High-speed railway ballastless track construction measurement method

Abstract

The invention belongs to the technical field of data measurement, and discloses a high-speed railway ballastless track construction measurement method, which comprises the following steps of: carrying out internal reference and external reference calibration on a binocular camera of a track detection trolley; obtaining related data containing a binocular image, and carrying out space-time alignment processing on the related data so as to construct a prediction state vector; carrying out stereo matching on the binocular image to generate a prediction disparity map, and carrying out three-dimensional reconstruction on the prediction disparity map to obtain a track point cloud map; and extracting a track inner side point coordinate sequence and a track height program sequence from the track point cloud picture, and generating a corresponding track gauge table and a corresponding smoothness table. According to the invention, vision, three-dimensional reconstruction and inertial navigation technologies are integrated, and integrated detection of gauge, elevation and smoothness parameters is realized.

Description

Technical Field

[0001] The present invention relates to the field of data measurement technology, and more particularly to a high-speed railway ballastless track construction measurement method. Background Art

[0002] With the rapid development of high-speed railways, ballastless track has become the mainstream track structure for lines with speeds above 300 kilometers per hour due to its advantages of high smoothness, low maintenance cost and long-term stability. As the basis for train operation, the smoothness and track gauge of the track are directly related to the stability and safety of the train operation.

[0003] Currently, the measurement of track smoothness and gauge during the construction phase mostly relies on track inspection carts. However, the following problems still exist in the measurement using track inspection carts: first, the camera parameter calibration mostly adopts traditional calibration methods, the calibration process is relatively complicated, and the camera calibration error is large; second, the sensor data synchronization accuracy is insufficient, resulting in millisecond-level time deviations between binocular images, IMU attitude data and GNSS positioning data. The error accumulation is significant in dynamic measurement, making it difficult to meet the high-precision defect positioning requirements; finally, the binocular image processing mostly adopts traditional stereo matching algorithms (such as SAD and SGM), which are prone to mismatching in weak texture areas of the track (such as the surface or side of the rail), resulting in sparse or distorted three-dimensional point clouds, making it impossible to accurately extract the gauge and vertical height parameters; in view of this, the present invention proposes a high-speed railway ballastless track construction measurement method to solve the above problems. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art and to achieve the above-mentioned purpose, the present invention provides the following technical solution: a high-speed railway ballastless track construction measurement method, comprising: S1. Perform internal and external calibration on the binocular camera on the track inspection vehicle to obtain the corresponding internal and external reference data; S2, using the track detection vehicle to obtain binocular images, mileage data, IMU attitude data, and GNSS positioning data and perform spatiotemporal alignment processing; S3, constructing a predicted state vector based on the mileage data, IMU attitude data and GNSS positioning data after time-space alignment; S4. Perform stereo matching on the binocular images to generate a predicted disparity map; perform three-dimensional reconstruction on the predicted disparity map based on the intrinsic parameter data, the extrinsic parameter data, and the predicted state vector to generate a track point cloud map; S5. Extracting a coordinate sequence of the inner track points and a track elevation sequence from the track point cloud image; calculating a track gauge sequence based on the coordinate sequence of the inner track points, and generating a corresponding track gauge table based on the track gauge sequence; S6. Perform vertical calculation on the IMU attitude data after time-space alignment processing to obtain the vertical displacement sequence of the track; perform weighted fusion of the track height sequence and the vertical displacement sequence to obtain the vertical height data of the track, and generate the corresponding smoothness table based on the vertical height data of the track; integrate the track gauge table and the smoothness table to generate a measurement report.

[0005] Furthermore, the method of obtaining the internal reference data includes: The track detection vehicle collects a calibration image containing the flexible checkerboard calibration plate and the track, and records the IMU calibration data during the collection process; obtains the coordinates of the checkerboard corner points in the flexible checkerboard calibration plate in the calibration image and integrates the IMU calibration data to obtain the IMU rotation matrix; A projection error function is constructed based on the coordinates of the obtained checkerboard corner points and the IMU rotation matrix. The historically calibrated intrinsic parameter matrix and distortion coefficients are updated by minimizing the projection error function to obtain the intrinsic parameter matrix and distortion coefficients. The intrinsic parameter matrix and distortion coefficients together constitute the intrinsic parameter data.

[0006] Furthermore, the method of obtaining the external reference data includes: Acquire feature point pairs and rail gauges from the calibration image; construct an extrinsic parameter optimization function based on the feature point pairs and rail gauges; iteratively optimize the extrinsic parameter optimization function using the Levenberg-Marquardt method, and obtain the corresponding baseline, rotation matrix, and translation vector when the extrinsic parameter optimization function iteratively converges to a minimum value; the baseline, rotation matrix, and translation vector together constitute the extrinsic parameter data; A binocular camera with calibrated extrinsic parameters is used to obtain N groups of test images. Test matching point pairs in the test images are obtained through a feature matching algorithm. The vertical distances from the test matching point pairs to the epipolar lines are calculated, recorded as epipolar residuals. The average of all epipolar residuals is calculated. When the average is greater than or equal to the preset pixel threshold, the extrinsic parameters need to be recalibrated.

[0007] Furthermore, the method of performing the spatiotemporal alignment process includes: measuring the cable transmission delay and the time taken for the image to be transmitted from the camera sensor to the processor; correcting the exposure timestamp of the binocular camera based on the cable transmission delay and the time taken for the image to be transmitted from the camera sensor to the processor to obtain a corrected exposure timestamp of the binocular camera; Based on the corrected binocular camera exposure timestamp, the GNSS positioning data, IMU attitude data, and mileage data are aligned with the binocular image timestamp through interpolation.

[0008] Furthermore, the method for obtaining the predicted state vector includes: The IMU attitude data and GNSS positioning data after time-space alignment are fused through a filtering algorithm to obtain the position vector, velocity vector, attitude quaternion vector and error vector of the track detection vehicle in the world coordinate system; the position vector, velocity vector, attitude quaternion vector and error vector of the IMU sensor are spliced ​​to obtain the state vector; A kinematic model is constructed and discretized to obtain a discrete-time prediction model. The state vector is used as the input of the discrete-time prediction model to obtain a motion state vector. Based on the mileage data after spatiotemporal alignment, the position vector in the motion state vector is corrected through a filtering algorithm to obtain a predicted state vector.

[0009] Furthermore, the stereo matching method includes: The binocular image is calibrated based on the distortion coefficient in the intrinsic reference data to obtain a calibrated binocular image; multi-scale feature extraction is performed on the calibrated binocular image through a lightweight convolutional neural network to obtain a multi-scale feature map; The cost of multi-scale feature maps is calculated to obtain a mixed cost volume; pixels in the mixed cost volume are selected based on a winner-takes-all strategy to obtain an initial disparity map; the initial disparity map is edge-constrained optimized using a conditional random field energy function and a parabolic interpolation function to obtain an optimized disparity map; the optimized disparity map is upsampled using a pre-built convolutional neural network to generate a predicted disparity map.

[0010] Furthermore, the method for obtaining the track point cloud map includes: The predicted disparity map is triangulated based on the obtained intrinsic parameter matrix and baseline to obtain the coordinates of the pixels in the predicted disparity map in the camera coordinate system; the coordinates of the pixels in the camera coordinate system are converted to the coordinates in the IMU coordinate system through the obtained rotation matrix and translation vector; the coordinates of the pixels in the IMU coordinate system are converted to the coordinates in the world coordinate system based on the attitude quaternion in the predicted state vector, and the coordinates of all pixels in the world coordinate system constitute a point cloud map; the point cloud map is filtered to obtain a track point cloud map.

[0011] Furthermore, the coordinate sequence of the track inner point and the track elevation sequence are obtained in the following manner: Based on the track point cloud map, slice the track along the track extension direction at a preset slice interval, extract the point cloud data at the top surface of the track in each slice, and calculate the arithmetic mean of the height coordinates based on the point cloud data at the top surface of the track. Arrange the obtained arithmetic mean values ​​in spatial order to obtain the track elevation sequence; Establish a reference coordinate system at the highest point of the track top surface in the track point cloud image, where the X axis of the reference coordinate system is along the extension direction of the track, the Y axis points to the inside of the track, and the Z axis is vertically upward along the XY plane; A feature extraction plane is constructed by moving down a preset distance along the Z axis, and a sampling window is established along the extension direction of the track at a preset slice interval. The point cloud coordinates of the inner side of the track in the feature extraction plane are obtained through the sampling window to obtain the coordinate sequence of the points on the inner side of the track.

[0012] Furthermore, the method of generating a corresponding track gauge table based on the track gauge sequence includes: The attitude quaternion vector in the predicted state vector is converted into an angle to obtain the roll angle; a gauge calculation function is constructed based on the track design superelevation value and the roll angle, and the gauge calculation function is used to calculate the coordinate sequence of the inner point of the track to obtain the gauge sequence; the gauge sequence is compared with the preset standard gauge to generate a corrected gauge sequence; and a corresponding gauge table is generated based on the gauge sequence, the corrected gauge sequence, and the position vector in the predicted state vector.

[0013] Furthermore, the vertical displacement sequence of the track is obtained by: A vertical displacement kinematic model is constructed and discretized to obtain a vertical time displacement prediction model. The IMU attitude data after spatiotemporal alignment is used as the input of the vertical time displacement prediction model to obtain vertical displacement data. The vertical displacement data is aligned with the design elevation of the track at preset intervals to generate a vertical displacement sequence.

[0014] Technical effects and advantages of the present invention: This embodiment automatically calibrates camera parameters using a flexible checkerboard calibration plate combined with inherent track features, avoiding the need for manual calibration sites in traditional calibration processes and improving calibration efficiency. By introducing IMU attitude data, errors caused by movement or deformation of the calibration plate are avoided. A synchronous error compensation model and a cubic spline interpolation algorithm are constructed to correct timestamp deviations of different data, achieving sub-millisecond spatiotemporal alignment of binocular images, IMU attitude data, GNSS positioning data, and mileage data, eliminating cumulative errors caused by time asynchrony in dynamic measurements. A lightweight MobileNetV3 network is used to extract multi-scale features, combined with an adaptive weight fusion strategy to generate a mixed cost volume. Conditional random fields and parabolic interpolation functions are used to optimize edge details of the initial disparity map, significantly improving stereo matching accuracy in weak-texture scenarios and achieving sub-millimeter 3D point cloud reconstruction of the rail surface. This provides a reliable data basis for the calculation of gauge and height parameters. The gauge table and smoothness table generated on this basis can be directly connected to construction acceptance standards and support high-precision defect location. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The figure is a schematic diagram of a high-speed railway ballastless track construction measurement method according to the present invention.

[0016] Figure 2 Schematic diagram of the process of generating the track point cloud map in the present invention. DETAILED DESCRIPTION

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Example 1

[0018] See also Figure 1 As shown, the high-speed railway ballastless track construction measurement method described in this embodiment includes: S1. Perform internal and external calibration on the binocular camera on the track inspection vehicle to obtain the corresponding internal and external reference data; S2, using the track detection vehicle to obtain binocular images, mileage data, IMU attitude data, and GNSS positioning data and perform spatiotemporal alignment processing; S3, constructing a predicted state vector based on the mileage data, IMU attitude data and GNSS positioning data after time-space alignment; S4. Perform stereo matching on the binocular images to generate a predicted disparity map; perform three-dimensional reconstruction on the predicted disparity map based on the intrinsic parameter data, the extrinsic parameter data, and the predicted state vector to generate a track point cloud map; S5. Extracting a coordinate sequence of the inner track points and a track elevation sequence from the track point cloud image; calculating a track gauge sequence based on the coordinate sequence of the inner track points, and generating a corresponding track gauge table based on the track gauge sequence; S6. Perform vertical calculation on the IMU attitude data after time-space alignment processing to obtain the vertical displacement sequence of the track; perform weighted fusion of the track height sequence and the vertical displacement sequence to obtain the vertical height data of the track, and generate the corresponding smoothness table based on the vertical height data of the track; integrate the track gauge table and the smoothness table to generate a measurement report.

[0019] Specifically, the binocular camera on the track inspection vehicle is calibrated with internal parameters to obtain the corresponding internal parameter data in the following ways: Specifically, using the standard track gauge and rail height as geometric references, a straight track section with a length greater than a preset length (e.g., 10m) and a track gauge error less than a preset track gauge error (e.g., 0.1mm) is selected as the calibration section. A flexible checkerboard calibration plate (with a grid size of 20mm) is laid longitudinally along the calibration section to ensure that the flexible checkerboard calibration plate presents multiple postures (pitch and yaw angles of ±20° each) within the field of view of the binocular camera. A non-invasive fixture is used to fix the flexible checkerboard calibration plate to ensure that the flexible checkerboard calibration plate is free of wrinkles or offset. Keeping the flexible checkerboard calibration plate within the binocular camera's field of view, the track inspection vehicle travels at a preset speed (e.g., 0.5 m / s). The binocular camera synchronously captures images containing the flexible checkerboard calibration plate and the left and right rails at 15 fps, which are recorded as calibration images (divided into left and right images). The IMU calibration data during the acquisition process is recorded. The checkerboard corner detection function (indChessboardCorners) is used to detect the checkerboard corners of the flexible checkerboard calibration plate in the calibration image. The sub-pixel corner optimization function (cornerSubPix) is used to perform sub-pixel optimization on the calibration image (e.g., to 0.01 pixel accuracy). Construct a reprojection error function, and update the historically calibrated intrinsic parameter matrix and distortion coefficient by minimizing the projection error function to obtain the intrinsic parameter matrix K and distortion coefficient D; the intrinsic parameter matrix K and distortion coefficient D together constitute the intrinsic parameter data; The function expression of the reprojection error function is: ; in; represents the reprojection error term; is the ordinal number of the checkerboard corner point in the flexible checkerboard calibration plate; Represents the observed pixel coordinates of the th checkerboard corner point of the flexible checkerboard calibration plate in the calibration image; Represents the coordinates of the corner point of the flexible checkerboard calibration plate; is the projection function, which is used to Mapped to the image plane; K is the camera's intrinsic parameter matrix (3×3, obtained from the previous calibration), which includes the camera's focal length and principal point; D is the camera's distortion coefficient (obtained from the previous calibration), which includes radial distortion and tangential distortion; represents the IMU-vision attitude alignment term; Indicates the binocular camera's angular velocity obtained by integrating the angular velocity in the IMU measurement data at time IMU rotation matrix (3×3); Represents the camera rotation matrix calculated based on the intrinsic parameter matrix, distortion coefficients, and checkerboard corner points; The time corresponding to the IMU measurement data; It is a weight coefficient used to balance the contribution of the two parts of the error. The larger the value, the more emphasis is placed on the consistency between IMU and visual posture, and the smaller the value, the more reliance is on the accuracy of visual reprojection. The value range is 0.05 to 0.2.

[0020] The methods for calibrating the binocular camera to obtain external parameter data include: The SIFT algorithm is used to extract feature points from the calibration image, and then feature point pairs are obtained; the SGM algorithm is used to obtain the track gauge of the rails in the calibration image; An extrinsic parameter optimization function is constructed based on the characteristic point pairs and the rail gauge. The function expression of the extrinsic parameter optimization function is: ; in, Represents the gauge error term, which is used to force the calibration result to conform to the standard gauge; Indicates the gauge of the rail in the calibration image; Represents the epipolar residual term, which is used to ensure the geometric consistency of binocular matching; Represents the pixel coordinates of the i-th feature point in the left image of the calibration image; Represents the pixel coordinates of the i-th feature point in the right eye image in the calibration image; i is the ordinal number of the feature point; F represents the basic matrix, which is used to describe the epipolar geometric relationship of the binocular camera. It is calculated by the binocular camera's rotation matrix R (3×3, obtained from the previous calibration), translation vector t (obtained from the previous calibration) and camera intrinsic parameters. The expression is: ,in and is the intrinsic parameter matrix of the left and right cameras in the binocular camera, is the antisymmetric matrix of the translation vector; is the weight coefficient of the epipolar residual term (range 0.1 to 0.3), which is used to balance the gauge error and geometric consistency; It is the IMU-vision alignment item, which is used to suppress the attitude drift caused by vibration; Represents the IMU rotation matrix; Represents the camera rotation matrix; The weight coefficient of the IMU-vision alignment term (range 0.03 to 0.1), used to balance the weight of sensor fusion; It should be noted that the epipolar line is determined by the extrinsic parameters (relative position and attitude) of the binocular camera. It represents the straight line on the right (left) camera image plane that contains all possible matching points corresponding to a point in the left (right) camera image, reflecting the geometric constraint relationship between the two cameras. The accuracy of the extrinsic parameter calibration can be verified by calculating the perpendicular distance from the matching point pair to the epipolar line (episode residual).

[0021] The extrinsic parameter optimization function is iteratively optimized using the Levenberg-Marquardt method. When the extrinsic parameter optimization function converges to the minimum value, the corresponding binocular camera baseline B (which is an intrinsic parameter of the camera and needs to be checked through calibration to see if it has changed), rotation matrix R, and translation vector t are obtained. The baseline B, rotation matrix R, and translation vector t together constitute the extrinsic parameter data.

[0022] Use a binocular camera with calibrated external parameters to obtain N groups of test images, and obtain at least 100 pairs of test matching points in each group of test images using a feature matching algorithm (such as SIFT, ORB) , and use the calibrated external parameters to calculate the basic matrix; for each feature point in the left image of the test image, calculate its corresponding epipolar equation in the right image. The expression of the epipolar equation is: ; Where F represents the basic matrix; calculate the test matching point pair To the extreme line The vertical distance is recorded as the epipolar residual; the average of the epipolar residuals of all test matching point pairs is calculated. If the average of the epipolar residuals is less than 0.25 pixels, it means that the extrinsic parameters meet the requirements; if the average of the epipolar residuals is greater than or equal to 0.25 pixels, the extrinsic parameters need to be recalibrated or the feature matching accuracy needs to be checked.

[0023] The sensors on the track detection vehicle are used to obtain binocular images, mileage data, IMU attitude data, and GNSS positioning data (position in the world coordinate system). The sensors on the track detection vehicle include a binocular camera, mileage meter, IMU, and GNSS. The binocular camera uses a global shutter CMOS camera (such as FLIR BlackflyS) with a resolution of 4000×3000, a frame rate of 15fps, a focal length of 12mm, a baseline of 0.3m, and is equipped with an 850nm near-infrared fill light. The IMU uses a fiber optic gyroscope (such as Hon The system uses a dual-frequency RTK receiver (e.g., Trimble R12) with multi-constellation support and a positioning accuracy of 10 mm ± 1 ppm. The binocular camera is mounted on the front of the vehicle, with its optical axis parallel to the track and a pitch angle of 15° to 45°. The IMU is fixed to the center of mass of the vehicle, and the GNSS antenna is placed on the roof. The IMU attitude data includes raw Euler angles, raw acceleration measurements, and raw angular velocity measurements. The binocular images include left and right images.

[0024] Specifically, the track inspection vehicle travels along the track to be measured, acquiring mileage data, IMU attitude data, and GNSS positioning data. The PPS (pulse per second) signal from the GNSS triggers the binocular camera to expose the track, thereby acquiring a binocular image of the track. A synchronization error compensation model is constructed to correct the exposure timestamp of the binocular camera through the synchronization error compensation model. The expression of the synchronization error compensation model is: ;in Indicates the corrected camera exposure timestamp (aligned with GNSS time, unit: milliseconds); Indicates the timestamp of the PPS signal received by the GNSS (UTC time reference, unit: milliseconds); Cable transmission delay, i.e., the physical transmission time of the PPS signal from the GNSS antenna to the camera, is measured and calibrated using the round-trip time (RTT) (unit: milliseconds). Indicates the time it takes to transfer an image from the camera sensor to the processor (including exposure, readout, and transmission link, in milliseconds); Based on the timestamp of the corrected binocular camera exposure, align the GNSS positioning data, IMU attitude data, and mileage data with the binocular image timestamp using interpolation methods (e.g., cubic spline interpolation). The above process achieves high-precision time alignment (sub-millisecond level) of multi-sensor data through hardware trigger compensation and software interpolation algorithm, providing a reliable timing benchmark for subsequent track geometry parameter calculation and defect detection.

[0025] The methods for obtaining the predicted state vector include: The IMU attitude data and GNSS positioning data after time-space alignment are fused through extended Kalman filtering or error state Kalman filtering to output the position vector of the track detection vehicle in the world coordinate system. (unit: meters) and the velocity vector (Unit: m / s), attitude quaternion vector (represents the rotation transformation from the IMU body coordinate system to the world coordinate system, obtained by converting the original Euler angle data) and the error of the IMU sensor, including the accelerometer zero bias vector (indicates the inherent hardware error of the accelerometer, in meters per second²) and the gyroscope bias vector (represents the low-frequency random error of the gyroscope, unit: radian / second); The position vector, velocity vector, attitude quaternion vector, accelerometer bias vector and gyroscope bias vector are spliced ​​to obtain the state vector ; Construct a kinematic model. The function expression of the kinematic model is: ; in represents the vector consisting of acceleration and angular velocity measurements; Indicates the rate of change of the track detection trolley position, unit: m / s; Indicates the rate of change of the track inspection trolley speed, unit: m / s²; Represented by the attitude quaternion vector The derived rotation matrix is ​​used to convert the IMU acceleration Convert to the world coordinate system; Represents the acceleration measurement value in the IMU attitude data after time and space alignment, unit: m / s²; represents the gravitational acceleration vector ( , unit: m / s²); Indicates the attitude change rate of the track detection car, unit: radian / second; Represents the raw angular velocity measurement value in the IMU attitude data after time and space alignment, unit: radian / second; Represents quaternion multiplication; Indicates the rate of change of accelerometer bias, unit: m / s³; Indicates the rate of change of gyroscope bias, unit: radian / second²; Indicates the accelerometer bias noise, unit: m / s³; Indicates the gyroscope bias noise in radians per second².

[0026] The dynamic model is converted into a discrete-time prediction model by numerical integration (4th-order Runge-Kutta method), and the motion state vector is obtained; the expression of the discrete-time prediction model is ;in represents the predicted state vector at discrete time; Represents the state vector at the current time point; Represents the vector of acceleration and angular velocity measurements at the current moment; Indicates a point in time; represents the equivalent state increment function calculated by RK4; Represents the discretized process noise vector, which obeys the zero-mean Gaussian distribution , Represents the noise covariance matrix composed of position vector, velocity vector, attitude vector, accelerometer bias vector and gyroscope bias vector, which is used to describe the noise statistical characteristics of each component in state prediction; ; represents the discretized time step (0.005 seconds), which is strictly synchronized with the IMU sampling rate; Based on the mileage data processed by time-space alignment, the position vector in the motion state vector is corrected through a filtering algorithm (such as the Kalman filter algorithm or the mean filter algorithm) to obtain the predicted state vector of time-space alignment. .

[0027] like Figure 2As shown in the figure, the steps of generating a track point cloud map include: The binocular image is calibrated based on the distortion coefficient in the intrinsic parameter to obtain a calibrated binocular image; Based on a lightweight convolutional neural network (MobileNetV3), multi-scale feature extraction is performed on the calibrated binocular image to obtain multi-scale feature maps (with resolutions of 1 / 8, 1 / 4, and 1 / 2 of the binocular image); Perform cost calculation on the multi-scale feature maps to obtain a mixed cost volume. Using a winner-takes-all strategy, select the disparity with the smallest cost value for each pixel from the mixed cost volume as the initial disparity estimation result, and finally obtain the initial disparity map. The initial disparity map is optimized with edge constraints using a conditional random field energy function (unary + binary) and a parabolic interpolation function to obtain an optimized disparity map that retains rail contour details. The optimized disparity map is upsampled through a pre-built convolutional neural network to generate a predicted disparity map (with the same resolution as the binocular image); The predicted disparity map is triangulated based on the obtained baseline and intrinsic parameter matrix to obtain the coordinates of each pixel in the predicted disparity map in the camera coordinate system; The pixel coordinates in the camera coordinate system are converted to the coordinates in the IMU coordinate system through the rotation matrix and translation vector of the binocular camera; Based on the attitude quaternion vector in the predicted state vector, the coordinates of the pixel in the IMU coordinate system are converted to the coordinates in the world coordinate system (X, Y, Z). The coordinates of all pixels in the world coordinate system constitute a point cloud image; Filter the point cloud image to remove outliers (such as fasteners or other debris) and noise data (using statistical filtering, the neighborhood standard deviation is greater than 3σ) in the point cloud image to obtain the track point cloud image; It should be noted that the cost calculation method includes: The cosine similarity is used to calculate the cross-channel correlation of the multi-level feature maps to obtain the initial matching cost. Based on the initial matching cost, the traditional SAD algorithm (calculation window size is 5×5) is combined with the local pixel intensity difference of the feature map to obtain the local cost. The initial matching cost and the local cost are weightedly fused using an adaptive weight fusion strategy to generate a hybrid cost volume. The hybrid cost volume consists of three 3D cost matrices (height × width × disparity range) covering a disparity range of 0-256 pixels. The camera coordinate system is defined as: origin: the center of the baseline; X-axis: horizontally to the right along the imaging plane; Y-axis: vertically downward along the imaging plane; Z-axis: perpendicular to the camera imaging plane, pointing directly in front of the scene; the IMU coordinate system is defined as: origin: located at the sensor center of the inertial measurement unit (IMU); X-axis: horizontally to the right along the front of the sensor (or the direction of the device's movement); Y-axis: vertically upward along the front of the sensor; Z-axis: following the right-hand rule, perpendicular to the XY plane and forward; triangulation calculations and coordinate system transformations are processing processes well known to those skilled in the art, and the specific calculation process and coordinate conversion process are not given here.

[0028] The coordinate sequence of the inner track points and the track elevation sequence are obtained in the following ways: Based on the track point cloud map, slice the track along the track extension direction at a preset slice interval (for example, 0.01m), extract the point cloud data at the top surface of the track in each slice, and calculate the arithmetic mean of the height coordinates based on the point cloud data at the top surface of the track. The obtained arithmetic mean values ​​are arranged in spatial order to form a complete track elevation sequence; A reference coordinate system is established based on the highest point group on the top surface of the track. In the reference coordinate system, the X axis is along the extension direction of the track, the Y axis points to the inside of the track, and the Z axis is vertically upward along the XY plane. Move down the Z axis by a preset distance (14mm or 16mm from the highest point of the track top surface) to construct a feature extraction plane. Establish a sampling window at a preset slice interval along the track extension direction. Obtain the point cloud coordinates on the inner side of the feature extraction plane through the sampling window to obtain the coordinate sequence of the inner points of the track. and ; Perform angle conversion on the attitude quaternion vector in the predicted state vector to obtain the roll angle Based on the track design superelevation value (derived from the line design file) and the roll angle Construct a gauge calculation function and use it to calculate the coordinate sequence of the inner points of the track to obtain the gauge sequence ; Set the track gauge sequence Compare with the preset standard gauge (1435mm) to generate a modified gauge sequence (how much the gauge differs from the standard gauge). Combined with the position vector in the predicted state vector, the corresponding gauge table (including position, gauge, and the modified gauge required for that position) is generated. It should be noted that the function expression of the gauge calculation function is: ;in and are the coordinate sequences of the inner side of the left track and the right track respectively; Design superelevation values ​​for the track.

[0029] Construct a vertical displacement kinematic model. The function expression of the vertical displacement kinematic model is: ; in Indicates the time derivative of the vertical (Z-axis) position of the track detection trolley (i.e. vertical velocity), unit: m / s; Indicates the instantaneous vertical speed of the track inspection trolley, unit: m / s; Indicates the vertical acceleration of the track inspection trolley, unit: m / s²; Represents the vertical acceleration measurement value in the IMU attitude data after time and space alignment, unit: m / s²; Represents the gravitational acceleration vector (unit: m / s²); Indicates the pitch angle of the vehicle, which is calculated from the attitude quaternion vector in the predicted state vector. Unit: radian; Accelerometer bias vector; Indicates the rate of change of accelerometer bias, unit: m / s³; Indicates the accelerometer bias noise, unit: m / s³; The vertical displacement kinematic model is converted into a discrete vertical time displacement prediction model through numerical integration (4th-order Runge-Kutta method). The IMU attitude data after time-space alignment is used as the input of the vertical time displacement prediction model to obtain vertical displacement data. The vertical displacement data obtained by the vertical time displacement prediction model is aligned with the design elevation of the track every 1 km to generate a vertical displacement sequence. ; The track elevation sequence and vertical displacement sequence are weightedly fused (weight ratio is 0.45:0.55) to obtain the vertical height data of the track; a vertical curve is drawn based on the vertical height data of the track, where the horizontal axis of the vertical curve is the position corresponding to the vertical height data, and the vertical axis is the vertical height data; Based on the line design file, a standard smoothness curve is drawn and superimposed on the vertical curve diagram. A threshold for irregularity is first set, and then, from the drawn vertical curve diagram, curve sections where the absolute difference between the actual curve and the standard smoothness curve exceeds the preset threshold are extracted. For each extracted curve section, its position in the diagram, the corresponding track vertical height data, and the absolute difference between the actual curve and the standard curve are recorded in detail. Finally, a standardized track smoothness table is generated based on this recorded information. The gauge table and smoothness table are then integrated by location to generate a ballastless track measurement report.

[0030] This embodiment automatically calibrates camera parameters using a flexible checkerboard calibration plate combined with inherent track features, avoiding the need for manual calibration sites in traditional calibration processes and improving calibration efficiency. By introducing IMU attitude data, errors caused by movement or deformation of the calibration plate are avoided. A synchronous error compensation model and a cubic spline interpolation algorithm are constructed to correct timestamp deviations of different data, achieving sub-millisecond spatiotemporal alignment of binocular images, IMU attitude data, GNSS positioning data, and mileage data, eliminating cumulative errors caused by time asynchrony in dynamic measurements. A lightweight MobileNetV3 network is used to extract multi-scale features, combined with an adaptive weight fusion strategy to generate a mixed cost volume. Conditional random fields and parabolic interpolation functions are used to optimize edge details of the initial disparity map, significantly improving stereo matching accuracy in weak-texture scenarios and achieving sub-millimeter 3D point cloud reconstruction of the rail surface. This provides a reliable data basis for the calculation of gauge and height parameters. The gauge table and smoothness table generated on this basis can be directly connected to construction acceptance standards and support high-precision defect location.

[0031] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art will be able to modify the technical solutions described in the foregoing embodiments or to substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.

[0032] It should be noted that, in this document, the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, article, or apparatus comprising a series of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of other identical elements in the process, method, article, or apparatus comprising the element.

[0033] In the description of the present invention, it should be understood that the terms "first", "second", etc. are only used to distinguish the descriptions and cannot be understood as indicating or implying relative importance.

[0034] In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0035] In the description of the present invention, “several” means one or more, and “a large number” means two or more.

[0036] Throughout this specification, reference to terms such as "one embodiment," "some embodiments," "examples," "specific examples," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0037] The formulas in this manual are all dimensionless and calculated using numerical values. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters and thresholds in the formulas are set by technicians in this field based on actual conditions.

[0038] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A high-speed railway ballastless track construction measurement method, characterized in that: include: S1. Perform internal and external calibration on the binocular camera on the track inspection vehicle to obtain the corresponding internal and external reference data; S2, using the track detection vehicle to obtain binocular images, mileage data, IMU attitude data, and GNSS positioning data and perform spatiotemporal alignment processing; S3, constructing a predicted state vector based on the mileage data, IMU attitude data and GNSS positioning data after time-space alignment; S4. Perform stereo matching on the binocular images to generate a predicted disparity map; perform three-dimensional reconstruction on the predicted disparity map based on the intrinsic parameter data, the extrinsic parameter data, and the predicted state vector to generate a track point cloud map; S5. Extracting a coordinate sequence of the inner track points and a track elevation sequence from the track point cloud image; calculating a track gauge sequence based on the coordinate sequence of the inner track points, and generating a corresponding track gauge table based on the track gauge sequence; S6. Perform vertical calculation on the IMU attitude data after time-space alignment processing to obtain the vertical displacement sequence of the track; perform weighted fusion of the track height sequence and the vertical displacement sequence to obtain the vertical height data of the track, and generate the corresponding smoothness table based on the vertical height data of the track; integrate the track gauge table and the smoothness table to generate a measurement report.

2. A high-speed railway ballastless track construction measurement method according to claim 1, characterized in that: The method for obtaining the internal reference data includes: The track detection vehicle collects a calibration image containing the flexible checkerboard calibration plate and the track, and records the IMU calibration data during the collection process; obtains the coordinates of the checkerboard corner points in the flexible checkerboard calibration plate in the calibration image and integrates the IMU calibration data to obtain the IMU rotation matrix; A projection error function is constructed based on the coordinates of the obtained checkerboard corner points and the IMU rotation matrix. The historically calibrated intrinsic parameter matrix and distortion coefficients are updated by minimizing the projection error function to obtain the intrinsic parameter matrix and distortion coefficients. The intrinsic parameter matrix and distortion coefficients together constitute the intrinsic parameter data.

3. A high-speed railway ballastless track construction measurement method according to claim 2, characterized in that: The method for obtaining the external reference data includes: Acquire feature point pairs and rail gauges from the calibration image; construct an extrinsic parameter optimization function based on the feature point pairs and rail gauges; iteratively optimize the extrinsic parameter optimization function using the Levenberg-Marquardt method, and obtain the corresponding baseline, rotation matrix, and translation vector when the extrinsic parameter optimization function iteratively converges to a minimum value; the baseline, rotation matrix, and translation vector together constitute the extrinsic parameter data; A binocular camera with calibrated extrinsic parameters is used to acquire N sets of test images. Test matching point pairs in the test images are obtained using a feature matching algorithm. The vertical distances from the test matching point pairs to the epipolar lines are calculated, recorded as epipolar residuals. The average of all epipolar residuals is calculated. When the average is greater than or equal to the preset pixel threshold, the extrinsic parameters need to be recalibrated.

4. A high-speed railway ballastless track construction measurement method according to claim 3, characterized in that: The method of performing the spatiotemporal alignment process includes: measuring the cable transmission delay and the time taken for the image to be transmitted from the camera sensor to the processor; correcting the exposure timestamp of the binocular camera based on the cable transmission delay and the time taken for the image to be transmitted from the camera sensor to the processor to obtain a corrected exposure timestamp of the binocular camera; Based on the corrected binocular camera exposure timestamp, the GNSS positioning data, IMU attitude data, and mileage data are aligned with the binocular image timestamp through interpolation.

5. A high-speed railway ballastless track construction measurement method according to claim 4, characterized in that: The method for obtaining the predicted state vector includes: The IMU attitude data and GNSS positioning data after time-space alignment are fused through a filtering algorithm to obtain the position vector, velocity vector, attitude quaternion vector and error vector of the track detection vehicle in the world coordinate system; the position vector, velocity vector, attitude quaternion vector and error vector of the IMU sensor are spliced ​​to obtain the state vector; A kinematic model is constructed and discretized to obtain a discrete-time prediction model. The state vector is used as the input of the discrete-time prediction model to obtain a motion state vector. Based on the mileage data after spatiotemporal alignment, the position vector in the motion state vector is corrected through a filtering algorithm to obtain a predicted state vector.

6. A high-speed railway ballastless track construction measurement method according to claim 5, characterized in that: The method for performing stereo matching includes: The binocular image is calibrated based on the distortion coefficient in the intrinsic reference data to obtain a calibrated binocular image; multi-scale feature extraction is performed on the calibrated binocular image through a lightweight convolutional neural network to obtain a multi-scale feature map; The cost of multi-scale feature maps is calculated to obtain a mixed cost volume; pixels in the mixed cost volume are selected based on a winner-takes-all strategy to obtain an initial disparity map; the initial disparity map is edge-constrained optimized using a conditional random field energy function and a parabolic interpolation function to obtain an optimized disparity map; the optimized disparity map is upsampled using a pre-built convolutional neural network to generate a predicted disparity map.

7. A high-speed railway ballastless track construction measurement method according to claim 6, characterized in that: The method for obtaining the track point cloud map includes: The predicted disparity map is triangulated based on the obtained intrinsic parameter matrix and baseline to obtain the coordinates of the pixels in the predicted disparity map in the camera coordinate system; the coordinates of the pixels in the camera coordinate system are converted to the coordinates in the IMU coordinate system through the obtained rotation matrix and translation vector; the coordinates of the pixels in the IMU coordinate system are converted to the coordinates in the world coordinate system based on the attitude quaternion in the predicted state vector, and the coordinates of all pixels in the world coordinate system constitute a point cloud map; the point cloud map is filtered to obtain a track point cloud map.

8. A high-speed railway ballastless track construction measurement method according to claim 7, characterized in that: The coordinate sequence of the inner track points and the track elevation sequence are obtained by: Based on the track point cloud map, slice the track along the track extension direction at a preset slice interval, extract the point cloud data at the top surface of the track in each slice, and calculate the arithmetic mean of the height coordinates based on the point cloud data at the top surface of the track. Arrange the obtained arithmetic mean values ​​in spatial order to obtain the track elevation sequence; Establish a reference coordinate system at the highest point of the track top surface in the track point cloud image, where the X axis of the reference coordinate system is along the extension direction of the track, the Y axis points to the inside of the track, and the Z axis is vertically upward along the XY plane; A feature extraction plane is constructed by moving down a preset distance along the Z axis, and a sampling window is established along the extension direction of the track at a preset slice interval. The point cloud coordinates of the inner side of the track in the feature extraction plane are obtained through the sampling window to obtain the coordinate sequence of the points on the inner side of the track.

9. A high-speed railway ballastless track construction measurement method according to claim 8, characterized in that: The method of generating a corresponding track gauge table based on the track gauge sequence includes: The attitude quaternion vector in the predicted state vector is converted into an angle to obtain the roll angle; a gauge calculation function is constructed based on the track design superelevation value and the roll angle, and the gauge calculation function is used to calculate the coordinate sequence of the inner point of the track to obtain the gauge sequence; the gauge sequence is compared with the preset standard gauge to generate a corrected gauge sequence; and a corresponding gauge table is generated based on the gauge sequence, the corrected gauge sequence, and the position vector in the predicted state vector.

10. A high-speed railway ballastless track construction measurement method according to claim 9, characterized in that: The vertical displacement sequence of the track is obtained by: A vertical displacement kinematic model is constructed and discretized to obtain a vertical time displacement prediction model. The IMU attitude data after spatiotemporal alignment is used as the input of the vertical time displacement prediction model to obtain vertical displacement data. The vertical displacement data is aligned with the design elevation of the track at preset intervals to generate a vertical displacement sequence.

Images