Road boundary line detection method based on vehicle-mounted millimeter wave radar

Abstract

The invention relates to a road boundary line detection method based on a vehicle-mounted millimeter wave radar. The method comprises the following steps: acquiring radar point cloud information of a current road scene and vehicle body information; according to the obtained target absolute speed, motion point cloud is filtered out through dynamic and static separation operation; road scene recognition is carried out by applying different screening points and clustering strategies in combination with the motion state of the current vehicle; determining an ROI area, and carrying out the first screening of the static point clouds; performing clustering processing on the screened static point clouds by adopting a density-based spatial clustering method to obtain a point cloud cluster, and performing secondary cluster screening to obtain a road boundary cluster; performing curve fitting on the clustering points meeting the requirements of the second cluster screening by adopting a least square method, performing third cluster screening at the same time, and removing part of noise clusters and non-road boundary clusters; and performing road boundary curve filtering processing on the currently obtained road boundary line by using a Kalman filtering algorithm to obtain an optimal road curve estimation result of the current frame.

Description

Technical Field

[0001] The present invention relates to the field of intelligent driving, and in particular to the field of millimeter-wave radar detection, and specifically refers to a road boundary line detection method, system, device, processor and computer-readable storage medium thereof based on a vehicle-mounted millimeter-wave radar. Background Art

[0002] With the rapid development of the automotive industry and the continuous advancement of intelligent technology, automotive millimeter-wave radar technology has gradually matured and been applied to the field of automotive safety. Millimeter-wave radar has the characteristics of high precision, high resolution, and high reliability, and can achieve real-time monitoring and high-precision perception of target objects in complex environments.

[0003] Among them, the detection of road boundaries is crucial to ensure driving safety. In severe weather conditions (such as rain, snow, fog, etc.), the camera's line of sight is easily affected, while the on-board millimeter-wave radar can penetrate the influence of rain, snow, fog and other weather conditions to achieve accurate detection of road boundaries. This helps drivers to more accurately judge road boundaries in severe weather conditions, avoid vehicle deviation from the road or collision accidents, and thus improve driving safety.

[0004] CN202311552032.8 discloses a road boundary detection method, device, electronic device and storage medium. Compared with the application that uses Hough transform to only detect straight lines, this technical solution uses the least squares curve fitting method when fitting road boundary lines, which can detect curved lane boundary lines in non-straight form.

[0005] CN202110528933.8 discloses a road boundary detection method. Compared with this application, this technical solution can obtain the speed state of the target point cloud, accurately distinguish the dynamic and static states of the target, greatly reducing the number of point cloud targets that need to be calculated. In addition, static point cloud clustering is added to improve the detection accuracy of boundary curves.

[0006] CN202210072589.0 discloses a road boundary detection method and system based on a vehicle-mounted millimeter-wave radar. Compared with the application, the present technical solution takes into account the motion state of the vehicle within the frame period. In particular, when the vehicle turns, the road boundary curve equation in the vehicle coordinate system will change. Therefore, a rotation matrix is ​​introduced to correct the road curve equation after the transformation of the vehicle coordinate system and then smoothing filtering is performed.

[0007] CN202011463411.6 discloses a road boundary detection method based on a vehicle-mounted millimeter-wave radar. Compared with this application, the curve fitted by this technical solution is a quadratic curve, which requires a smaller amount of calculation than the cubic curve fitting mentioned in this application. For actual roads, most of them are straight lines or curve boundaries with small curvature, so there is no need to consider the road fitting of the cubic curve equation.

[0008] Since the above scheme does not consider the road curve fitting results of the previous and next frames, there may be a situation where the results of the two frames are very different. Therefore, this technical solution introduces the Kalman filter combined with the matrix transformation method. By smoothing the fitting results of the previous and next frames, a stable lane line output for each frame can be obtained. This technical solution also improves the density-based spatial clustering method, applying different weights to different regions, different road scenes and different clustering directions to reduce the impact of noise points and non-road boundary static points on the clustering results. Summary of the invention

[0009] The purpose of the present invention is to overcome the shortcomings of the above-mentioned prior art and provide a road boundary line detection method, system, device, processor and computer-readable storage medium based on vehicle-mounted millimeter-wave radar.

[0010] In order to achieve the above-mentioned object, the road boundary line detection method, system, device, processor and computer-readable storage medium based on vehicle-mounted millimeter wave radar of the present invention are as follows: The main feature of the road boundary line detection method based on vehicle-mounted millimeter wave radar is that the method comprises the following steps: (1) The vehicle-mounted millimeter-wave radar is rigidly connected to the vehicle and installed to obtain radar point cloud information of the current road scene and vehicle body information; (2) Based on the acquired absolute speed of the target, the moving point cloud is filtered out through the static-dynamic separation operation to determine the absolute speed of the point cloud; (3) Combined with the current vehicle motion state, road scene recognition is performed by applying different screening points and clustering strategies; (4) Determine the ROI area to perform the first screening of the static point cloud; (5) A density-based spatial clustering method is used to cluster the stationary point cloud after ROI screening to obtain point cloud clusters, and a second cluster screening is performed to obtain road boundary clusters; (6) The least squares method is used to perform curve fitting on the cluster points that meet the requirements of the second cluster screening, and the third cluster screening is performed at the same time to eliminate some noise clusters and non-road boundary clusters; (7) The Kalman filter algorithm is used to perform road boundary curve filtering on the currently acquired road boundary line to obtain the optimal road curve estimation result of the current frame and output the corresponding boundary curve.

[0011] Preferably, the step (1) is specifically as follows: The Cartesian coordinate system is specified with the center of mass of the vehicle as the origin of the coordinate system. The front of the vehicle is the positive x direction of the coordinate system, and the horizontal direction to the left of the vehicle is the positive y direction of the coordinate system. When the vehicle-mounted millimeter-wave radar detects the target point cloud, the target position is converted to the vehicle body coordinate system to determine the coordinates through the relative point cloud position, vehicle speed, yaw angular velocity and installation position information. The conversion method from point cloud speed to absolute vehicle speed is as follows: ; in, represents the absolute speed of the target, represents the radial velocity measured by the radar, represents the vehicle's speed, Indicates the point cloud orientation angle, represents the yaw angular velocity of the vehicle, Indicates the longitudinal installation position of the radar relative to the vehicle. Indicates the lateral installation position of the radar relative to the vehicle.

[0012] Preferably, the step (2) is specifically as follows: Set the velocity positive threshold , when the absolute value of the target absolute speed Less than threshold , the point cloud is considered to be stationary; otherwise, the point cloud is in motion, and the motion point cloud with speed is filtered out through the static-dynamic separation operation.

[0013] Preferably, the step (3) specifically comprises the following steps: (3.1) Count the heights of the absolutely stationary point clouds in a fixed area and determine the ratio Hr of the total number of point clouds that meet the requirements. If Hr is greater than the empirical parameter Thr, the current vehicle road scene is judged to be a tunnel scene. Otherwise, proceed to the next step. (3.2) If the current vehicle speed is greater than the first preset speed or the vehicle speed and yaw angular velocity are greater than the first preset angular velocity, the current vehicle road scene is determined to be a high-speed scene, otherwise, proceed to the next step; (3.3) If the current vehicle speed is greater than the second preset speed or the vehicle speed and yaw angular velocity are greater than the second preset angular velocity, the current vehicle road scene is determined to be an urban scene, otherwise, proceed to the next step; (3.4) If the current vehicle speed is greater than the third preset speed, the current vehicle road scene is determined to be a low-speed scene, otherwise, proceed to the next step; (3.5) None of the above conditions are met, and the current vehicle road scene is stationary.

[0014] Preferably, the step (5) specifically comprises the following steps: (5.1) First determine the threshold radius And the minimum number of points MinPts, select an unprocessed data point p from the stationary point cloud, and calculate the data point p in The number of data points in the neighborhood; (5.2) If the data point p is If the number of data points in the neighborhood is greater than or equal to the minimum number of points MinPts, then the data point p is marked as a core point; if the data point p is If the number of data points in the neighborhood is less than MinPts, but the point falls in the neighborhood of other core points, then the data point p is marked as a boundary point; if the data point p is neither a core point nor a boundary point, then it is marked as a noise point; (5.3) Through each core point, find all the data points that can be reached by its density to form a cluster. Repeat this process until all core points have been processed and all data points p have been marked as a cluster or a noise point. (5.4) After clustering is completed, the number of point clouds contained in each cluster, the longitudinal length and the lateral length of the cluster are traversed to determine whether it is a road boundary cluster.

[0015] Preferably, the step (6) is specifically as follows: The least squares method is used to perform curve fitting on the cluster points on the left and right sides of the vehicle that are most likely to be road boundary clusters in the following way: ; in, represents the quadratic coefficient of the fitting curve, represents the linear coefficient of the fitting curve, represents the constant term of the fitted curve, , is the vehicle coordinate.

[0016] Preferably, the step (6) further comprises: The curvature of the fitting result is calculated as follows, and the yaw rate and curvature are used to calculate the curvature. Eliminate road boundaries that do not conform to reality: ; in, represents the first-order derivative of the curve at a certain position, Represents the second derivative of a curve at a certain position.

[0017] Preferably, the step (7) comprises: According to the rotation matrix , calculate the prior road curve fitting result of the previous frame after the vehicle turns at the next moment, specifically: ; in, Indicates the target horizontal coordinate after the vehicle turns. Indicates the target ordinate after the vehicle turns. Indicates the target horizontal coordinate of the previous frame before the vehicle turns. Indicates the target vertical coordinate of the previous frame before the vehicle turns. Indicates that the vehicle is in the radar frame period The angle of internal rotation, ; According to the above formula, we can get the new fitting curve coefficients: ; ; ; Using the current frame and the previous frame Six coefficients, based on the Kalman filter algorithm, to obtain stable coefficient output in the time domain; Get the road boundary quadratic curve: ; in, is the quadratic coefficient of the road fitting curve after smoothing by Kalman filtering, is the linear coefficient of the smoothed road fitting curve, is the constant term of the smoothed road fitting curve.

[0018] The main feature of the road boundary line detection system based on vehicle-mounted millimeter wave radar for implementing the above-mentioned method is that the system includes: Radar point cloud preprocessing module, used to obtain radar point cloud information and vehicle body information of the current road scene; A road scene recognition module is connected to the radar point cloud preprocessing module and is used to perform road scene recognition processing by applying different screening points and clustering strategies according to the current vehicle motion state; A point cloud clustering module, connected to the road scene recognition module, for clustering the screened point cloud using a spatial clustering method; A clustering screening module, connected to the point cloud clustering module, is used to screen out road boundary clusters that meet the requirements from the acquired point cloud; A curve fitting module, connected to the cluster screening module, for performing curve fitting processing on cluster points that meet the screening requirements; and The road boundary line filtering module is connected to the curve fitting module and is used to perform road boundary curve filtering processing on the currently acquired road boundary line using the Kalman filtering algorithm.

[0019] The road boundary line detection device based on vehicle-mounted millimeter wave radar has the following main features: the device comprises: a processor configured to execute computer executable instructions; The memory stores one or more computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the road boundary line detection method based on the vehicle-mounted millimeter wave radar are implemented.

[0020] The road boundary line detection processor based on the vehicle-mounted millimeter-wave radar has the main feature that the processor is configured to execute computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the road boundary line detection method based on the vehicle-mounted millimeter-wave radar are implemented.

[0021] The main feature of the computer-readable storage medium is that a computer program is stored thereon, and the computer program can be executed by a processor to implement the steps of the above-mentioned road boundary line detection method based on vehicle-mounted millimeter-wave radar.

[0022] The road boundary line detection method, system, device, processor and computer-readable storage medium based on the vehicle-mounted millimeter-wave radar of the present invention are adopted to find the absolute stationary point relative to the ground after speed conversion calculation through the point cloud information detected by the vehicle-mounted millimeter-wave radar, and estimate the current road scene in combination with the vehicle speed and the vehicle yaw angular velocity. Then, the point cloud that meets the conditions is screened out through ROI to complete the clustering of stationary target clusters and road curve fitting. The influence of other stationary obstacles other than road boundaries can be effectively eliminated, and compared with the traditional road boundary line detection method, the calculation time of curve fitting is greatly reduced by clustering some stationary point clouds. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The figure is a flow chart of the road boundary line detection method based on vehicle-mounted millimeter-wave radar of the present invention.

[0024] Figure 2 It is a structural schematic diagram of the road boundary line detection system based on the vehicle-mounted millimeter-wave radar of the present invention.

[0025] Figure 3 It is an implementation flow chart of the present invention in a specific embodiment.

[0026] Figure 4 FIG. 4 is a schematic diagram of curve fitting in a specific embodiment of the present invention.

[0027] Figure 5 FIG. 4 is a schematic diagram of performing scene recognition processing in a specific embodiment of the present invention. DETAILED DESCRIPTION

[0028] In order to more clearly describe the technical content of the present invention, further description is given below in conjunction with specific embodiments.

[0029] Before describing in detail embodiments according to the present invention, it should be noted that, hereinafter, the terms "comprises", "includes" or any other variations are intended to cover non-exclusive inclusion, whereby a process, method, article or apparatus comprising a series of elements includes not only these elements, but also other elements not explicitly listed or inherent to such process, method, article or apparatus.

[0030] See also Figure 1 As shown, the road boundary line detection method based on vehicle-mounted millimeter wave radar, wherein the method comprises the following steps: (1) The vehicle-mounted millimeter-wave radar is rigidly connected to the vehicle and installed to obtain radar point cloud information of the current road scene and vehicle body information; (2) Based on the acquired absolute speed of the target, the moving point cloud is filtered out through the static-dynamic separation operation to determine the absolute speed of the point cloud; (3) Combined with the current vehicle motion state, road scene recognition is performed by applying different screening points and clustering strategies; (4) Determine the ROI area to perform the first screening of the static point cloud; (5) A density-based spatial clustering method is used to cluster the stationary point cloud after ROI screening to obtain point cloud clusters, and a second cluster screening is performed to obtain road boundary clusters; (6) The least squares method is used to perform curve fitting on the cluster points that meet the requirements of the second cluster screening, and the third cluster screening is performed at the same time to eliminate some noise clusters and non-road boundary clusters; (7) The Kalman filter algorithm is used to perform road boundary curve filtering on the currently acquired road boundary line to obtain the optimal road curve estimation result of the current frame and output the corresponding boundary curve.

[0031] As a preferred embodiment of the present invention, the step (1) is specifically as follows: The Cartesian coordinate system is specified with the center of mass of the vehicle as the origin of the coordinate system. The front of the vehicle is the positive x direction of the coordinate system, and the horizontal direction to the left of the vehicle is the positive y direction of the coordinate system. When the vehicle-mounted millimeter-wave radar detects the target point cloud, the target position is converted to the vehicle body coordinate system to determine the coordinates through the relative point cloud position, vehicle speed, yaw angular velocity and installation position information. The conversion method from point cloud speed to absolute vehicle speed is as follows: ; in, represents the absolute speed of the target, represents the radial velocity measured by the radar, represents the vehicle's speed, Indicates the point cloud orientation angle, represents the yaw angular velocity of the vehicle, Indicates the longitudinal installation position of the radar relative to the vehicle. Indicates the lateral installation position of the radar relative to the vehicle.

[0032] As a preferred embodiment of the present invention, the step (2) is specifically as follows: Set the velocity positive threshold , when the absolute value of the target absolute speed Less than threshold , the point cloud is considered to be stationary; otherwise, the point cloud is in motion, and the motion point cloud with speed is filtered out through the static-dynamic separation operation.

[0033] As a preferred embodiment of the present invention, the step (3) specifically includes the following steps: (3.1) Count the heights of the absolutely stationary point clouds in a fixed area and determine the ratio Hr of the total number of point clouds that meet the requirements. If Hr is greater than the empirical parameter Thr, the current vehicle road scene is judged to be a tunnel scene. Otherwise, proceed to the next step. (3.2) If the current vehicle speed is greater than the first preset speed or the vehicle speed and yaw angular velocity are greater than the first preset angular velocity, the current vehicle road scene is determined to be a high-speed scene, otherwise, proceed to the next step; (3.3) If the current vehicle speed is greater than the second preset speed or the vehicle speed and yaw angular velocity are greater than the second preset angular velocity, the current vehicle road scene is determined to be an urban scene, otherwise, proceed to the next step; (3.4) If the current vehicle speed is greater than the third preset speed, the current vehicle road scene is determined to be a low-speed scene, otherwise, proceed to the next step; (3.5) None of the above conditions are met, and the current vehicle road scene is stationary.

[0034] As a preferred embodiment of the present invention, the step (5) specifically includes the following steps: (5.1) First determine the threshold radius And the minimum number of points MinPts, select an unprocessed data point p from the stationary point cloud, and calculate the data point p in The number of data points in the neighborhood; (5.2) If the data point p is If the number of data points in the neighborhood is greater than or equal to the minimum number of points MinPts, then the data point p is marked as a core point; if the data point p is If the number of data points in the neighborhood is less than MinPts, but the point falls in the neighborhood of other core points, then the data point p is marked as a boundary point; if the data point p is neither a core point nor a boundary point, then it is marked as a noise point; (5.3) Through each core point, find all the data points that can be reached by its density to form a cluster. Repeat this process until all core points have been processed and all data points p have been marked as a cluster or a noise point. (5.4) After clustering is completed, the number of point clouds contained in each cluster, the longitudinal length and the lateral length of the cluster are traversed to determine whether it is a road boundary cluster.

[0035] As a preferred embodiment of the present invention, the step (6) is specifically as follows: The least squares method is used to perform curve fitting on the cluster points on the left and right sides of the vehicle that are most likely to be road boundary clusters in the following way: ; in, represents the quadratic coefficient of the fitting curve, represents the linear coefficient of the fitting curve, represents the constant term of the fitted curve, , is the vehicle coordinate.

[0036] As a preferred embodiment of the present invention, the step (6) further comprises: The curvature of the fitting result is calculated as follows, and the yaw rate and curvature are used to calculate the curvature. Eliminate road boundaries that do not conform to reality: ; in, represents the first-order derivative of the curve at a certain position, Represents the second derivative of a curve at a certain position.

[0037] As a preferred embodiment of the present invention, the step (7) comprises: According to the rotation matrix , calculate the prior road curve fitting result of the previous frame after the vehicle turns at the next moment, specifically: ; in, Indicates the target horizontal coordinate after the vehicle turns. Indicates the target ordinate after the vehicle turns. Indicates the target horizontal coordinate of the previous frame before the vehicle turns. Indicates the target vertical coordinate of the previous frame before the vehicle turns. Indicates that the vehicle is in the radar frame period The angle of internal rotation, ; Get the new fitting curve coefficients according to the above formula: ; ; ; Using the current frame and the previous frame Six coefficients, based on the Kalman filter algorithm, to obtain stable coefficient output in the time domain; Get the road boundary quadratic curve: ; in, is the quadratic coefficient of the road fitting curve after smoothing by Kalman filtering, is the linear coefficient of the smoothed road fitting curve, is the constant term of the smoothed road fitting curve.

[0038] See also Figure 2 As shown, the road boundary line detection system based on vehicle-mounted millimeter-wave radar for implementing the above-mentioned method, wherein the system includes: Radar point cloud preprocessing module, used to obtain radar point cloud information and vehicle body information of the current road scene; A road scene recognition module is connected to the radar point cloud preprocessing module and is used to perform road scene recognition processing by applying different screening points and clustering strategies according to the current vehicle motion state; A point cloud clustering module, connected to the road scene recognition module, for clustering the screened point cloud using a spatial clustering method; A clustering screening module, connected to the point cloud clustering module, is used to screen out road boundary clusters that meet the requirements from the acquired point cloud; A curve fitting module, connected to the cluster screening module, for performing curve fitting processing on cluster points that meet the screening requirements; and The road boundary line filtering module is connected to the curve fitting module and is used to perform road boundary curve filtering processing on the currently acquired road boundary line using the Kalman filtering algorithm.

[0039] The technical solution will be further described in detail below: like Figure 1 As shown in the flowchart, the present invention proposes a road boundary line detection method based on vehicle-mounted millimeter wave radar. The method mainly involves six modules, including radar point cloud preprocessing module, road scene recognition module, point cloud clustering module, clustering cluster screening module, curve fitting module, road boundary line filtering module, such as Figure 2 shown.

[0040] The vehicle-mounted millimeter-wave radar is rigidly connected to the vehicle and installed in the front (including the left front, right front, and front). This technical solution has no specific requirements for the vehicle model.

[0041] First, it is necessary to specify a Cartesian coordinate system with the center of mass of the vehicle as the origin. The front of the vehicle is the positive x direction of the coordinate system, and the horizontal direction to the left of the vehicle is the positive y direction of the coordinate system. When the vehicle-mounted millimeter-wave radar detects the target point cloud, the target position can be converted to a certain coordinate in the vehicle body coordinate system through the relative point cloud position, vehicle speed, yaw rate and installation position information, that is, (x, y) is uniquely determined. The conversion formula from point cloud speed to absolute vehicle speed is: (1); In formula (1), represents the absolute speed of the target, represents the radial velocity measured by the radar, represents the vehicle's speed, Indicates the point cloud orientation angle, represents the yaw angular velocity of the vehicle, Indicates the longitudinal installation position of the radar relative to the vehicle. Indicates the lateral installation position of the radar relative to the vehicle.

[0042] According to the calculated target absolute speed , we can filter out some moving point clouds with speed through the static-dynamic separation operation. The specific operation is: set a suitable positive speed threshold , when the absolute value of the target absolute speed Less than threshold , the point cloud is considered to be stationary; otherwise, the point cloud is in motion.

[0043] After obtaining the absolute speed of the point cloud, combined with the vehicle's motion state (such as vehicle speed, yaw angular velocity, etc.), the road scene recognition algorithm in step three can be used to determine whether the current vehicle driving scene is a tunnel, highway, city, low speed or vehicle stationary state. Different screening and clustering strategies can be applied in different road scenes. For example, in high-speed scenes, the screening point threshold and clustering threshold can be appropriately relaxed; in low-speed and stationary scenes, the screening point threshold and clustering threshold can be reduced to improve the accuracy of road boundary fitting. Road scene recognition methods such as Figure 3 As shown, it includes the following main contents: The height of the absolutely stationary point cloud in a fixed area is counted. Usually, the longitudinal range of the area is 15m~100m, and the lateral range is -6m~6m. Then calculate the ratio Hr of the number of points with an absolute height greater than 5m and less than 7m to the total number of points with an absolute height less than 20m. If Hr is greater than Thr, where Thr is an empirical parameter, usually 0.21, it can be judged that the current vehicle road scene is a tunnel scene. Otherwise, further judgment is made based on the vehicle speed and vehicle yaw angle: if the vehicle speed is greater than 18m / s or the vehicle speed is greater than 10m / s and the yaw rate is greater than 0.1rad / s, it is judged as a high-speed scene; if the vehicle speed is greater than 10m / s or the vehicle speed is 7m / s and the yaw rate is greater than 0.1rad / s, it is judged as an urban scene; if the vehicle speed is greater than 1m / s, it is judged as a low-speed scene; if none of the above conditions are met, it is judged that the vehicle is currently stationary.

[0044] Step 4: Determine the ROI region (Region of Interest) to filter the static point cloud. The number of points detected by the vehicle-mounted millimeter-wave radar per frame period is about several hundred. Dividing the entire space that needs to detect the road boundary curve in the vehicle Cartesian coordinate system, namely the ROI, can greatly reduce the number of point clouds entering the subsequent boundary line detection, thereby reducing the pressure on hardware calculations and improving the real-time performance of the algorithm.

[0045] In step 5, this technical solution uses a density-based spatial clustering method (DBSCAN) to cluster the static point cloud after ROI screening. Its core idea is based on the density of data points. If the density of data points in a region exceeds a certain threshold, these points are divided into a cluster. The specific implementation process is: first determine the threshold radius and the minimum number of points MinPts, then select an unprocessed data point p from the stationary point cloud and calculate p The number of data points in the neighborhood. If the number of data points in the neighborhood is greater than or equal to MinPts, point p is marked as a core point; if the number of data points in the neighborhood is greater than or equal to MinPts, point p is marked as a core point. If the number of data points in the neighborhood is less than MinPts, but the point falls in the neighborhood of other core points, then point p is marked as a boundary point; if point p is neither a core point nor a boundary point, then it is marked as a noise point. Then, through each core point, find all its density-reachable data points to form a cluster. Density reachability is a recursive concept. If point q is within the core point p, then it is marked as a noise point. neighborhood, then q is said to be directly density-reachable from p; if there exists a series of points p1, p2, …, p n , where p i+1 From p i Direct density can be reached, then it is called p n The density can be reached from p1. This process is repeated until all core points are processed and all data points are marked as a cluster or noise point.

[0046] Unlike traditional clustering methods such as partitioning and hierarchical clustering, DBSCAN divides areas with sufficiently high density into clusters and can discover clusters of arbitrary shapes. The algorithm divides data points into three categories: core points, boundary points, and noise points. A core point refers to a point that contains more than the MinPts number of other points in its neighborhood; a boundary point refers to a point that falls within the neighborhood of a core point but does not meet the MinPts number requirement; a noise point is neither a core point nor a boundary point. DBSCAN grows clusters by continuously expanding the neighborhood of core points, which can effectively handle high-dimensional data and outliers. In the clustering process, this technical solution applies different weights to different areas, different road scenes, and different clustering directions to reduce the impact of noise points and non-road boundary stationary points on the clustering results.

[0047] After clustering is completed, the number of point clouds contained in each cluster, the longitudinal length and the lateral length of the cluster are traversed to determine whether it may be a road boundary cluster.

[0048] In step 6, the present invention uses the least square method to perform curve fitting on the cluster points on the left and right sides of the vehicle obtained in step 5 that are most likely to be road boundary clusters. Figure 4 As shown, the quadratic curve formula obtained by fitting is as follows: (2); In formula (2), a represents the quadratic term coefficient of the fitting curve, b represents the linear term coefficient of the fitting curve, and c represents the constant term of the fitting curve.

[0049] In step seven, since the influence of some noise clusters and non-road boundary clusters has been eliminated in step five, and the curvature of the curve in actual situations is related to different road conditions, the curvature is calculated for the fitting results, and the road boundary lines that do not conform to reality are eliminated based on the yaw angular velocity and curvature.

[0050] The curvature calculation formula is: (3); In formula (3), represents the first-order derivative of the curve at a certain position, Represents the second derivative of a curve at a certain position.

[0051] In step eight, the present invention uses the Kalman filter algorithm to filter the left and right quadratic curve coefficients of the current frame and the left and right quadratic curve coefficients of the previous frame to obtain the optimal road curve estimation result of the current frame. In the scene of a curve or a small angle turn of the vehicle, the fitting curve equation will cause errors in the vehicle coordinate system. The curve fitting result of the previous frame at the current frame moment is corrected according to the frame period of the radar and the yaw angular velocity of the vehicle, so the fitting method is also applicable to the scene of a curve or a small angle turn of the vehicle.

[0052] According to the rotation matrix: , the priori road curve fitting result of the previous frame after the vehicle turns at the next moment can be calculated. The specific formula is: (4); In formula (4), Indicates the target horizontal coordinate after the vehicle turns. Indicates the target ordinate after the vehicle turns. Indicates the target horizontal coordinate of the previous frame before the vehicle turns. Indicates the target vertical coordinate of the previous frame before the vehicle turns. Indicates that the vehicle is in the radar frame period The angle of internal rotation, According to formula (4), we can calculate: (5); (6); (7); In this step, the current frame and the previous frame are used to obtain Six coefficients can be used to obtain relatively stable coefficient output in the time domain according to the filtering algorithm.

[0053] Take the road boundary line on one side as an example to illustrate the process of Kalman filter processing. Assume that the road boundary curves at time t and time t-1 are: (8); In the formula, is the corrected fitting coefficient of the vehicle motion at time t-1, is the fitting coefficient at time t. Assume that the state vector , state transfer matrix , the state estimation error covariance matrix is , the noise is , the process noise covariance matrix is , then the prediction equation is: (9); (10); The observation equation is: , is the measurement vector at time t, is the observation matrix, is the residual.

[0054] Kalman Gain : (11); In the formula, is the measurement noise covariance matrix.

[0055] Calculate the optimal estimate of the posterior state according to the update equation and the posterior state estimation error covariance matrix : (12); (13); Step 9: The obtained road boundary quadratic curve is as follows: (14); exist Figure 5 In the embodiment shown, the obtained vehicle speed carSpeed ​​is 24.94m / s, the vehicle yaw rate yawRate is -0.01rad / s, and the calculated Hr by the road scene recognition module is 0.0154, which is less than the threshold value 0.21. Figure 3 In this embodiment, in combination with the high-speed scene determined by scene recognition, the longitudinal distance of the ROI is set to 0m~150m, the lateral distance is ±15m, and the height is above 0m and below 3.5m. The number of point clouds of the boundary point cloud clusters obtained by clustering is set to be greater than , the length of the cluster is greater than 5m, and the width of the cluster is less than 0.8m. , frame period , so the rotation angle ,according to Figure 5 The actual point cloud of the sound insulation walls on both sides of the elevated road. The road boundary curve on the left side of the previous frame is: According to equations (5), (6) and (7), the fitting coefficient after vehicle motion correction within the frame period is calculated as: ; In this step, the current frame and the previous frame are used to obtain Six coefficients can be used to obtain relatively stable coefficient output in the time domain according to the filtering algorithm.

[0056] It has been calculated above that the road boundary curve on the left side of the previous frame after vehicle motion correction is: ; Assume that the measurement noise covariance matrix According to formula (11) (12), the optimal estimate of the posterior state can be calculated , that is, the smoothed road boundary curve is: ; Similarly, by performing the same smoothing filter on the road boundary curve fitted on the right, a stable lane line output can be obtained for each frame.

[0057] The road boundary line detection device based on vehicle-mounted millimeter wave radar, wherein the device comprises: a processor configured to execute computer executable instructions; The memory stores one or more computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the road boundary line detection method based on the vehicle-mounted millimeter wave radar are implemented.

[0058] The road boundary line detection processor based on the vehicle-mounted millimeter-wave radar, wherein the processor is configured to execute computer-executable instructions, and when the computer-executable instructions are executed by the processor, the steps of the road boundary line detection method based on the vehicle-mounted millimeter-wave radar are implemented.

[0059] The computer-readable storage medium stores a computer program thereon, and the computer program can be executed by a processor to implement the steps of the above-mentioned road boundary line detection method based on vehicle-mounted millimeter-wave radar.

[0060] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.

[0061] It should be understood that each part of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution device.

[0062] A person of ordinary skill in the art may understand that all or part of the steps of the method for implementing the above-mentioned embodiment may be completed by instructing the relevant hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one of the steps of the method embodiment or a combination thereof.

[0063] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.

[0064] In the description of this specification, the description with reference to the terms "an embodiment", "some embodiments", "example", "specific example", or "embodiment" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0065] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

[0066] The road boundary line detection method, system, device, processor and computer-readable storage medium based on the vehicle-mounted millimeter-wave radar of the present invention are adopted to find the absolute stationary point relative to the ground after speed conversion calculation through the point cloud information detected by the vehicle-mounted millimeter-wave radar, and estimate the current road scene in combination with the vehicle speed and the vehicle yaw angular velocity. Then, the point cloud that meets the conditions is screened out through ROI to complete the clustering of stationary target clusters and road curve fitting. The influence of other stationary obstacles other than road boundaries can be effectively eliminated, and compared with the traditional road boundary line detection method, the calculation time of curve fitting is greatly reduced by clustering some stationary point clouds.

[0067] In this specification, the present invention has been described with reference to specific embodiments thereof. However, it is apparent that various modifications and variations may be made without departing from the spirit and scope of the present invention. Therefore, the specification and drawings should be regarded as illustrative rather than restrictive.

Claims

1. A road boundary detection method based on vehicle-mounted millimeter wave radar, characterized in that: The method comprises the following steps: (1) The vehicle-mounted millimeter-wave radar is rigidly connected to the vehicle and installed to obtain radar point cloud information of the current road scene and vehicle body information; (2) Based on the acquired absolute speed of the target, the moving point cloud is filtered out through the static-dynamic separation operation to determine the absolute speed of the point cloud; (3) Combined with the current vehicle motion state, road scene recognition is performed by applying different screening points and clustering strategies; (4) Determine the ROI area to perform the first screening of the static point cloud; (5) A density-based spatial clustering method is used to cluster the stationary point cloud after ROI screening to obtain point cloud clusters, and a second cluster screening is performed to obtain road boundary clusters; (6) The least squares method is used to perform curve fitting on the cluster points that meet the requirements of the second cluster screening, and the third cluster screening is performed at the same time to eliminate some noise clusters and non-road boundary clusters; (7) The Kalman filter algorithm is used to perform road boundary curve filtering on the currently acquired road boundary line to obtain the optimal road curve estimation result of the current frame and output the corresponding boundary curve.

2. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 1 is characterized in that: The step (1) is specifically as follows: The Cartesian coordinate system is specified with the center of mass of the vehicle as the origin of the coordinate system. The front of the vehicle is the positive x direction of the coordinate system, and the horizontal direction to the left of the vehicle is the positive y direction of the coordinate system. When the vehicle-mounted millimeter-wave radar detects the target point cloud, the target position is converted to the vehicle body coordinate system to determine the coordinates through the relative point cloud position, vehicle speed, yaw angular velocity and installation position information. The conversion method from point cloud speed to absolute vehicle speed is as follows: ； in, represents the absolute speed of the target, represents the radial velocity measured by the radar, represents the vehicle's speed, Indicates the point cloud orientation angle, represents the yaw angular velocity of the vehicle, Indicates the longitudinal installation position of the radar relative to the vehicle. Indicates the lateral installation position of the radar relative to the vehicle.

3. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 2 is characterized in that: The step (2) is specifically as follows: Set the velocity positive threshold , when the absolute value of the target absolute speed Less than threshold , the point cloud is considered to be stationary; otherwise, the point cloud is in motion, and the motion point cloud with speed is filtered out through the static-dynamic separation operation.

4. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 1 is characterized in that: The step (3) specifically includes the following steps: (3.1) Count the heights of the absolutely stationary point clouds in a fixed area and determine the ratio Hr of the total number of point clouds that meet the requirements. If Hr is greater than the empirical parameter Thr, the current vehicle road scene is judged to be a tunnel scene. Otherwise, proceed to the next step. (3.2) If the current vehicle speed is greater than the first preset speed or the vehicle speed and yaw angular velocity are greater than the first preset angular velocity, the current vehicle road scene is determined to be a high-speed scene, otherwise, proceed to the next step; (3.3) If the current vehicle speed is greater than the second preset speed or the vehicle speed and yaw angular velocity are greater than the second preset angular velocity, the current vehicle road scene is determined to be an urban scene, otherwise, proceed to the next step; (3.4) If the current vehicle speed is greater than the third preset speed, the current vehicle road scene is determined to be a low-speed scene, otherwise, proceed to the next step; (3.5) None of the above conditions are met, and the current vehicle road scene is stationary.

5. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 2 is characterized in that: The step (5) specifically includes the following steps: (5.1) First determine the threshold radius And the minimum number of points MinPts, select an unprocessed data point p from the stationary point cloud, and calculate the data point p in The number of data points in the neighborhood; (5.2) If the data point p is If the number of data points in the neighborhood is greater than or equal to the minimum number of points MinPts, then the data point p is marked as a core point; if the data point p is If the number of data points in the neighborhood is less than MinPts, but the point falls in the neighborhood of other core points, then the data point p is marked as a boundary point; if the data point p is neither a core point nor a boundary point, then it is marked as a noise point; (5.3) Through each core point, find all the data points that can be reached by its density to form a cluster. Repeat this process until all core points have been processed and all data points p have been marked as a cluster or a noise point. (5.4) After clustering is completed, the number of point clouds contained in each cluster, the longitudinal length and the lateral length of the cluster are traversed to determine whether it is a road boundary cluster.

6. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 5 is characterized in that: The step (6) is specifically as follows: The least squares method is used to perform curve fitting on the cluster points on the left and right sides of the vehicle that are most likely to be road boundary clusters in the following way: ； in, represents the quadratic coefficient of the fitting curve, represents the linear coefficient of the fitting curve, represents the constant term of the fitted curve, , is the vehicle coordinate.

7. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 5 is characterized in that: The step (6) further includes: The curvature of the fitting result is calculated as follows, and the yaw rate and curvature are used to calculate the curvature. Eliminate road boundaries that do not conform to reality: ； in, represents the first-order derivative of the curve at a certain position, Represents the second derivative of a curve at a certain position.

8. The road boundary line detection method based on vehicle-mounted millimeter wave radar according to claim 5 is characterized in that: The step (7) comprises: According to the rotation matrix , calculate the prior road curve fitting result of the previous frame after the vehicle turns at the next moment, specifically: ； in, Indicates the target horizontal coordinate after the vehicle turns. Indicates the target ordinate after the vehicle turns. Indicates the target horizontal coordinate of the previous frame before the vehicle turns. Indicates the target vertical coordinate of the previous frame before the vehicle turns. Indicates that the vehicle is in the radar frame period The angle of internal rotation, ; Get the new fitting curve coefficients according to the above formula: ； ； ； Using the current frame and the previous frame Six coefficients, based on the Kalman filter algorithm, to obtain stable coefficient output in the time domain; Get the road boundary quadratic curve: ； in, is the quadratic coefficient of the road fitting curve after smoothing by Kalman filtering, is the linear coefficient of the smoothed road fitting curve, is the constant term of the smoothed road fitting curve.

9. A road boundary line detection system based on a vehicle-mounted millimeter-wave radar for implementing the method according to any one of claims 1 to 8, characterized in that: The system comprises: Radar point cloud preprocessing module, used to obtain radar point cloud information and vehicle body information of the current road scene; A road scene recognition module is connected to the radar point cloud preprocessing module and is used to perform road scene recognition processing by applying different screening points and clustering strategies according to the current vehicle motion state; A point cloud clustering module, connected to the road scene recognition module, for clustering the screened point cloud using a spatial clustering method; A clustering screening module, connected to the point cloud clustering module, is used to screen out road boundary clusters that meet the requirements from the acquired point cloud; A curve fitting module, connected to the cluster screening module, for performing curve fitting processing on cluster points that meet the screening requirements; and The road boundary line filtering module is connected to the curve fitting module and is used to perform road boundary curve filtering processing on the currently acquired road boundary line using the Kalman filtering algorithm.

10. A road boundary detection device based on vehicle-mounted millimeter wave radar, characterized in that: The device comprises: a processor configured to execute computer executable instructions; A memory stores one or more computer executable instructions, and when the computer executable instructions are executed by the processor, the steps of the road boundary line detection method based on vehicle-mounted millimeter wave radar described in any one of claims 1 to 8 are implemented.

11. A road boundary detection processor based on vehicle-mounted millimeter wave radar, characterized in that: The processor is configured to execute computer executable instructions. When the computer executable instructions are executed by the processor, the steps of the road boundary line detection method based on vehicle-mounted millimeter wave radar described in any one of claims 1 to 8 are implemented.

12. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and the computer program can be executed by a processor to implement the steps of the road boundary line detection method based on a vehicle-mounted millimeter-wave radar as described in any one of claims 1 to 8.

Images