-quality feature point set. These feature points have high confidence, stable gradient strength, and high matching degree with standard contours in the image, making them more suitable for subsequent calculations.

[0121] 2. Calling the camera intrinsic and extrinsic parameter matrices: Using pre-calibrated camera intrinsic and extrinsic parameter matrices. The camera intrinsic parameter matrix includes parameters such as focal length and optical center coordinates, used to convert pixel coordinates to coordinates in the camera coordinate system. The extrinsic parameter matrix describes the camera's position and orientation in the world coordinate system, used to convert coordinates in the camera coordinate system to coordinates in the world coordinate system.

[0122] 3. Coordinate Transformation: For each high-quality feature point, its pixel coordinates are transformed into camera coordinates using the camera intrinsic matrix. Then, the extrinsic matrix is ​​used to transform the camera coordinates into 3D coordinates in the physical coordinate system. This process ensures that the coordinates of the feature points are accurately mapped from the image plane to the actual 3D space.

[0123] Step S1B0: Fit the rail head profile based on the transformed three-dimensional coordinate data, and calculate the initial rail width value by the spatial position difference of symmetrical feature points on both sides of the rail head.

[0124] Rail head profile: The geometry of the rail head, typically an inverted trapezoid or rectangle, used to calculate the rail width. Initial rail width value: The rail width value calculated based on the spatial position difference of symmetrical feature points on both sides of the rail head, without temperature compensation.

[0125] The specific process is as follows:

[0126] 1. Fitting the rail head profile: Using the transformed 3D coordinate data, the rail head profile is fitted using mathematical fitting methods (such as polynomial fitting or spline fitting). During the fitting process, symmetrical feature points on both sides of the rail head are selected to ensure that the fitted profile accurately reflects the actual shape of the rail head.

[0127] 2. Calculate the initial track width value:

[0128] From the fitted railhead profile, symmetrical feature points on both sides of the railhead are selected, and the spatial position difference of these feature points in the horizontal direction is calculated. The specific method is as follows:

[0129] For each pair of symmetrical feature points, calculate their distance in the horizontal direction.

[0130] The initial track width value is obtained by averaging multiple pairs of symmetrical feature points.

[0131] Step S1C0: Call the preset thermal expansion coefficient of this type of rail, and combine it with the real-time collected rail head temperature data to perform temperature compensation correction on the initial rail width value.

[0132] Among them, the coefficient of thermal expansion of the rail is the coefficient of expansion or contraction of the rail material due to temperature changes, usually expressed as the rate of change of length per degree Celsius. Real-time rail temperature data refers to the rail surface temperature data collected in real time by temperature sensors. Temperature compensation correction adjusts the initial rail width value based on the coefficient of thermal expansion and real-time rail temperature data to eliminate the influence of temperature changes on rail width measurement.

[0133] In step S1C0, the preset thermal expansion coefficient of this type of rail is used, and combined with the real-time collected rail head temperature data, the initial rail width value is corrected for temperature compensation. This process ensures the accuracy of rail width measurement and eliminates the influence of ambient temperature changes on the measurement results.

[0134] The specific steps are as follows:

[0135] 1. Call the rail thermal expansion coefficient: Based on the material and specifications of the rail, call the preset rail thermal expansion coefficient.

[0136] 2. Real-time rail temperature data acquisition: A temperature sensor installed near the rail head is used to acquire real-time rail head temperature data. Assume the currently acquired rail head temperature is T. current The reference temperature (usually the standard installation temperature) is T. ref .

[0137] 3. Calculate the temperature change: Calculate the difference between the current temperature and the reference temperature.

[0138] ;

[0139] 4. Perform temperature compensation correction:

[0140] Calculate the rail width compensation based on the rail's thermal expansion coefficient and temperature change: ;

[0141] Where α is the coefficient of thermal expansion of the rail. This is the initial track width value. It is the compensation amount for the track width.

[0142] Correct the initial track width value: .

[0143] Step S1D0: A filtering algorithm is used to eliminate measurement fluctuations in the compensated rail width value and generate rail width reference parameters.

[0144] The specific steps are as follows:

[0145] 1. Select a filtering algorithm: Choose an appropriate filtering algorithm based on the characteristics and requirements of the measurement data. For example, for scenarios requiring fast response, a moving average filter can be selected; for scenarios requiring high-precision smoothing, a Gaussian filter or a Kalman filter can be selected.

[0146] 2. Apply a filtering algorithm: Input the temperature-compensated track width value into the selected filtering algorithm for smoothing. For example, when using a moving average filter, calculate the average value of the data within a certain window to reduce measurement fluctuations. The specific formula is as follows:

[0147] ;

[0148] in, is the compensated track width value of the i-th measurement, and N is the window size.

[0149] 3. Generate track width reference parameters: Output the filtered track width value as the final track width reference parameter.

[0150] The machine vision module identifies and tracks the points being measured on the overhead contact line. Combining motion prediction algorithms with a spatial reference coordinate system to calculate the trajectory, the pan-tilt unit is adjusted to continuously align the laser measurement beam with the points being measured.

[0151] Step S410: The contact wire image acquired by the machine vision module is processed in real time using a target detection algorithm to identify the contact wire test points and output their pixel coordinates in the image coordinate system; at the same time, the environmental interference is suppressed by the image enhancement algorithm to ensure that the confidence of the test point identification is not lower than the preset confidence threshold; the preset camera intrinsic parameter matrix and extrinsic parameter matrix are called to transform the pixel coordinates to three-dimensional coordinates in the spatial reference coordinate system.

[0152] Among them, object detection algorithm: a computer vision technique used to identify specific objects in an image and determine their position and size. Image coordinate system: the coordinate system of pixels in an image, usually with the top left corner of the image as the origin. Pixel coordinates: the position of a pixel in an image, usually represented as (x, y). Image enhancement algorithm: an algorithm used to improve image quality, making the image more suitable for subsequent processing. Confidence threshold: a preset value used to determine the reliability of object detection results; only detection results with a confidence level higher than this threshold are accepted.

[0153] The acquisition method and process are described below:

[0154] 1. Acquire catenary images: Use a machine vision module to acquire images of the catenary in real time. Ensure that the image resolution and frame rate meet the requirements of subsequent processing.

[0155] 2. Apply target detection algorithms: Apply target detection algorithms, such as YOLO, SSD, or Faster R-CNN, to the acquired images to identify the points to be measured on the contact wire. The target detection algorithm will locate the points to be measured in the image and output their pixel coordinates in the image coordinate system.

[0156] 3. Image Enhancement Processing: To improve the accuracy of point identification, image enhancement algorithms are applied to suppress environmental interference, such as changes in lighting, shadows, and reflections. Image enhancement algorithms may include histogram equalization, contrast enhancement, and noise reduction.

[0157] 4. Ensure Confidence: Check the confidence value output by the target detection algorithm to ensure that the recognition confidence of the measured point is not lower than the preset confidence threshold. If the confidence is lower than the threshold, it may be necessary to re-acquire the image or adjust the parameters of the image enhancement algorithm.

[0158] 5. Output pixel coordinates: Output the pixel coordinates of the identified measured points.

[0159] 6. Call the camera intrinsic and extrinsic parameter matrices: Use the pre-calibrated camera intrinsic and extrinsic parameter matrices. The camera intrinsic parameter matrix includes parameters such as focal length and optical center coordinates, used to convert pixel coordinates to coordinates in the camera coordinate system. The extrinsic parameter matrix describes the camera's position and orientation in the world coordinate system, used to convert coordinates in the camera coordinate system to coordinates in the world coordinate system.

[0160] 7. Coordinate Transformation: For each identified measurement point, the camera intrinsic matrix is ​​used to convert its pixel coordinates to coordinates in the camera coordinate system. Then, the extrinsic matrix is ​​used to convert the coordinates in the camera coordinate system to three-dimensional coordinates in the physical coordinate system.

[0161] Step S420: Analyze the three-dimensional coordinate sequence of the measured point in the spatial reference coordinate system of the preceding frames based on the sliding window method, and extract motion state parameters including instantaneous velocity, acceleration and motion direction angle.

[0162] Among them, the sliding window method is a time series analysis method that analyzes local features of data by sliding a fixed-size window across the data sequence. Instantaneous velocity: the velocity at a specific moment or location, calculated as the ratio of position change to time change. Acceleration: the rate of change of velocity, representing how quickly velocity changes with time. Direction angle: the angle between the direction of an object's motion and a reference direction (such as the horizontal). Motion state parameters: parameters describing the motion state of an object, including velocity, acceleration, and direction angle.

[0163] The acquisition method and process are described below:

[0164] 1. Collect the three-dimensional coordinate sequence of the measured point: Collect the three-dimensional coordinates of the measured point in each frame from consecutive image frames to form a time series.

[0165] 2. Applying the sliding window method: The sliding window method is applied to the three-dimensional coordinate sequence of the measured points to analyze the data within the window. The window size can be set according to actual needs, and usually a window size that can contain enough data points for reliable analysis is selected.

[0166] 3. Calculate instantaneous velocity: Within the sliding window, calculate the change in three-dimensional coordinates of the measured point between two consecutive frames. Calculate the instantaneous velocity based on the coordinate change and the time interval between frames.

[0167] The formula is: Where v is the instantaneous velocity, It is the change in three-dimensional coordinates. It is a time interval.

[0168] 4. Calculate acceleration: Within the sliding window, calculate the instantaneous change in velocity. Based on the instantaneous change in velocity and the time interval, calculate the acceleration. The formula is: , where a is acceleration and Δv is the change in velocity.

[0169] 5. Calculate the motion direction angle: Based on the changes in the three-dimensional coordinates of the measured point in consecutive frames, calculate its motion direction. Use the arctangent function to calculate the direction angle. The formula is: , where θ is the direction angle, and Δy and Δx are the changes in pixel coordinates on the y-axis and x-axis, respectively.

[0170] 6. Output motion state parameters: Output the calculated instantaneous velocity, acceleration, and motion direction angle, etc., to provide data support for subsequent motion prediction.

[0171] Step S430: Adaptively match the motion prediction model based on the fluctuation characteristics of the motion state parameters, and use the matched motion prediction model to output the predicted three-dimensional coordinates of the measured point at future times in the spatial reference coordinate system.

[0172] Among them, the motion prediction model is a mathematical model used to predict the future position of an object, based on the object's historical motion data. Fluctuation characteristics refer to the characteristics of how the object's motion state parameters (such as velocity and acceleration) change over time, used to select the most suitable motion prediction model. Predicted 3D coordinates are the three-dimensional coordinates of the object's expected position at a future moment, calculated based on the motion prediction model, within a spatial reference coordinate system.

[0173] The necessary process is as follows: 1. Input historical motion data: Input the historical motion data (such as position, velocity, and acceleration) of the measured point into the selected motion prediction model. This data is usually a sequence of motion state parameters obtained from the sliding window method analysis. 2. Model prediction: Using the selected motion prediction model, predict the position of the measured point at a future time based on the historical motion data. The model outputs the predicted three-dimensional coordinates in a spatial reference coordinate system. 3. Output predicted three-dimensional coordinates: Output the predicted three-dimensional coordinates, which represent the expected position of the measured point at a future time. These predicted coordinates will be used for subsequent gimbal adjustment and measurement beam alignment.

[0174] The specific process of adaptively matching the motion prediction model based on the fluctuation characteristics of the motion state parameters can be found in steps S431 to S433, and will not be elaborated here.

[0175] Step S440: The predicted 3D coordinates are inversely calculated into predicted pixel coordinates in the image coordinate system, and the gimbal kinematic model is called to convert them into real-time adjustment values ​​for the gimbal pitch and azimuth angles.

[0176] The camera intrinsic parameter matrix describes the camera's internal parameters, including focal length and optical center coordinates, and is used to convert pixel coordinates to coordinates in the camera coordinate system. The gimbal kinematic model describes the relationship between the motion of the gimbal (controllable gimbal) and changes in its angle, and is used to convert pixel coordinates to physical adjustment values ​​of the gimbal. The pitch angle is the angle at which the gimbal rotates around its horizontal axis, used to adjust the vertical direction of the measurement beam. The azimuth angle is the angle at which the gimbal rotates around its vertical axis, used to adjust the horizontal direction of the measurement beam.

[0177] The acquisition method and process are described below:

[0178] 1. Call the camera intrinsic parameter matrix: Using the pre-calibrated camera intrinsic parameter matrix, the predicted 3D coordinates are converted into coordinates in the camera coordinate system. The camera intrinsic parameter matrix typically includes parameters such as focal length and optical center coordinates, which are used to convert 3D coordinates into normalized camera coordinates.

[0179] 2. Inverse calculation to predicted pixel coordinates in image coordinate system: The coordinates in the camera coordinate system are inversely calculated to pixel coordinates in the image coordinate system. This process is completed by inversely applying the camera intrinsic parameter matrix, ensuring that the predicted 3D coordinates can be accurately mapped onto the image plane.

[0180] 3. Application of the gimbal kinematics model: Based on the gimbal kinematics model, the pixel coordinates in the image coordinate system are converted into physical adjustment quantities of the gimbal. The gimbal kinematics model describes the relationship between the gimbal rotation angle and the image coordinates, and is used to calculate the angle that the gimbal needs to rotate.

[0181] 4. Calculate the pitch and azimuth adjustments: Based on the pixel coordinates in the image coordinate system, calculate the required pitch and azimuth adjustments for the gimbal. The specific method is as follows: If the pixel coordinates indicate that the measured point is above or below the image center, the pitch angle of the gimbal needs to be adjusted. If the pixel coordinates indicate that the measured point is to the left or right of the image center, the azimuth angle of the gimbal needs to be adjusted.

[0182] 5. Generate Adjustment Commands: Based on the calculated pitch and azimuth adjustment amounts, generate gimbal adjustment commands. These commands include the required angle and direction of gimbal rotation for precise alignment with the measured point.

[0183] 6. Output Adjustment Commands: Output the generated gimbal adjustment commands to drive the gimbal's servo motors for adjustment. These commands ensure that the gimbal can respond quickly and accurately align with the measured point.

[0184] Step S450: Based on the adjustment amount, a control command is generated to drive the servo motor of the gimbal to dynamically perform the adjustment. Combined with the real-time position deviation feedback in the spatial reference coordinate system, the laser measurement beam is continuously aligned with the measured point.

[0185] The specific process described above can be found in steps 1 to 5, and will not be repeated here.

[0186] Simultaneously suppressing environmental interference through image enhancement algorithms to ensure that the confidence level of the identified test point is not lower than a preset confidence threshold includes:

[0187] Step S411: Collect environmental data in real time through environmental sensors, identify the type of interference, including backlight, dust and rain / snow, and quantify the intensity of interference.

[0188] Among them, environmental sensors are devices used to detect surrounding environmental conditions (such as light intensity, dust concentration, humidity, etc.). Interference type refers to environmental factors that may affect image quality and the performance of the machine vision system, such as backlighting, dust, and rain / snow. Interference intensity is the quantified strength or degree of influence of the interfering factor.

[0189] The acquisition method and process are described below:

[0190] 1. Real-time environmental data acquisition: Use environmental sensors (such as light sensors, dust sensors, humidity sensors, etc.) to monitor current environmental conditions in real time. Sensor data can include light intensity, dust concentration, humidity level, temperature, etc.

[0191] 2. Identify interference types: Identify the types of interference that may affect image quality based on sensor data.

[0192] For example, high light intensity may indicate backlighting conditions, high dust concentration may indicate dust disturbance, and a specific reading from a humidity sensor may indicate rain or snow.

[0193] 3. Quantify Interference Intensity: Quantify the identified interference types to determine their intensity levels. For example, backlight interference can be categorized as "mild," "moderate," and "severe" based on light intensity readings. Similarly, dust concentration and humidity levels can also be converted into corresponding interference intensity levels.

[0194] 4. Output interference information: Output the identified interference type and the quantified interference intensity.

[0195] Step S412: Based on the type and intensity of interference, adaptively select and apply an image enhancement algorithm.

[0196] Among them, image enhancement algorithms are used to improve image quality and make the image more suitable for subsequent processing, such as increasing contrast and enhancing edges. Adaptive selection automatically selects the most suitable image enhancement algorithm based on real-time acquired environmental data and interference types.

[0197] The acquisition method and process are described below:

[0198] 1. Analyze the type and intensity of interference: Using the type and intensity of interference identified in step S411, analyze the specific impact of the current environment on image quality.

[0199] 2. Adaptive selection of image enhancement algorithm: Based on the type and intensity of interference, the most suitable algorithm is selected from a pre-set image enhancement algorithm library. For example, for backlight interference, histogram equalization or adaptive histogram equalization (CLAHE) may be selected; for dust interference, denoising algorithms such as median filtering or bilateral filtering may be selected.

[0200] 3. Apply image enhancement algorithms: Apply the selected image enhancement algorithm to the original image to improve image quality. For example, if histogram equalization is selected, contrast will be improved by stretching the gray-level distribution of the image; if a denoising algorithm is selected, noise in the image will be reduced through filtering operations.

[0201] 4. Adjust algorithm parameters: Dynamically adjust the parameters of the image enhancement algorithm according to the specific interference situation to achieve the best enhancement effect. For example, for histogram equalization, it may be necessary to adjust the equalization range or intensity; for denoising algorithms, it may be necessary to adjust the filter window size or intensity.

[0202] 5. Output the enhanced image: Output the image after image enhancement processing for subsequent identification and tracking of the measured points.

[0203] Step S413: Calculate the objective quality indicators of the enhanced image, including contrast, sharpness, and noise level; output the recognition confidence score based on the quality indicators using a pre-trained support vector machine model.

[0204] Among them, objective quality metrics are a series of quantitative measures used to evaluate the quality of image enhancement, including contrast, sharpness, and noise levels. Support Vector Machine (SVM) model is a supervised learning algorithm used for classification and regression analysis; here, it is used to predict recognition confidence scores based on image quality metrics.

[0205] The acquisition method and process are described below:

[0206] 1. Calculate Contrast Ratio: Contrast ratio is a measure of the difference between bright and dark areas in an image. It is usually evaluated by calculating the width or standard deviation of the image histogram. The higher the contrast ratio, the greater the difference between bright and dark areas in the image, and the clearer the details.

[0207] 2. Calculate Sharpness: Sharpness reflects the clarity of edges in an image. It can be calculated using edge detection algorithms (such as the Sobel operator) or evaluated using sharpness metrics (such as the Laplacian operator). The higher the sharpness, the clearer the edges of the image.

[0208] 3. Calculate the noise level: The noise level reflects random variations in brightness or color in an image and can be evaluated by calculating the image's variance or using a specific noise assessment algorithm. The lower the noise level, the better the image quality.

[0209] 4. Evaluate the confidence score using a pre-trained SVM model: The calculated objective quality metrics (contrast, sharpness, and noise level) are used as input features and fed into the pre-trained SVM model. The SVM model outputs a confidence score based on these features, which reflects the impact of image quality on subsequent recognition tasks.

[0210] 5. Output confidence score: Output the confidence score of the SVM model for subsequent confidence threshold comparison.

[0211] Step S414: Compare the confidence score with the preset confidence threshold in real time.

[0212] Step S415: If the confidence score is greater than or equal to the preset confidence threshold, then the enhanced image is output to the machine vision module for subsequent recognition.

[0213] The process is described below:

[0214] 1. Real-time comparison of confidence scores: The confidence scores calculated in step S613 are compared with the preset confidence threshold in real time.

[0215] 2. Determine if the confidence level meets the standard: If the confidence score is greater than or equal to the preset confidence threshold, the enhanced image quality is considered to meet the requirements and can be used for subsequent recognition tasks.

[0216] 3. Output Enhanced Image: When the confidence score meets the requirements, the enhanced image is output to the machine vision module. This step ensures that only high-quality image data is used for subsequent processing and analysis, thereby improving the recognition accuracy and reliability of the entire system.

[0217] 4. Prepare for subsequent recognition: After receiving the enhanced image, the machine vision module prepares to perform subsequent target detection and feature point recognition tasks.

[0218] Step S416: If the confidence level is less than the preset confidence threshold, the enhancement algorithm parameters are dynamically adjusted, and the enhancement and evaluation are re-executed, with a maximum of 3 iterations.

[0219] Among them, enhancement algorithm parameters are adjustable settings in the image enhancement algorithm, such as filter size and contrast enhancement intensity, which affect the algorithm's performance. Re-evaluation involves re-evaluating the quality and calculating the confidence score of the enhanced image after adjusting the enhancement algorithm parameters.

[0220] The acquisition method and process are described below:

[0221] 1. Determine if the confidence score is insufficient: If the confidence score in step S414 is lower than the preset confidence threshold, it indicates that the current image enhancement effect does not meet the requirements.

[0222] 2. Dynamically adjust enhancement algorithm parameters: Based on the current image quality and confidence score, dynamically adjust the parameters of the image enhancement algorithm. For example, if the noise level is high, the intensity of the denoising algorithm can be increased; if the contrast is insufficient, the magnitude of the contrast adjustment can be increased.

[0223] 3. Re-enhance the image: Re-enhance the image using the adjusted parameters.

[0224] 4. Recalculate objective quality metrics: For the enhanced image, recalculate objective quality metrics such as contrast, sharpness, and noise level.

[0225] 5. Re-evaluate confidence scores using an SVM model: Recalculate the recognition confidence scores using a pre-trained support vector machine model with updated objective quality metrics.

[0226] 6. Loop Count Control: Set an upper limit for the number of loops (e.g., 3 times) to avoid infinite loops. If the confidence score reaches a preset threshold before reaching the upper limit, the loop stops.

[0227] Adaptive matching motion prediction models based on the fluctuation characteristics of motion state parameters include:

[0228] Step S431: Analyze the motion state parameter sequence of the preceding frames based on the sliding window method and calculate the acceleration variance.

[0229] Among them, the sliding window method is a method for analyzing time series data, which analyzes local features of the data by sliding a fixed-size window across the data sequence. Motion state parameters include velocity, acceleration, etc., which describe the motion state of an object. Acceleration variance measures the dispersion of acceleration data and is used to measure the smoothness of the object's motion.

[0230] The acquisition method and process are described as follows: 1. Collect motion state parameter sequence: Collect the motion state parameters of the measured point from consecutive image frames to form a time series. 2. Apply sliding window method: Apply the sliding window method to the motion state parameter sequence to analyze the data within the window. The window size can be set according to actual needs; usually, a window size that can contain enough data points for reliable analysis is selected. 3. Calculate acceleration: Within the sliding window, calculate the velocity change of the measured point between two consecutive frames, and then calculate the acceleration based on the velocity change and the time interval between frames. The formula is: Where a is the acceleration, Δv is the change in velocity, and Δt is the time interval. 4. Calculate the variance of acceleration: Within the sliding window, calculate the variance of all acceleration data. The variance formula is: ,in, Here, N is the variance of acceleration, N is the number of data points within the window, and a is the variance of acceleration. i 5. Output acceleration variance: Output the calculated acceleration variance to provide data support for subsequent motion prediction model matching.

[0231] Step S432: If the acceleration variance is lower than the preset stationarity threshold, then the Kalman filter algorithm is used for recursive prediction.

[0232] Among them, the Kalman filter algorithm is an efficient recursive filter capable of estimating the state of a dynamic system from a series of noisy measurements. It is widely used in signal processing and time series analysis, especially for predicting and estimating dynamic systems in noisy environments. Recursive prediction uses data from previous time steps to predict the state at the next time step; it is a time series-based prediction method. A stationarity threshold is a pre-set threshold used to determine whether the acceleration variance is within a small range, thus deciding whether to use the Kalman filter algorithm for prediction.

[0233] The acquisition method and process are described as follows: 1. Determine the acceleration variance: Check whether the acceleration variance calculated in step S431 is lower than the preset stability threshold. The preset stability threshold is set according to system requirements and experimental data to determine whether the system motion is relatively stable. 2. Match the Kalman filter algorithm: If the acceleration variance is lower than the stability threshold, it indicates that the motion of the measured point is relatively stable. At this time, the Kalman filter algorithm is selected for motion prediction. The Kalman filter algorithm is suitable for processing linear systems with Gaussian noise and can provide the optimal estimate of the system state. 3. Initialize the Kalman filter: Set the initial state of the Kalman filter, including the initial estimate and the initial error covariance matrix. The initial estimate can be based on previous measurement data, and the error covariance matrix reflects the uncertainty of the initial estimate. 4. Perform recursive prediction: Use the Kalman filter algorithm to perform recursive prediction to predict the state of the measured point at future times. Kalman filtering includes two main steps: prediction and update. In the prediction step, the next state is predicted based on the system model; in the update step, the prediction is corrected using new measurement data to obtain a more accurate state estimate. 5. Output prediction results: Output the future state of the measured point predicted by the Kalman filter algorithm.

[0234] Step S433: If the acceleration variance exceeds the preset sway threshold, then match the long short-term memory network time-series prediction model.

[0235] Among them, Long Short-Term Memory (LSTM) networks are a special type of recurrent neural network (RNN) capable of learning long-term dependencies. LSTMs are suitable for processing and predicting important events with long intervals and delays in time series. Time series forecasting models are models that predict future data points based on historical data, typically used to process data sequences arranged in chronological order. A sway threshold is a pre-set threshold used to determine whether the acceleration variance exceeds a certain value, thereby deciding whether to use the LSTM model for prediction.

[0236] The acquisition method and process are described below:

[0237] 1. Determine Acceleration Variance: Check if the acceleration variance calculated in step S431 exceeds the preset sway threshold. The sway threshold is used to identify situations where there are large fluctuations or instabilities in the motion of the measured point. 2. Match LSTM Temporal Prediction Model: If the acceleration variance exceeds the sway threshold, it indicates that the motion of the measured point may have unpredictable fluctuations. In this case, the LSTM model is selected for motion prediction. The LSTM model can handle complex nonlinear relationships and long-term dependencies, making it suitable for prediction in such situations. 3. Prepare Training Data: Collect historical motion state parameters of the measured point, such as position, velocity, and acceleration, as training data. The data needs to undergo appropriate preprocessing, such as normalization and splitting into training and test sets. 4. Train the LSTM Model: Train the LSTM model using the prepared training data. During training, adjust the model's hyperparameters, such as learning rate, batch size, and number of hidden layer units, to obtain optimal performance. 5. Perform Temporal Prediction: Use the trained LSTM model to predict the future motion state of the measured point. Based on the model's output, predict the position of the measured point at future times. 6. Output prediction results: Output the future state of the measured point predicted by the LSTM model for subsequent gimbal adjustment and measurement beam alignment.

[0238] Based on the adjustment amount generation control command, the servo motor driving the gimbal dynamically performs adjustments. Combined with real-time position deviation feedback in the spatial reference coordinate system, this ensures that the laser measurement beam is continuously aligned with the measured point, including:

[0239] Step 1: Based on the adjustment of the gimbal's pitch and azimuth angles, and by integrating the real-time position deviation of the measured point in the spatial reference coordinate system, an incremental PID control algorithm is used to calculate the real-time control quantity and generate a control command containing the target angle parameter.

[0240] The incremental PID control algorithm is a method of implementing PID control that calculates the change in the control quantity based on the error increment between the current and previous control cycles, rather than directly calculating the control quantity. Control commands are instructions sent to the servo motors, directing them on how to adjust the gimbal's attitude. Pitch and azimuth angles are the angles the gimbal needs to adjust, corresponding to vertical and horizontal rotation, respectively. Real-time position deviation is the deviation between the current position of the measured point and the target position in the spatial reference coordinate system.

[0241] The acquisition method and process are described below:

[0242] 1. Determine the angle adjustment amount: Based on the pitch and azimuth adjustment amounts calculated in the previous step, determine the angle that the gimbal needs to be adjusted.

[0243] 2. Initialize PID parameters: Set the parameters of the PID controller, including the proportional coefficient (P), integral coefficient (I), and derivative coefficient (D). These parameters affect the response speed and stability of the control system.

[0244] 3. Calculate the PID control quantity: The incremental PID control algorithm is used to calculate the change in the control quantity. The algorithm considers the current error, the error of the previous cycle, and the rate of change of the error. The formula for calculating the incremental PID control quantity is:

[0245] .

[0246] in, It is the change in the control quantity at time k. Δt is the error increment at time k, Kp, Ki, and Kd are the proportional, integral, and differential coefficients, respectively, and Δt is the sampling time.

[0247] 4. Generate Control Commands: The calculated changes in control quantities are converted into specific control commands, which are then sent to the servo motors. Control commands should include information such as the angle and speed the gimbal needs to adjust. For example, a control command could be a command containing the target pitch and azimuth angles, instructing the servo motors on how to adjust the gimbal's attitude.

[0248] 5. Output Control Commands: Output the generated control commands, ready to be sent to the servo driver. These commands will be used to adjust the servo motors of the gimbal, ensuring that the laser measurement beam can be accurately aligned with the measured point.

[0249] Step 2: Send the control commands to the servo driver of the gimbal via the preset servo communication protocol.

[0250] Servo communication protocol: a standardized communication method used to transmit control commands to the servo motor, ensuring that control commands are accurately sent to the servo driver. Servo driver: an electronic device used to receive control commands and drive the servo motor to perform actions according to the commands.

[0251] The specific steps are as follows:

[0252] 1. Prepare control commands: Based on the control commands generated in step 1, these commands contain information such as the angle and speed that the gimbal needs to be adjusted. Ensure that the format of the control commands conforms to the requirements of the servo driver.

[0253] 2. Select the servo communication protocol: Use a preset servo communication protocol, such as RS-232, RS-485, CAN bus, or Ethernet, to ensure that control commands can be reliably transmitted to the servo drive. The selected communication protocol should be chosen based on the servo drive's interface and the actual application environment.

[0254] 3. Sending Control Commands: Using the selected communication protocol, control commands are sent from the control system to the servo drive of the gimbal. This ensures data integrity and accuracy during communication, preventing data loss or errors.

[0255] 4. Confirm Command Sending: After sending the control command, confirm that the servo drive has received it. This can be done by checking the communication status or receiving feedback signals from the servo drive. If a communication error is found or the command is not received correctly, the control command needs to be resent.

[0256] Step 3: After parsing the instructions, the servo driver drives the servo motor to rotate according to the preset acceleration and deceleration curve.

[0257] Acceleration / deceleration curve: The speed change curve of the servo motor during startup and shutdown, used to smooth the acceleration and deceleration process of the motor and reduce mechanical shock.

[0258] The specific steps are as follows: 1. Parse control commands: After receiving the control commands, the servo driver parses the command content and extracts parameters such as the target pitch angle and azimuth angle. Ensure the parsing process is accurate to correctly execute subsequent actions. 2. Drive the servo motor: Based on the parsed control commands, the servo driver drives the servo motor to rotate according to a preset acceleration / deceleration curve. The acceleration / deceleration curve helps reduce mechanical shock during motor start-up and shutdown, extending equipment life and improving control accuracy.

[0259] Step 4: The actual angle position of the gimbal is collected in real time by the motor encoder, and the position deviation data under the spatial reference coordinate system is compared with the target angle parameters in a closed loop to calculate the angle error.

[0260] The specific steps are as follows: 1. Real-time acquisition of angle position: The actual angle position of the gimbal is acquired in real time using a motor encoder, including pitch and azimuth angles. The high-precision measurement of the encoder ensures the accuracy and reliability of the gimbal position. 2. Acquisition of position deviation data: Real-time position deviation data of the measured point is acquired from the spatial reference coordinate system. This data reflects the deviation between the current position of the measured point in space and the target position. 3. Closed-loop comparison: The actual angle position fed back by the encoder is compared with the target angle parameters to calculate the angle error. The specific formula is as follows: Angle error = Target angle − Actual angle; 4. Determining whether the error is within the tolerance range: Check whether the calculated angle error is within the preset tolerance range. The preset tolerance range is set according to the actual measurement accuracy requirements. For example, the tolerance range for pitch and azimuth angles can be set to ±0.1 degrees respectively. 5. Outputting angle error: If the angle error is within the preset tolerance range, it means that the gimbal has been accurately aligned with the measured point, and the angle error can be output and monitoring can continue. If the angle error exceeds the preset tolerance range, further adjustments are required.

[0261] Step 5: If the angle error exceeds the preset tolerance range, recalculate the PID control quantity based on the latest position deviation, generate a secondary adjustment command, and repeat steps 2 to 4 until the laser measurement beam is precisely aligned with the measured point and the position deviation in the spatial reference coordinate system is less than the preset threshold. If the angle error exceeds the preset tolerance range, recalculate the PID control quantity based on the latest position deviation. Use an incremental PID control algorithm to calculate the change in control quantity based on the current error, the error of the previous cycle, and the rate of change of the error.

[0262] By fusing compensated laser data, filtered inertial data, and visual tracking data, and using a multi-source fusion algorithm, the output parameters include the three-dimensional coordinates, guide height, and pull-out value of the measured points on the contact wire.

[0263] Step S510: The real-time collected ambient temperature and vibration data are time-stamped and synchronized with the compensated laser data, filtered inertial data, and visual tracking data to ensure that the multi-source data are aligned in the time dimension.

[0264] The specific steps are as follows: 1. Collect multi-source data: Real-time acquisition of ambient temperature and vibration data. Acquisition of compensated laser ranging data. Acquisition of filtered inertial measurement unit (IMU) data.

[0265] 1. **Data Acquisition and Time Stamping:** Acquire visual tracking data output from the machine vision module. 2. **Timestamping:** Add a precise timestamp to each data point to ensure accurate recording of the acquisition time. The timestamp accuracy should be high enough to meet the requirements of high-precision measurement. 3. **Timestamp Alignment:** Align the timestamps of all data to ensure that data from different sensors are aligned in the time dimension. The specific method is as follows: Select a reference timestamp, usually the timestamp from the machine vision module, as its acquisition frequency is typically high. Adjust the timestamps of data from other sensors to align them with the reference timestamp. If the timestamps do not perfectly match, adjustments can be made using interpolation or time correction methods. 4. **Data Synchronization Verification:** Verify the synchronization of the data after timestamp alignment to ensure that all data are aligned in the time dimension. Synchronization can be verified by calculating the differences between timestamps, ensuring that the differences are within the allowable error range.

[0266] Step S520: The synchronized multi-source data is uniformly converted to a spatial reference coordinate system. The Kalman filter algorithm is used to remove outliers in the data. Through the aforementioned preset temperature compensation coefficient and vibration noise reduction algorithm, the preprocessed data error is controlled within the preset data error threshold.

[0267] The necessary processes are as follows: 1. Data Conversion to a Spatial Reference Coordinate System: The synchronized multi-source data (including compensated laser data, filtered inertial data, and visual tracking data) are uniformly converted to a spatial reference coordinate system. This step ensures that all data are represented in the same coordinate system, facilitating subsequent fusion processing. A preset coordinate transformation matrix is ​​used to transform each type of data from its original coordinate system to the spatial reference coordinate system. For example, for laser ranging data, the extrinsic parameter matrix of the laser ranging module is used for conversion; for inertial data, the extrinsic parameter matrix of the IMU is used; and for visual tracking data, the intrinsic and extrinsic parameter matrices of the camera are used. 2. Kalman Filtering to Remove Outliers: The Kalman filter algorithm is applied to process the converted data and remove outliers. The Kalman filter provides the best estimate of the system state through prediction and update steps, combining prior knowledge and new measurements. The state vector and covariance matrix of the Kalman filter are initialized. The state vector typically includes information such as position and velocity, while the covariance matrix represents the uncertainty of the state estimate. At each time step, the state transition matrix is ​​used to predict the next state, and then the state estimate is updated based on the new measurement data. Outliers are identified and removed by comparing the predicted and measured values. 3. Secondary optimization using temperature compensation coefficient and vibration denoising algorithm: The Kalman-filtered data is optimized again by applying preset temperature compensation coefficient and vibration denoising algorithm. This step further reduces the impact of environmental factors on the data and improves the accuracy of the data. Temperature compensation coefficient is used to compensate for the temperature of the data to ensure that the data is not affected by temperature changes. For example, for laser ranging data, the ranging result is adjusted according to the current ambient temperature and the preset temperature compensation coefficient. Vibration denoising algorithm is applied to reduce the interference of vibration on the data. For example, a low-pass filter or adaptive filter is used to remove high-frequency vibration noise. 4. Controlling data error within preset threshold: The data error after preprocessing is evaluated to ensure that the error is within the preset data error threshold. The preset threshold is set according to the measurement accuracy requirements. For example, the position error can be set to ±1mm, and the velocity error can be set to ±0.01m / s. If the data error exceeds the preset threshold, readjust the temperature compensation coefficient and the parameters of the vibration noise reduction algorithm, or reapply the Kalman filter algorithm until the data error meets the requirements.

[0268] Step S530: Based on the sensor accuracy parameters and historical performance data of the machine vision module, inertial measurement unit, and laser ranging module, and combined with real-time ambient temperature and vibration data, the weight coefficients of each data source are dynamically allocated. The weighted least squares method is used to fuse and solve the preprocessed data to obtain the optimal three-dimensional coordinate estimate of the contact wire measured point, ensuring that the fusion accuracy meets the preset fusion accuracy requirements.

[0269] Weighted least squares is a mathematical optimization technique used to find the optimal function match for data in the presence of errors. Fusion accuracy refers to the accuracy of the fused data results, meeting preset accuracy requirements.

[0270] The necessary process is described as follows: 1. Evaluate sensor accuracy and performance: Evaluate the accuracy and historical performance data of sensors such as machine vision modules, inertial measurement units (IMUs), and laser ranging modules. 2. Dynamically assign weight coefficients: Dynamically assign weight coefficients to each sensor data based on the evaluation results and real-time ambient temperature and vibration data. The weight coefficients reflect the relative reliability and importance of each sensor data in the fusion process. 3. Apply weighted least squares method: Use the weighted least squares method for data fusion to solve for the optimal 3D coordinate estimate. The weighted least squares method considers the weights of different sensor data to minimize the sum of squared weighted errors. The specific formula is as follows: ;in, It is the optimal three-dimensional coordinate estimate. It is the measurement data of the i-th sensor. It is a measurement matrix. It is the measurement error covariance matrix. These are weighting coefficients. 4. Adjust weighting coefficients: Adjust the weighting coefficients based on real-time environmental changes and sensor performance to optimize fusion accuracy. For example, if a sensor performs poorly in a specific environment, its weighting coefficient can be reduced. 5. Verify fusion accuracy: Check whether the fused data meets the preset fusion accuracy requirements. If it does not meet the requirements, it may be necessary to readjust the weighting coefficients or improve the data fusion algorithm. 6. Output fusion results: Output the optimal 3D coordinate estimate after fusion for use in subsequent steps.

[0271] Furthermore, when dynamically allocating weighting coefficients based on sensor accuracy and historical performance data, a dynamic correction mechanism for laser ranging based on train airflow disturbance is also included, specifically:

[0272] Step S531, Real-time acquisition of airflow data and judgment of disturbance level: Real-time acquisition of airflow speed and direction data at the measurement point using a preset miniature wind speed and direction sensor at a sampling frequency of ≥100Hz.

[0273] Step S532: Based on the preset train speed-airflow disturbance correlation model, the current airflow disturbance is divided into three levels: weak (airflow speed ≤ 2m / s), moderate (airflow speed 2-5m / s), and strong (airflow speed > 5m / s).

[0274] Step S533, Laser ranging dynamic correction model call: Match correction model for different disturbance levels. For weak disturbances, use the linear compensation formula correction value = original value × (1 + k1 × ΔV) (k1 is the linear compensation coefficient, ΔV is the change in airflow velocity).

[0275] Medium disturbances are compensated using a quadratic polynomial model. The corrected value is equal to the original value × (1 + k2 × ΔV + k3 × ΔV²) (where k2 and k3 are polynomial coefficients).

[0276] Strong disturbances trigger the LSTM airflow interference prediction correction model. Input real-time airflow data, train position and original laser ranging value, and output the predicted correction value with the prediction error ≤ 10% of the laser ranging module accuracy.

[0277] Step S534, Correction data verification and dynamic weight adjustment: Perform spatiotemporal correlation verification between the corrected laser ranging data and the inertial measurement unit attitude data with a time stamp synchronization of ≤1ms, and calculate the ranging deviation before and after correction; if the deviation is >15% of the accuracy of the laser ranging module, adjust the weight coefficient of the correction model with a step size of 0.05, and iterate the correction a maximum of 3 times until the deviation meets the requirements.

[0278] Step S535, data fusion weight optimization: In the weighted least squares fusion, the weight coefficient of the laser ranging data after airflow correction is increased according to the disturbance level, with a 5% increase for weak disturbance, a 10% increase for moderate disturbance, and a 15% increase for strong disturbance, to ensure that the fusion accuracy meets the preset requirements (such as ±0.5mm).

[0279] Step S540: Based on the fused 3D coordinate data, calculate the guide height and pull-out value through the geometric model, and output a comprehensive detection result including the 3D coordinates, guide height, and pull-out value geometric parameters.

[0280] The acquisition method and process are described below:

[0281] 1. Obtain the fused three-dimensional coordinate data: Obtain the fused three-dimensional coordinate data from step S530. These data represent the spatial position of the measured points of the contact wire.

[0282] 2. Establish a geometric model: Establish a mathematical model based on the geometric relationship between the track and the overhead contact line. The model should be able to calculate the guide height and pull-out value based on the given three-dimensional coordinates.

[0283] 3. Calculate the guide height: Using a geometric model, extract the Z-axis (vertical direction) coordinate values ​​from the three-dimensional coordinate system to calculate the guide height. The guide height is the vertical distance from the contact wire to the track plane.

[0284] 4. Calculate the pull-out value: Using the same geometric model, extract the X-axis (horizontal direction) and Y-axis (horizontal direction) coordinate values ​​from the three-dimensional coordinate system, and calculate the pull-out value. The pull-out value is the horizontal distance between the contact wire and the centerline of the track.

[0285] 5. Generate comprehensive test results: Integrate the calculated 3D coordinates, guide height, and pull-out values ​​into a comprehensive test result.

[0286] Based on the fused 3D coordinate data, the guide height and pull-out value are calculated using a geometric model, including:

[0287] Step S541: Based on the fused three-dimensional coordinate data and combined with the track width reference parameters, determine the track center and track surface reference, calculate the initial value of the guide height using the preset vertical projection formula, and calculate the initial value of the pull-out value using the preset horizontal distance formula.

[0288] Calculate guide height: The guide height is calculated using a preset vertical projection formula. Typically, if the orbital plane is defined as Z=0, the formula for calculating the guide height is: H=|Z|; where Z is the vertical coordinate value in the fused 3D coordinate data.

[0289] Calculate pull-out value: The pull-out value is calculated using a preset horizontal distance formula. Assuming the track centerline is at X=0 and Y=0, the pull-out value can be obtained by calculating the Euclidean distance: Where X and Y are the horizontal coordinate values ​​in the fused 3D coordinate data.

[0290] Step S542: Based on the pre-output predicted three-dimensional coordinates of the measured point at the future time, the deviation is calculated by combining the real-time three-dimensional coordinates under the spatial reference coordinate system, and the guide height and pull-out value are dynamically corrected.

[0291] The acquisition method and process are described below:

[0292] 1. Calculate the deviation: Obtain the predicted three-dimensional coordinates of the measured point at future times. ; Obtain the real-time three-dimensional coordinates of the measured point at the current moment. Calculate the deviation: ; ; .

[0293] 2. Dynamic correction of guide height: The initial value of the guide height is corrected using the deviation ΔZ. : ,in, .

[0294] 3. Dynamic correction of pull-out value: The initial value of the pull-out value is corrected using deviations ΔX and ΔY. : .in, .

[0295] 4. Verify the corrected value: Check the corrected guide height. and pull value Is it within a reasonable range? If the corrected value exceeds the preset reasonable range, it may be necessary to re-evaluate the prediction model or adjust the correction algorithm.

[0296] Step S543: Based on real-time ambient temperature and vibration data, apply a preset temperature compensation coefficient to correct the guide height value, and apply a preset vibration noise reduction algorithm to correct the pull-out angle parameter.

[0297] The process is briefly described as follows:

[0298] 1. Apply temperature compensation coefficient: Based on the real-time ambient temperature, the guide height value is corrected using a preset temperature compensation coefficient. The correction formula is as follows:

[0299] Where α is the temperature coefficient and ΔT is the temperature change. : The compensated conduction height value, that is, the corrected conduction height value after taking into account the effect of temperature change; : Corrected guide height value.

[0300] 2. Apply vibration noise reduction algorithm: Use a preset vibration noise reduction algorithm to correct the pull-out value in order to eliminate the influence of vibration.

[0301] Correction formula: ;in, Represents the vibration noise reduction function. : This indicates the pull-out value after vibration noise reduction processing, that is, the pull-out value after eliminating the influence of vibration; This indicates the pull-out value that has been dynamically adjusted.

[0302] Step S544: Compare the current guide height and pull-out value with the preset standard value in real time, and calculate the absolute deviation.

[0303] The process is described as follows: 1. Obtain the current calculation results: Obtain the corrected guide height Hcorrected and pull-out value Lcorrected from step S543. 2. Obtain historical data or standard values: Obtain historical guide height and pull-out value data for comparison, or preset standard values. 3. Calculate the absolute deviation: Calculate the absolute deviation between the current measurement value and the historical data or standard value.

[0304] In step S545, if the deviation exceeds the preset threshold, the exception handling mechanism is triggered.

[0305] Step S546, conversely, the calculated guide height and pull-out value are used as the actual output calculation results.

[0306] The anomaly handling mechanism includes: Step S545.1, retrieving real-time coordinates, predicted coordinates, and environmental data, combining the recognition confidence of the machine vision module and the attitude fluctuation value of the inertial measurement unit, and determining the anomaly type through a preset decision tree model. Anomaly types include data acquisition deviation, motion prediction inaccuracy, and environmental interference overload. Here, attitude fluctuation value refers to the amount of change in device attitude detected by the inertial measurement unit. The decision tree model is a tree-structured model used for classification and decision-making to determine the anomaly type.

[0307] The process is briefly described as follows: 1. Data Collection: Collect real-time coordinates, predicted coordinates, environmental data, and identify confidence levels and attitude fluctuation values. 2. Application of Decision Tree Model: Use a decision tree model to determine the anomaly type based on the collected data. The decision tree model determines the anomaly type through a series of judgment conditions. 3. Output Anomaly Type: Output the determined anomaly type to provide a basis for subsequent corrections.

[0308] Step S545.2: Adopt and execute a correction strategy matching the anomaly type. The correction strategy refers to the corresponding corrective measures taken for different anomaly types. The process is briefly described and the necessary formulas are as follows: 1. Matching the correction strategy: Select the appropriate correction strategy based on the anomaly type. 2. Executing the correction: For data acquisition deviations, recalibrate the sensor or adjust the data acquisition method. For inaccurate motion prediction, retrain or select a more accurate prediction model. For environmental interference overload, apply an environmental compensation algorithm to adjust the measured value: Compensated measured value = Measured value × (1 + Compensation coefficient × ΔT). 3. Outputting the correction result: Output the corrected measurement result for use in subsequent steps.

[0309] Step S545.3 involves re-synchronizing the timestamps, transforming the coordinates, and weighting the fusion of the multi-source data after the correction strategy is implemented. The corrected 3D coordinates are then calculated, and the elevation and pull-out values ​​are recalculated. Specifically, weighted fusion combines data from different sensors, performing fusion calculations based on their weights and accuracy. Re-weighted fusion uses weighted least squares to re-fuse the data.

[0310] Step S545.4: Compare the current guide height and pull-out value calculation results with historical data or standard values ​​in real time, and calculate the absolute deviation.

[0311] In step S545.5, if the absolute deviation is less than or equal to the preset threshold, the correction is considered successful; otherwise, the correction strategy is re-executed up to the preset number of times.

[0312] In step S545.6, if the correction is successful, the corrected guide height and pull-out values ​​are output as the final measurement results; otherwise, an anomaly log is recorded, including the anomaly type, processing procedure, and deviation data.

[0313] Exception Log: Records relevant information when an exception occurs, including the exception type, handling process, and measurement results.

[0314] Based on the same inventive concept, embodiments of the present invention provide a high-precision measurement system for track contact network parameters using a multi-dimensional coupling algorithm, including a memory and a processor. The memory stores data that can be run on the processor to implement the following... Figures 1 to 2 The procedure for the method shown.

[0315] The embodiments described in this specific implementation are preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A multi-dimensional coupling algorithm track catenary parameter high-precision measurement method, characterized in that, The method is executed by a measuring device comprising a machine vision module, an inertial measurement unit, a laser ranging module, and a gimbal, and comprises the following steps: Collecting a rail image through the machine vision module, identifying rail head profile feature points, and performing quality grading; fitting a profile based on a high-quality feature point set to calculate an initial rail width value, combining a rail thermal expansion coefficient and real-time rail temperature data to compensate for generating a rail width reference parameter; simultaneously starting the inertial measurement unit to collect initial attitude data of the measuring device, establishing a spatial reference coordinate system with the rail width reference as the scale reference and the initial attitude as the azimuth reference; Controlling the gimbal to rotate according to a preset angle sequence, and simultaneously controlling the laser ranging module to collect distance data corresponding to the angles, generating an angle-distance mapping table through a space-time correlation algorithm, and forming a calibration parameter table that is bound to the rail width reference and calibrated in the spatial reference coordinate system; In the spatial reference coordinate system, the attitude angle of the measuring device relative to the track is calculated, and the difference value between the absolute attitude angle measured by the inertial measurement unit is compared; if the difference value exceeds a preset threshold, the rail width reference is refitted based on the rail width reference and historical data of high-quality feature points, and the spatial reference coordinate system and the calibration parameter table are updated; otherwise, environmental temperature and vibration data are collected, a preset temperature compensation coefficient and a vibration denoising algorithm are called to compensate and filter the output data of the laser ranging module and the inertial measurement unit online, and compensated laser data and filtered inertial data are obtained; Identifying and tracking the measured points of the overhead contact line through the machine vision module, calculating the trajectory in combination with a motion prediction algorithm and the spatial reference coordinate system, adjusting the gimbal to continuously align the laser measurement beam with the measured points, and outputting visual tracking data; Fusing the compensated laser data, the filtered inertial data, and the visual tracking data, calculating through a multi-source fusion algorithm, and outputting the three-dimensional coordinates, the height, and the pull-out value parameters of the measured points of the overhead contact line.

2. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 1, characterized in that, Collecting a rail image through the machine vision module, identifying rail head profile feature points, and performing quality grading includes: Adaptively adjusting the lighting conditions through a multi-light source dynamic light supplementing device in the machine vision module, pre-processing the rail image using a preset denoising algorithm and a distortion correction matrix to obtain a high-quality rail head region image; Real-time inference of the pre-processed image through a preset improved U-Net deep learning model, identification of key feature points of the rail head profile through an attention mechanism and a depth separable convolution structure, and synchronous output of confidence scores of each feature point; Calculating the local gradient intensity variance of the feature points through a Sobel operator, and calculating the Euclidean distance in combination with a preset rail head standard profile; Dividing the quality levels based on multi-index collaborative judgment rules: dividing the feature points with a confidence score not lower than a preset threshold, a gradient intensity variance not exceeding a preset threshold, and a Euclidean distance not exceeding a preset threshold into a high-quality level, and the rest into a low-quality level, to form a high-quality and low-quality feature point label set.

3. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 2, characterized in that, Extracting a high-quality feature point set from the high-quality and low-quality feature point label set, calling a pre-labeled camera intrinsic parameter matrix and an extrinsic parameter matrix to convert the feature point pixel coordinates into three-dimensional coordinates in a physical coordinate system; ​ Fitting the rail head profile based on the converted three-dimensional coordinate data, and calculating the initial rail width value through the spatial position difference of the symmetrical feature points on both sides of the rail head; Calling the preset steel rail thermal expansion coefficient, combining the real-time collected rail head temperature data, and performing temperature compensation correction on the initial rail width value; Using a filtering algorithm to eliminate the measurement fluctuations of the compensated rail width value, and generating the rail width reference parameter.

4. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 1, characterized in that, Through the machine vision module, the measured points of the overhead contact system are recognized and tracked, and the trajectory is calculated in combination with the motion prediction algorithm and the spatial reference coordinate system. The gimbal is adjusted to continuously align the laser measurement beam with the measured points, including: Using a target detection algorithm to process the overhead contact system images collected by the machine vision module in real time, identifying the measured points of the overhead contact system and outputting their pixel coordinates in the image coordinate system; simultaneously suppressing environmental interference through an image enhancement algorithm to ensure that the recognition confidence of the measured points is not less than the preset confidence threshold; calling the preset camera intrinsic and extrinsic parameter matrices to convert the pixel coordinates to three-dimensional coordinates in the spatial reference coordinate system; Based on the sliding window method, the three-dimensional coordinate sequences of the measured points in the spatial reference coordinate system of the previous frames are analyzed, and motion state parameters including instantaneous speed, acceleration, and motion direction angle are extracted; According to the fluctuation characteristics of the motion state parameters, the motion prediction model is adaptively matched, and the matched motion prediction model is used to output the predicted three-dimensional coordinates of the measured points at future time in the spatial reference coordinate system; The predicted three-dimensional coordinates are converted back to predicted pixel coordinates in the image coordinate system, and the real-time adjustment amount of the gimbal pitch angle and azimuth angle is converted by calling the gimbal kinematics model; Based on the adjustment amount, control instructions are generated to drive the servo motor of the gimbal to dynamically execute the adjustment, and combined with the real-time position deviation feedback in the spatial reference coordinate system, the laser measurement beam continuously aligns with the measured points.

5. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 4, characterized in that, Simultaneously suppressing environmental interference through an image enhancement algorithm to ensure that the recognition confidence of the measured points is not less than the preset confidence threshold includes: Real-time collection of environmental data through environmental sensors to identify interference types, including backlight, dust, and rain and snow, and to quantify interference intensity; Based on the interference type and intensity, adaptively select and apply the image enhancement algorithm; Calculate the objective quality indicators of the enhanced image, including contrast, sharpness, and noise level; output the recognition confidence score based on the quality indicators through a pre-trained support vector machine model; Real-time comparison of the confidence score with the preset confidence threshold; If the confidence score is greater than or equal to the preset confidence threshold, output the enhanced image to the machine vision module for subsequent recognition; If the confidence is less than the preset confidence threshold, dynamically adjust the enhancement algorithm parameters, and re-execute the enhancement and evaluation, with a maximum of 3 cycles.

6. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 4, characterized in that, Adaptive matching of the motion prediction model according to the fluctuation characteristics of the motion state parameters includes: Based on the sliding window method, the motion state parameter sequences of the previous frames are analyzed, and the acceleration variance is calculated; If the acceleration variance is lower than the preset stationary threshold, the Kalman filtering algorithm is matched for recursive prediction; If the acceleration variance exceeds the preset shaking threshold, the long-short term memory network time series prediction model is matched.

7. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 4, characterized in that, Based on the adjustment amount, a control instruction is generated to drive the servo motor of the holder to dynamically execute the adjustment, and the real-time position deviation feedback under the space reference coordinate system is combined to continuously align the laser measurement beam to the measured point, including: Step 1: Based on the adjustment amount of the holder pitch angle and azimuth angle, the real-time position deviation of the measured point under the space reference coordinate system is fused, and an incremental PID control algorithm is used to calculate the real-time control amount to generate a control instruction containing the target angle parameter; Step 2: The control instruction is sent to the servo driver of the holder through the preset servo communication protocol; Step 3: After the servo driver parses the instruction, the servo motor is driven to rotate according to the preset acceleration and deceleration curve; Step 4: The actual angle position of the holder is collected in real time through the motor encoder, and the position deviation data under the space reference coordinate system are combined with the target angle parameter for closed-loop comparison to calculate the angle error; Step 5: If the angle error exceeds the preset tolerance range, the PID control amount is recalculated based on the latest position deviation, a secondary adjustment instruction is generated, and steps 2 to 4 are repeated until the laser measurement beam accurately aligns the measured point, and the position deviation under the space reference coordinate system is less than the preset threshold.

8. The multi-dimensional coupling algorithm track catenary parameter high-precision measurement method according to claim 4, characterized in that, Fusing compensated laser data, filtered inertial data, and visual tracking data, and using a multi-source fusion algorithm to solve, the three-dimensional coordinates, height, and pull-out value parameters of the measured point of the catenary are output, including: The real-time collected environmental temperature and vibration data are time-stamped synchronized with the compensated laser data, filtered inertial data, and visual tracking data to ensure that the multi-source data are aligned in the time dimension; The synchronized multi-source data are converted to the space reference coordinate system, the Kalman filter algorithm is used to remove outliers in the data, and the aforementioned preset temperature compensation coefficient and vibration noise reduction algorithm are used for secondary optimization to control the error of the preprocessed data within the preset data error threshold; Based on the sensor accuracy parameters and historical performance data of the machine vision module, the inertial measurement unit, and the laser ranging module, the real-time environmental temperature and vibration data are used to dynamically assign weight coefficients to each data source, and the weighted least squares method is used to fuse and solve the preprocessed data to obtain the optimal three-dimensional coordinate estimate of the measured point of the catenary, ensuring that the fusion accuracy meets the preset fusion accuracy requirements; Based on the fused three-dimensional coordinate data, the preset catenary geometric model is called to calculate the height and pull-out value, and the comprehensive detection results containing the three-dimensional coordinates, height, and pull-out value are output.

9. A high-precision measurement system for parameters of a track catenary with multi-dimensional coupling algorithm, characterized in that, A program stored on the memory and executable on the processor, which can be loaded and executed by the processor to implement the multi-dimensional coupling algorithm track catenary parameter high-precision measurement method of any one of claims 1 to 8.

Images