Road boundary width identification method and device, radar and readable storage medium

Abstract

The invention relates to a road boundary width identification method and device, a radar and a readable storage medium. The method comprises the following steps: obtaining point cloud information collected by a vehicle radar, performing position division on all static point clouds according to a plurality of distance resolution units preset in a plurality of target detection directions of the vehicle radar and transverse positions of the static point clouds, and determining a target distance resolution unit where a peak value of the number of the point clouds in each target detection direction is located; for each target detection direction, determining the single-side road boundary measurement width of the current frame in the corresponding target detection direction according to the distance value corresponding to the target distance resolution unit; filtering the single-side road boundary measurement width of the corresponding current frame according to the road width historical information to obtain the filtered single-side road boundary width of the current frame; and false targets in the stationary point cloud can be detected based on the single-side road boundary width of the current frame. By adopting the method, the detection performance of the radar in a closed scene can be improved.

Description

Technical Field

[0001] This application relates to the field of radar technology, and in particular to a method, apparatus, radar, and readable storage medium for identifying road boundary width. Background Technology

[0002] With the development of radar technology, millimeter-wave radar is increasingly widely used in various fields (e.g., vehicles, ships). Taking its application in vehicles as an example, with the development of automotive intelligence, millimeter-wave radar has many advantages over other sensors, such as low cost, immunity to weather conditions, and strong anti-interference capabilities in complex scenarios. However, its target detection performance also has some inherent defects in certain specific scenarios (such as vehicles driving in tunnels or other enclosed environments). Especially when the radar is in enclosed environments such as tunnels, it often detects multipath and mirror-image ghosts (false targets), leading to inaccurate detection and consequently affecting driving functions.

[0003] To improve vehicle safety in enclosed environments such as tunnels, a method is needed to enhance the detection performance of millimeter-wave radar in such environments. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, vehicle, computer-readable storage medium, and computer program product for road boundary width identification that can improve the detection performance of millimeter-wave radar in closed scenarios, in order to address the above-mentioned technical problems.

[0005] Firstly, this application provides a boundary width identification method, including:

[0006] The road curvature of the vehicle's driving path is determined and the point cloud information collected by the vehicle's radar is acquired. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0007] Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all stationary point clouds is divided, and the target range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units.

[0008] For each target detection direction, the single-sided road boundary measurement width of the current frame in the corresponding target detection direction is determined according to the distance value corresponding to the target distance resolution unit;

[0009] Obtain historical road width information, and filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0010] Based on the width of the single-sided road boundary, the road curvature, and the distance of the filtered current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted.

[0011] In one embodiment, before dividing all the stationary point clouds into positions based on a plurality of range resolution units preset in the target detection direction of the vehicle radar and the lateral positions of each of the stationary point clouds, the method further includes:

[0012] Obtain the turning radius of the vehicle, and determine the radius of curvature of the road based on the turning radius;

[0013] If the radius of curvature is less than a preset radius of curvature threshold, the lateral position of all the stationary point clouds is compensated so that the distribution of the stationary point clouds reaches a vertical state, and the compensated lateral position of the stationary point clouds is obtained.

[0014] In one embodiment, the method for determining the preset radius of curvature threshold includes:

[0015] A threshold constraint condition is determined, which is used to constrain that, without lateral position compensation, at least half of the stationary point clouds collected by the vehicle radar belong to the same range resolution unit in terms of lateral position.

[0016] The calculation relationship for the preset radius of curvature threshold is determined based on the threshold constraint conditions;

[0017] The maximum longitudinal distance that the vehicle can reach based on the road width and the preset width of the distance resolution unit are substituted into the calculation relationship to determine the preset radius of curvature threshold.

[0018] In one embodiment, the step of dividing all the stationary point clouds into positions based on multiple preset range resolution units in multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, and determining the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units, includes:

[0019] Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all the stationary point clouds is divided, and the intermediate range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units.

[0020] For each of the target detection directions, multiple candidate distance resolution units, including the intermediate distance resolution unit, are determined from the multiple distance resolution units according to their respective intermediate distance resolution units.

[0021] The candidate distance resolution units are divided into regions to obtain K sub-distance resolution units. The target distance resolution unit where the peak of the point cloud number is located is determined from the K sub-distance resolution units.

[0022] In one embodiment, the step of obtaining historical road width information and filtering the measured width of the single-sided road boundary of the corresponding current frame based on the historical road width information to obtain the filtered single-sided road boundary width of the current frame includes:

[0023] Obtain historical road width information;

[0024] Based on the total road width of the previous frame in the road width history information, the preset weight, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions, the filtered total road width of the current frame is obtained.

[0025] For each of the target detection directions, the width change is determined based on the road boundary width of the previous frame in the target detection direction in the road width history information, the filtered total road width of the current frame, the total road width of the previous frame, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions;

[0026] The filtered single-sided road boundary width of the current frame is obtained based on the road boundary width of the previous frame and the width change.

[0027] In one embodiment, the step of detecting a first stationary point cloud distributed along the target detection direction based on the width of the road boundary on one side, the road curvature, and the distance of the filtered current frame, and deleting false targets from the first stationary point cloud, includes:

[0028] Based on the width of the road boundary on one side of the filtered current frame, the road curvature, and the distance, determine the theoretical lateral position corresponding to the width of the road boundary on one side of the filtered current frame.

[0029] Obtain the lateral position of the first stationary point cloud distributed along the target detection direction. Based on the theoretical lateral position and the lateral position of the first stationary point cloud, determine the false targets in the first stationary point cloud and delete the false targets in the first stationary point cloud.

[0030] In one embodiment, prior to deleting false targets from the first stationary point cloud, the method further includes:

[0031] Obtain the jitter parameters of the total road width, the point cloud number threshold, and the ratio threshold;

[0032] For each target detection direction, determine the number of first stationary point clouds distributed in the first stationary point cloud in the target detection direction, and determine the ratio of the number of second stationary point clouds in the target range resolution unit to the total number of stationary point clouds;

[0033] The road width confidence level is determined based on the jitter parameter, the point cloud quantity threshold, the ratio threshold, the first stationary point cloud quantity, and the ratio.

[0034] If the confidence level of the road width is greater than the preset confidence level, then the false target is deleted.

[0035] Secondly, this application also provides a boundary width recognition device, comprising:

[0036] The data determination module is used to determine the road curvature of the vehicle's driving path and acquire point cloud information collected by the vehicle's radar. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0037] The position determination module is used to divide the position of all the stationary point clouds according to the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, and determine the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units.

[0038] The road boundary width determination module is used to determine the single-sided road boundary measurement width of the current frame in the target detection direction according to the distance value corresponding to the target distance resolution unit for each target detection direction;

[0039] The filtering module is used to acquire historical road width information, and to filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0040] The false target deletion module is used to detect the first stationary point cloud distributed in the target detection direction based on the width of the single-sided road boundary, the road curvature, and the distance of the filtered current frame, and to delete false targets in the first stationary point cloud.

[0041] Thirdly, this application also provides a radar, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0042] The road curvature of the vehicle's driving path is determined and the point cloud information collected by the vehicle's radar is acquired. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0043] Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all stationary point clouds is divided, and the target range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units.

[0044] For each target detection direction, the single-sided road boundary measurement width of the current frame in the corresponding target detection direction is determined according to the distance value corresponding to the target distance resolution unit;

[0045] Obtain historical road width information, and filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0046] Based on the width of the single-sided road boundary, the road curvature, and the distance of the filtered current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted.

[0047] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0048] The road curvature of the vehicle's driving path is determined and the point cloud information collected by the vehicle's radar is acquired. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0049] Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all stationary point clouds is divided, and the target range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units.

[0050] For each target detection direction, the single-sided road boundary measurement width of the current frame in the corresponding target detection direction is determined according to the distance value corresponding to the target distance resolution unit;

[0051] Obtain historical road width information, and filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0052] Based on the width of the single-sided road boundary, the road curvature, and the distance of the filtered current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted.

[0053] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0054] The road curvature of the vehicle's driving path is determined and the point cloud information collected by the vehicle's radar is acquired. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0055] Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all stationary point clouds is divided, and the target range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units.

[0056] For each target detection direction, the single-sided road boundary measurement width of the current frame in the corresponding target detection direction is determined according to the distance value corresponding to the target distance resolution unit;

[0057] Obtain historical road width information, and filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0058] Based on the width of the single-sided road boundary, the road curvature, and the distance of the filtered current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted.

[0059] The aforementioned road boundary width identification method, device, radar, computer-readable storage medium, and computer program product acquire all stationary point clouds collected by the vehicle radar. Based on multiple preset range resolution units in multiple target detection directions of the vehicle radar, the positions of all collected stationary point clouds are divided to determine the concentrated distribution area of ​​stationary point clouds in each target detection direction. The target range resolution unit is determined based on the number of stationary point clouds in the concentrated distribution area of ​​stationary point clouds in each target detection direction. Then, the single-sided road boundary measurement width of the current frame in each target detection direction is determined based on the distance value corresponding to the target range resolution unit. The single-sided road boundary measurement width of the current frame is filtered to ensure its accuracy. Based on this, the first stationary point cloud distributed in the target detection direction is detected based on the filtered single-sided road boundary width, road curvature, and distance of the current frame, deleting false targets outside the road, reducing the false alarm rate of radar detection, and thus improving the radar's detection performance. This method determines the road boundary width in each target detection direction based on all collected stationary point clouds, weakening the influence of false point clouds on the road boundary width identification result. Moreover, this method is not limited by environmental influences. Attached Figure Description

[0060] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0061] Figure 1 This is an application environment diagram of the road boundary width recognition method in one embodiment;

[0062] Figure 2 This is a flowchart illustrating a road boundary width recognition method in one embodiment;

[0063] Figure 3 This is a flowchart illustrating step 204 in one embodiment;

[0064] Figure 4 This is a flowchart illustrating a method for determining road width confidence in one embodiment;

[0065] Figure 5 This is a flowchart illustrating the road boundary width recognition method in another embodiment;

[0066] Figure 6 This is a flowchart illustrating the tunnel boundary width identification method in another embodiment;

[0067] Figure 7 This is a structural block diagram of a road boundary width recognition device in one embodiment;

[0068] Figure 8 This is a diagram of the internal structure of a vehicle in one embodiment. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0070] With the development of automotive intelligence, millimeter-wave radar is increasingly widely used in automobiles. However, in enclosed environments such as tunnels, millimeter-wave radar often detects multipath and mirror-image false targets, affecting driving functions. To address this technical problem, some technologies replace the millimeter-wave radar, but hardware replacement increases costs. Alternatively, one approach utilizes the location information of stationary point clouds around the vehicle, using clustering to select the closest stationary points on the left and right sides of the vehicle and then performing curve fitting on these points to obtain the road boundary curve. However, this solution heavily relies on the location information of the selected stationary point clouds and has high environmental requirements, requiring the absence of interfering stationary points. When the radar is in enclosed environments such as tunnels, it detects many interfering stationary points, resulting in poor fitting performance and consequently, inaccurate road boundary recognition, leading to poor radar detection performance.

[0071] Therefore, to improve the detection performance of millimeter-wave radar in closed scenarios, a road boundary width recognition method is proposed. This method comprehensively considers the position information of all stationary point clouds, is insensitive to the position information of a single stationary point, and is therefore insensitive to the detected interfering stationary point clouds. In tunnel scenarios, it can output boundary width information more accurately, thereby improving the detection performance of millimeter-wave radar in closed scenarios.

[0072] The road boundary width identification method provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, radar 102 communicates with mobile carrier 104 via a network. Radar 102 is mounted on mobile carrier 104. Radar 102 can be of different types of millimeter-wave radar, and it can be positioned at different locations on mobile carrier 104 according to actual needs; no specific limitation is made here. Mobile carrier 104 is an entity or platform capable of carrying or transporting objects, information, energy, etc., and moving in physical space or a specific environment. Mobile carrier 104 can be, but is not limited to, vehicles, ships, aircraft, logistics equipment, etc. In this embodiment, a vehicle is used as an example to illustrate the mobile carrier 104. Radar 102 is mounted on the vehicle and, as the vehicle moves, can collect all stationary point clouds along the vehicle's travel path in real time. Based on the real-time collected stationary point clouds, the width of the single-sided road boundary of the current frame at the corresponding moment can be identified, and false targets in the stationary point cloud in the target detection direction can be filtered out.

[0073] Radar 102 determines the road curvature of the vehicle's driving path and acquires point cloud information collected by the vehicle radar. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud. Based on multiple preset range resolution units in multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all stationary point clouds is divided. From the multiple range resolution units, the target range resolution unit where the peak number of point clouds in each target detection direction is located is determined. For each target detection direction, the single-sided road boundary measurement width of the current frame in the corresponding target detection direction is determined based on the distance value corresponding to the target range resolution unit. Historical road width information is acquired, and the single-sided road boundary measurement width of the corresponding current frame is filtered based on the historical road width information to obtain the filtered single-sided road boundary width of the current frame. Based on the filtered single-sided road boundary width, road curvature, and distance of the current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted.

[0074] In one exemplary embodiment, such as Figure 2 As shown, a method for road boundary width recognition is provided, which can be applied to... Figure 1 Taking the radar in the example, the explanation includes the following steps 202 to 206. Wherein:

[0075] Step 202: Determine the road curvature of the vehicle's driving path and acquire point cloud information collected by the vehicle's radar. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0076] It should be noted that the road in this embodiment refers to a road in a closed environment where vehicles travel. A closed environment can be understood as one where radar wave propagation is strongly interfered with by reflections from surrounding surfaces (such as walls or ceilings), causing its signal characteristics to differ from those in an open environment. Therefore, such an environment can be classified as a "closed environment." For example, a scenario like a tunnel, where radar waves are easily reflected by walls, can be considered a closed environment. In this embodiment, a tunnel is used as an example of a closed environment, so the road the vehicle travels on is a tunnel.

[0077] For a curved tunnel, the tunnel curvature C at various points along the tunnel can be determined based on the curvature at the starting point, the rate of change of curvature, and the tunnel arc length. The tunnel curvature C can be expressed as: Where C is the curvature at every point on the road. The curvature at the starting point, Let C be the rate of change of curvature, and L be the tunnel arc length. When the rate of change of road curvature is small, a curved road can be approximated by a circular arc, i.e., C = .

[0078] The lateral position can be understood in the vehicle coordinate system as follows: with the vehicle's forward direction (i.e., the longitudinal axis, X-axis) as a reference, the lateral (Y-axis) refers to the horizontal offset direction perpendicular to the vehicle's direction of travel, including both left and right sides. The determination of the lateral position can be achieved through existing methods, which will not be elaborated here. Stationary point clouds can be point cloud data of stationary objects in the environment (such as curbs, guardrails, stationary vehicles, etc.) collected by millimeter-wave radar.

[0079] Step 204: Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, divide all stationary point clouds into positions, and determine the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units.

[0080] When scanning the surrounding environment, the radar covers different directional areas, such as front, rear, left, and right, meaning there are multiple detection directions. The target detection direction is determined from these multiple detection directions to determine the target's orientation used to determine the road boundary width. For example, the target detection directions include two directions: the left and right sides of the radar. These are the first target detection direction corresponding to the left side of the radar and the second target detection direction corresponding to the right side. These two target detection directions can be horizontal offset directions perpendicular to the vehicle's direction of travel. The principle for determining the width of the single-sided road boundary measurement in the current frame for each target detection direction is the same. To avoid repetition, only the processing of one target detection direction will be discussed.

[0081] The same number of range resolution units can exist in each target detection direction, and the distance between the range resolution units can be set according to actual needs. For example, the distance within a preset lateral distance ±W on both sides of the radar can be divided into N range bins (i.e., range resolution units) with a preset step width of m. The position division of stationary point clouds can be based on the lateral position of each stationary point cloud and the left / right W of the radar, divided by a preset step width. The integer obtained is then used to determine the corresponding X-th range bin. For example, a stationary point cloud with a lateral position of PosY can be divided using... =ceil((PosY +W) / m) determines the range-resolved cell to which the stationary point cloud belongs, where ceil represents rounding up. Let X represent the Xth distance bin, where X ≤ N. For example, if m is 1, then the distance corresponding to the first distance resolution cell is 1, the distance corresponding to the second distance resolution cell is 2, and the stationary point cloud with a lateral position of (1, 2) will fall into the second distance resolution cell, and so on.

[0082] The preset step width of the target range resolution unit can be equal to or unequal to the preset step widths of multiple preset range resolution units. For example, the preset step width of the target range resolution unit can be smaller than the preset step widths of multiple preset range resolution units. Position division of all stationary point clouds can be performed at least once, and at least once, a filtering process can be conducted.

[0083] For example, based on multiple preset range resolution units and the lateral position of each stationary point cloud in multiple target detection directions of the vehicle radar, the stationary point cloud belonging to each target detection direction is determined. Then, for each target detection direction, based on the preset step width, the number of range resolution units and the preset lateral distance corresponding to the preset range resolution units in each target detection direction, the position of the stationary point cloud belonging to each target detection direction is divided, the range resolution unit into which each stationary point cloud falls is determined, and the range resolution unit corresponding to the largest number of stationary point clouds is determined as the target range resolution unit.

[0084] Step 206: For each target detection direction, determine the single-sided road boundary measurement width of the current frame in the corresponding target detection direction based on the distance value corresponding to the target distance resolution unit.

[0085] The measured width of the single-sided road boundary in the current frame can be the width of the single-sided road boundary in the current frame, which refers to the lateral distance between the vehicle and the boundary on one side (left or right) of the road. It can be understood that the principle of measuring the width of the single-sided road boundary is to assign the lateral positions of all stationary point clouds to specific distance bins, and then determine which distance bin contains the most points; the lateral position corresponding to that distance bin is considered to be the lateral position of the road wall.

[0086] Step 208: Obtain historical road width information, and filter the measured width of the single-sided road boundary in each current frame based on the historical road width information to obtain the filtered single-sided road boundary width.

[0087] It is understandable that vehicles may change lanes while driving on the road, so the width of the road on the left and right sides changes in reality. That is, the width of the road in each target detection direction changes in reality. In order to avoid the error from both sides from accumulating and causing fluctuations in the total width, and to further ensure the accuracy of the single-sided road boundary measurement width in the current frame, it is necessary to filter the total road width. Based on the filtered total road width, the road width historical messages, and the single-sided road boundary measurement width of each current frame, the filtered road boundary width in each target detection direction is determined.

[0088] The historical road width information can include the filtered width of the single-sided road boundary of the previous frame in each target detection direction and the filtered total road width of the previous frame. The filtered width of the single-sided road boundary of the current frame in each target detection direction can be obtained by first filtering the total road width, and then filtering the measured width of the single-sided road boundary of the current frame under the premise that the total width is stable.

[0089] The filtered total road width in the current frame can be determined based on the filtered total road width of the previous frame and the measurement residual of the current frame. The measurement residual of the current frame is determined by subtracting the filtered total road width of the previous frame from the measured total road width. The width of a single-sided road boundary can be determined based on the single-sided road boundary width of the previous frame and the corresponding change in the single-sided road boundary width in the current frame. The change in the single-sided road boundary width is determined by multiplying the ratio of the change in the total road width determined in the previous and current frames to the measurement residual by the difference between the measured width of the single-sided road boundary in the current frame and the single-sided road boundary width in the previous frame.

[0090] For example, historical road width information is obtained, and the total road width is filtered by measuring the width of the single-sided road boundary in each current frame based on the historical road width information to obtain the total road width of the current frame. For the width of the single-sided road boundary in each target detection direction, it is determined based on the change in the width of the single-sided road boundary in the previous frame and the width of the single-sided road boundary in the current frame.

[0091] Step 210: Based on the width of the single-sided road boundary, road curvature, and distance of the current frame after filtering, detect the first stationary point cloud distributed in the target detection direction and delete false targets in the first stationary point cloud.

[0092] The distance can be the road arc length. For example, for each target detection direction, the corresponding theoretical lateral position is determined based on the width of the single-sided road boundary, the road curvature, and the distance of the current frame. The first stationary point cloud distributed in the target detection direction is detected based on the theoretical lateral position. If there is a false point cloud whose lateral position is greater than the theoretical lateral position, the first stationary point cloud is identified as a false target and deleted.

[0093] The aforementioned road boundary width recognition method acquires all stationary point clouds collected by the vehicle radar. Based on multiple preset range resolution units in multiple target detection directions of the vehicle radar, the positions of all collected stationary point clouds are divided to determine the concentrated distribution area of ​​stationary point clouds in each target detection direction. The target range resolution unit is determined based on the number of stationary point clouds in the concentrated distribution area of ​​stationary point clouds in each target detection direction. Then, the single-sided road boundary measurement width of the current frame in each target detection direction is determined based on the distance value corresponding to the target range resolution unit. The single-sided road boundary measurement width of the current frame is filtered to ensure its accuracy. On this basis, based on the filtered single-sided road boundary width, road curvature, and distance of the current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets outside the road are deleted, reducing the false alarm rate of radar detection and thus improving the radar's detection performance. In this method, the road boundary width in each target detection direction is determined based on all collected stationary point clouds, which weakens the influence of false point clouds on the road boundary width recognition result. Moreover, this method is not limited by environmental influences.

[0094] It's understandable that roads traveled by vehicles may have curves, such as curved tunnels. The lateral position of the stationary point cloud acquired by radar will be affected by the curvature of the tunnel, causing the point cloud distribution to deviate from the actual physical boundaries. For example, the curvature deformation may be mistaken for the actual shape of an obstacle. Therefore, to address this situation, it is necessary to compensate for the lateral position of the stationary point cloud, so that the distribution of the stationary point cloud is vertical, preserving the true boundary features.

[0095] In an exemplary embodiment, before dividing all stationary point clouds into positions based on a plurality of range resolution units preset in the target detection direction of the vehicle radar and the lateral positions of each stationary point cloud, the method further includes:

[0096] Obtain the turning radius of the vehicle and determine the curvature radius of the road based on the turning radius; if the curvature radius is less than the preset curvature radius threshold, compensate the lateral position of all stationary point clouds so that the distribution of the stationary point clouds reaches a vertical state, and obtain the compensated lateral position of the stationary point clouds.

[0097] It's understandable that when a vehicle is driving on a road, there's a lack of prior information about the road's curvature, and the exact curvature information is generally impossible to calculate directly. Therefore, we consider using the vehicle's turning radius, turnR, to approximate the road's curvature radius, R. The formula for calculating the vehicle's turning radius can be found through... This formula is achieved, where... It is the vehicle's speed. It is the yaw rate of the vehicle.

[0098] For scenarios with a small turning radius (turnR), assuming a small road curvature radius (R) and a large road curvature (C), it is necessary to compensate for the lateral position of the stationary point cloud to ensure its vertical distribution. The compensation principle includes the following:

[0099] For a curved road, the curvature C at various points along the road can be expressed by the formula... Let C represent the curvature at every point on the road. The curvature at the starting point, Let C be the rate of change of curvature, and L be the arc length. When the rate of change of road curvature is small, a curved road can be approximated by a circular arc, i.e., C = Therefore, with the radar as the origin, the coordinates (PosY, PosX) of any location on the road boundary can be expressed as: , Where d represents the lateral deviation between the radar and the road boundary, and α represents the angle between the radar normal and the road edge. Generally, the value of α is relatively small and can be ignored, and the arc length L can be replaced by the radial distance of the point cloud.

[0100] After straightening the curved road, the lateral position of the tunnel boundary point cloud can be obtained. This formula is used to calculate the compensated lateral position. In the diagram, PosY represents the lateral position of the point cloud detected by the radar, C represents the road curvature, which can be replaced by the turning radius of the vehicle; L represents the arc length of the road, which can be approximated by the longitudinal distance of the point cloud. Furthermore, considering that the turning radius of the vehicle cannot perfectly represent the road curvature radius (i.e., turnR cannot accurately represent the actual road curvature radius R), this method can be used to optimize the statistical performance: R1 = a preset coefficient × turnR is used to replace R. The preset coefficient can range from 0.8 to 1.2. The lateral position of the stationary point cloud is compensated, and then the number of peak point clouds in each distance bin under different preset coefficients is counted. The coefficient corresponding to the highest number of peak point clouds is used as the final preset coefficient. It can be understood that the tunnel curvature C can be used... Calculated.

[0101] To ensure the accuracy of road boundary width recognition, it is necessary to accurately determine a preset curvature radius threshold. In an exemplary embodiment, the preset curvature radius threshold is determined as follows:

[0102] Determine threshold constraints to ensure that, without lateral position compensation, at least half of the stationary point clouds acquired by the vehicle radar belong to the same range resolution unit in terms of lateral position. Based on the maximum longitudinal distance that the vehicle can identify based on the road width, the preset width of the range resolution unit, and the threshold constraints, determine a preset radius of curvature threshold.

[0103] The threshold constraint can be expressed as follows: , To identify the maximum achievable longitudinal distance for the designed road width, `length(bin)` represents the preset width corresponding to a distance resolution unit, and `median` represents the operation of calculating the median. A preset curvature radius threshold is set. It should be noted that determining the preset curvature radius threshold is essentially a statistical constraint problem. It is necessary to use median operations to ensure coverage of at least 50% of the scene. The designed maximum vertical distance and preset width may differ from the actual effective values ​​in the real environment. Therefore, the preset curvature radius threshold needs to be determined by using median operations.

[0104] In an exemplary embodiment, for cases where the radius of curvature is less than a preset radius of curvature threshold, this can be addressed by supplementing the lateral position of the stationary point cloud, or by straightening the curved tunnel. For each target detection direction, different stationary point clouds are selected to initially fit multiple radii of curvature, and then the most reasonable radius of curvature is selected to straighten the curved tunnel. The most reasonable radius of curvature can be determined based on the maximum number of peak point clouds.

[0105] In the above embodiments, if the radius of curvature is less than a preset radius of curvature threshold, the lateral position of all stationary point clouds is compensated so that the distribution of the stationary point clouds reaches a vertical state, and the lateral position of the stationary point clouds after compensation is obtained. By compensating the lateral position of the point clouds until the distribution of the point traces is vertical, the accuracy can be improved and the situation of inaccurate road boundary width caused by not performing compensation operations on the curved wall point clouds detected by radar in curved tunnels can be avoided.

[0106] In one exemplary embodiment, such as Figure 3 As shown, step 204 includes steps 302 to 306. Wherein:

[0107] Step 302: Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, divide all stationary point clouds into positions, and determine the intermediate range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units.

[0108] The preset step width of the intermediate distance resolution unit is equal to the preset step width of the preset distance resolution unit.

[0109] For example, based on multiple preset range resolution units and the lateral position of each stationary point cloud in multiple target detection directions of the vehicle radar, the stationary point cloud belonging to each target detection direction is determined. Then, for each target detection direction, based on the preset step width, the number of range resolution units, and the preset lateral distance corresponding to the preset range resolution units in each target detection direction, the position of the stationary point cloud belonging to each target detection direction is divided, the range resolution unit into which each stationary point cloud falls is determined, and the range resolution unit corresponding to the largest number of stationary point clouds is determined as the intermediate range resolution unit.

[0110] Step 304: For each target detection direction, determine multiple candidate range resolution units, including the intermediate range resolution unit, from multiple range resolution units according to their respective intermediate range resolution units.

[0111] Among them, multiple candidate range resolution units are consecutive. The number of multiple candidate range resolution units can be determined according to actual needs, such as data accuracy and time consumption. The number of multiple candidate range resolution units can be three or four, and there is no limitation here.

[0112] Step 306: Divide the multiple candidate range resolution units into regions to obtain K sub-range resolution units, and determine the target range resolution unit where the peak of the point cloud number is located from the K sub-range resolution units.

[0113] Taking a scenario with three candidate range resolution units as an example, focusing on the radar's first and second target detection directions, i.e., the left and right sides of the radar. The distances within ±W ranges on both sides of the radar are divided into N range bins with a step width of m. Then, a stationary point with a lateral position of PosY will be... =ceil((PosY+W) / m) falls within its respective interval, where ceil represents rounding up. Let X represent the Xth distance bin, where X ≤ N. Find the bins containing the peak number of stationary point clouds in the left and right distance bins respectively, i.e., the middle distance resolution unit on the left side of the radar, maxIdx_Left = max(bin[1~N / 2]), and the middle distance resolution unit on the right and left sides of the radar, maxIdx_right = max(bin[N / 2+1~N]). Taking the left side of the radar as an example, determine Bin[maxIdx_Left-1], Bin[maxIdx_Left], and Bin[maxIdx_Left+1] as candidate distance resolution units. Divide the position areas represented by Bin[maxIdx_Left-1], Bin[maxIdx_Left], and Bin[maxIdx_Left+1] into K sub-distance resolution units bin2 on an average basis, and count the number of point clouds falling into distance bin2 again. Determine the target distance resolution unit containing the peak number of point clouds from the K sub-distance resolution units. Set the distance value corresponding to the target distance resolution unit as the single-frame road boundary measurement width on the left side of the radar. The width of the road boundary measured in a single frame on the right side of the radar can be determined in the same way as the width of the road boundary measured in a single frame on the left side of the radar, which will not be elaborated here.

[0114] In this embodiment, for each side of the radar, a preliminary selection is performed from a set of preset range resolution units to determine the intermediate range resolution unit in the initial screening. Low-density units with obvious noise interference are quickly filtered out. The intermediate range resolution unit is used as a reference for secondary division to determine multiple sub-range resolution units. From the multiple sub-range resolution units, a fine selection is performed to determine the target range resolution unit where the peak of the point cloud number is located. This accurately ensures the effectiveness of the target. This method, through the synergy of coarse and fine selection, can ensure the effectiveness and accuracy of detection.

[0115] When driving on the road, lane changes occur, so the width of the road on the left and right sides changes in reality. However, the total width of the road generally remains unchanged. Therefore, the total width of the road is filtered, and the filtered single-side road boundary width can be determined based on the filtered total width of the road.

[0116] In an exemplary embodiment, historical road width information is obtained, and the measured width of the single-sided road boundary of the corresponding current frame is filtered based on the historical road width information to obtain the filtered single-sided road boundary width of the current frame, including:

[0117] Obtain historical road width information; based on the total road width of the previous frame, preset weights, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions from the historical road width information, obtain the filtered total road width of the current frame; for each target detection direction, determine the width change based on the road boundary width of the previous frame in the target detection direction from the historical road width information, the filtered total road width of the current frame, the total road width of the previous frame, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions; based on the road boundary width of the previous frame and the width change, obtain the filtered single-sided road boundary width of the current frame.

[0118] The filtered total road width of the current frame is determined by multiplying the difference between the sum of the absolute values ​​of the measured widths of the single-sided road boundaries of the current frame in each target detection direction and the filtered total road width of the previous frame, along with a first preset weight. This difference is then summed with the filtered total road width of the previous frame to determine the filtered total road width of the current frame. It can be represented as:

[0119] ;in, This represents the total width of the road in the previous frame after filtering. abs represents taking the absolute value. leftWidth represents the measured width of the left road boundary in the current frame, and rightWidth represents the measured width of the right road boundary in the current frame.

[0120] For each target detection direction, the width of the single-sided road boundary in the current frame after filtering is determined by summing the width of the single-sided road boundary after filtering in the previous frame and the width change. The width change is determined by dividing the difference between the total road width in the current frame and the total road width in the previous frame by the sum of the absolute values ​​of the measured widths of the single-sided road boundaries in the current frame in each target detection direction and subtracting the difference determined by the total road width in the previous frame after filtering. This ratio is then multiplied by the difference determined by the measured width of the single-sided road edge in the current frame and the width of the single-sided road boundary in the previous frame.

[0121] Taking the target detection directions as the left and right sides of the radar as an example, the width of the left road boundary in the filtered current frame is... and the width of the right-hand road boundary They are represented as follows:

[0122] = + ×(leftWidth- );in, This represents the width of the left road boundary after filtering in the previous frame;

[0123] = + ×(rightWidth- ); This represents the width of the right-side road boundary after filtering in the previous frame.

[0124] This method determines the accuracy of road boundary width in each target detection direction by updating the total road width in the current frame based on historical road width information through filtering. Compared with filtering the road boundary width in a single target detection direction, this method is more reasonable and can ensure the accuracy of road width even if the total road width changes.

[0125] Based on the above embodiments, it can be seen that road boundary width recognition can identify false targets in a stationary point cloud. In an exemplary embodiment, based on the single-sided road boundary width, road curvature, and distance of the filtered current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted, including:

[0126] Based on the width of the road boundary on one side, the road curvature, and the distance of the current frame after filtering, determine the theoretical lateral position corresponding to the width of the road boundary on one side of the current frame after filtering; obtain the lateral position of the first stationary point cloud distributed in the target detection direction; based on the theoretical lateral position and the lateral position of the first stationary point cloud, determine the false targets in the first stationary point cloud and delete the false targets in the first stationary point cloud.

[0127] Specifically, for each target detection direction, the theoretical lateral position of the road wall at different distances can be calculated based on the road curvature C and the width of the road wall (i.e., the width of the filtered single-sided road boundary), thereby deleting false targets outside the corresponding road wall. The theoretical lateral position can be determined by calculating the product of the road curvature and the distance, and then summing this product with the width of the filtered single-sided road boundary. Taking the left side of the radar as an example, the theoretical lateral position... It can be represented as: LeftWidth (after filtering) + 0.5 × C × ; is the theoretical lateral position of the single-sided road boundary width after filtering at different distances, LeftWidth is the left road boundary width after filtering, C is the tunnel curvature, and L is the radial distance or longitudinal distance of the point cloud detected by the radar, which can be obtained directly.

[0128] For example, based on the width of the road boundary on one side, the road curvature, and the distance of the filtered current frame, the theoretical lateral position corresponding to the width of the road boundary on one side of the filtered current frame is determined; for the width of the road boundary on one side at different times, the lateral position of the first stationary point cloud distributed in the target detection direction is obtained; based on the theoretical lateral position and the lateral position of the first stationary point cloud, stationary point clouds with a lateral distance greater than the theoretical lateral position are identified as false targets, and false targets in the first stationary point cloud are deleted.

[0129] In the above embodiments, the width of the road boundary on one side is determined based on the above method, thereby determining the theoretical lateral position at different distances. Based on the theoretical lateral position, false targets detected by the radar can be accurately screened out, so as to improve the detection performance of the radar.

[0130] To further improve the accuracy of false target detection and enhance radar detection performance, the confidence level of the width of the road boundary on one side can be calculated before deleting false targets. This improves the accuracy of identification. In an exemplary embodiment, such as... Figure 4 As shown, a method for determining the confidence level of road width is provided, including the following steps:

[0131] Step 402: Obtain the jitter parameters of the total road width, the point cloud quantity threshold, and the ratio threshold.

[0132] It is understandable that the confidence level of the single-sided road boundary width is determined in the same way for different target detection directions. In this example, the single-sided road boundary width and the current frame are used as examples for illustration.

[0133] The point cloud quantity threshold and the ratio threshold can both be determined according to actual needs, and will not be elaborated here.

[0134] Step 404: For each target detection direction, determine the number of first stationary point clouds distributed in the first stationary point cloud in the target detection direction, and determine the ratio of the number of second stationary point clouds in the target range resolution unit to the total number of stationary point clouds.

[0135] The determination of the number of first stationary point clouds distributed in the target detection direction can be achieved by determining the number of stationary point clouds selected in each target detection direction from all stationary point clouds collected by the radar for each frame.

[0136] Step 406: Determine the road width confidence level based on the jitter parameters, point cloud quantity threshold, ratio threshold, first stationary point cloud quantity and ratio.

[0137] For example, the corresponding jitter parameter weights are determined based on the jitter parameters, and a first adjustment coefficient, a second adjustment coefficient, and a custom weight are determined. The difference between the first number of stationary point clouds and the point cloud number threshold is multiplied by the first adjustment coefficient and then added to a first preset value (e.g., 50), and then multiplied by the custom weight to obtain a first value. The difference between the ratio of the second number of stationary point clouds in the target distance resolution unit to the total number of stationary point clouds and the ratio threshold is multiplied by the second adjustment coefficient and then added to a second preset value (e.g., 50) to obtain a sum. This sum is then multiplied by 1 and the difference between the custom weights to obtain a product. This product is multiplied by the jitter parameter weights to obtain a second value. The first value and the second value are summed to obtain the road width confidence score.

[0138] Road width confidence The effect of the number of stationary point clouds on the execution accuracy of the single-sided road boundary width can be determined based on the influence of the number of stationary point clouds on the execution accuracy of the single-sided road boundary width and the influence of the percentage of peak point cloud counts on the execution accuracy of the single-sided road boundary width. The influence of the number of stationary point clouds on the execution accuracy of the single-sided road boundary width can be expressed as follows: The impact of the percentage of peak point cloud counts on the execution degree of the width of a single-sided road boundary can be expressed as: Road width confidence It can be represented as: ;in, This represents the confidence level of the single-sided road boundary width in each target detection direction (e.g., left / right side of the radar); β is a user-defined weight; the first adjustment coefficient. Second adjustment coefficient This is a custom parameter; Num represents the number of stationary point clouds filtered out on the left and right sides of the current frame. The threshold for the number of points in the custom point cloud; P is the proportion of the number of points in the left / right peak distance bin of the current frame to the total number of points in the point cloud; For a custom proportional threshold; The weights for the custom jitter parameters are determined based on the jitter levels across the tunnel width (i.e., the jitter parameter itself). It's important to note that... ≤50; ≤50.

[0139] Step 408: If the confidence level of the road width is greater than the preset confidence level, then delete the false target.

[0140] For example, based on the width of the road boundary on one side of the filtered current frame, the road curvature, and the distance, the theoretical lateral position corresponding to the width of the road boundary on one side of the filtered current frame is determined; the lateral position of the first stationary point cloud distributed along the target detection direction is obtained; and false targets in the first stationary point cloud are determined based on the theoretical lateral position and the lateral position of the first stationary point cloud. If the road width confidence score is greater than a preset confidence score, the false targets are deleted.

[0141] Optionally, in an exemplary embodiment, in addition to the above-described method, the dispersion of the stationary point cloud can be added to the above-described method to calculate the road width confidence when determining the road width confidence, so as to improve the accuracy of the confidence.

[0142] In the above embodiments, the confidence level of the corresponding road width is calculated from three perspectives: the jitter of the tunnel width, the number of stationary point clouds, and the proportion of the number of peak points in the point cloud in each target detection direction, thereby improving the accuracy of the recognition results.

[0143] In one exemplary embodiment, such as Figure 5 As shown, a method for road boundary width recognition is provided, which can be applied to... Figure 1 Taking the radar in the image as an example, the following steps are included:

[0144] Step 502: Obtain the turning radius of the vehicle, determine the curvature radius of the road based on the turning radius, and obtain the point cloud information collected by the vehicle radar. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0145] Step 504: If the radius of curvature is less than the preset radius of curvature threshold, the lateral position of all stationary point clouds is compensated so that the distribution of the stationary point clouds reaches a vertical state, and the lateral position of the stationary point clouds after compensation is obtained.

[0146] Step 506: Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, divide all stationary point clouds into positions, and determine the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units.

[0147] Step 508: For each target detection direction, determine the single-sided road boundary measurement width of the current frame in the corresponding target detection direction based on the distance value corresponding to the target distance resolution unit.

[0148] Step 510: Obtain historical road width information, and filter the measured width of the single-sided road boundary of the corresponding current frame based on the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0149] Step 512: Calculate the road width confidence score for the single-sided road boundary width in the current frame.

[0150] Step 514: Calculate the theoretical lateral position corresponding to the width of the road boundary on one side, and delete false targets based on the road width confidence and the theoretical lateral position.

[0151] It should be noted that the specific implementation steps of this example can be achieved in the manner described above, and will not be repeated here. This example uses the current frame as an example for illustration. During the vehicle's movement, the above method can be executed based on each data frame acquired by the radar to remove false targets.

[0152] In the aforementioned road boundary width recognition method, the lateral distance on both sides of the radar is divided into multiple distance bins. Then, stationary point clouds are assigned to the corresponding distance bins based on their lateral positions. Finally, the width of the road wall is determined based on the number of point clouds in each distance bin. For roads with a radius of curvature smaller than a preset curvature radius threshold, the lateral position of the stationary point clouds is compensated until the point cloud distribution is vertical before statistical analysis to ensure the accuracy of width recognition. Based on this, the final output road width is filtered to obtain the filtered single-sided road width. Furthermore, the confidence level of the filtered road width is calculated based on information such as the proportion of peak point clouds in the distance bins, road width jitter, and the number of point clouds, improving the accuracy of road width recognition. In addition, road boundary width information can be output in real time.

[0153] The following example uses a closed scene with a tunnel as an illustration. The measured width of a single-sided road boundary is the same as the measured width of a single-sided tunnel boundary, the total road width is the same as the total tunnel width, and the width of a single-sided road boundary is the same as the width of a single-sided tunnel boundary. This is illustrated using the target detection directions as the left and right sides of the radar. Figure 6 As shown, a tunnel boundary width identification method based on the road boundary width identification method described above is provided, which specifically includes the following:

[0154] The turning radius turnR of the vehicle is used to approximate the curvature radius R of the tunnel. The turning radius turnR of the vehicle can be calculated using the following formula:

[0155] (1)

[0156] In formula (1), It is the vehicle's speed. This is achieved by adjusting the yaw rate of the vehicle as described above. The radius of curvature R is judged; if it is less than a preset radius of curvature threshold, the lateral position of all stationary point clouds is compensated to make the distribution of the stationary point clouds vertical, thus obtaining the compensated lateral position of the stationary point clouds. The preset radius of curvature threshold can be determined using the method described above, and will not be elaborated here. Specifically, the compensation principle includes the following:

[0157] For a curved tunnel, the curvature C at various points along the tunnel can be expressed by formula (2):

[0158] (2)

[0159] In equation (2) above, C represents the curvature of each point on the road. The curvature at the starting point, Let C be the rate of change of curvature, and L be the arc length. When the rate of change of road curvature is small, a curved road can be approximated by a circular arc, i.e., C = Therefore, with the radar as the origin, the coordinates (PosY, PosX) of any location on the road boundary can be approximately represented by formulas (3) and (4):

[0160] (3)

[0161] (4)

[0162] Where d represents the lateral deviation between the radar and the road boundary, and α represents the angle between the radar normal and the road edge. Generally, the value of α is relatively small and can be ignored. The arc length L can be replaced by the radial distance of the point cloud. Therefore, after straightening the curved tunnel, the lateral position of the point cloud at the tunnel boundary can be calculated using formula (5):

[0163] (5)

[0164] Where PosY is the lateral position of the point cloud detected by the radar; C is the curvature of the road, which can be replaced by the turning curvature of the vehicle; and L is the arc length of the road, which can be approximated by the longitudinal distance of the point cloud.

[0165] Divide the distance within the preset lateral distance ±W on both sides of the radar into N distance bins with a preset step width of m. Then, the stationary point cloud with a lateral position of PosY will fall into the following interval according to Formula 6:

[0166] =ceil((PosY+W) / m); (6)

[0167] Here, ceil represents rounding up. Let the x-th distance bin be N;

[0168] Find the intermediate range-resolved cell bins where the peak number of stationary point clouds is located in the left and right range bins of the radar, respectively:

[0169] Left side: maxIdx_Left=max(bin[1~N / 2]);

[0170] Right side: maxIdx_right=max(bin[N / 2+1~N]);

[0171] Taking the left side of the radar as an example, the position regions represented by Bin[maxIdx_Left-1], Bin[maxIdx_Left], and Bin[maxIdx_Left+1] on the left are divided into K sub-range resolution units, namely range bin2, and the number of point clouds falling into range bin2 is counted again. The same operation is performed on the right side. The measurement widths of the single-sided tunnel boundary of the left and right tunnels in the current frame are determined based on the bin where the peak number of stationary points in the left and right range bin2 is located.

[0172] The total tunnel width is filtered according to the method described above to obtain the filtered total tunnel width of the current frame. Then, the left and right tunnel boundary widths of the filtered current frame are determined according to the method described above. Next, based on the method described above, the tunnel width confidence scores corresponding to the left and right tunnel boundary widths of the current frame are calculated, and the theoretical lateral positions corresponding to the left and right tunnel boundary widths of the current frame are calculated according to the lateral position calculation method described above.

[0173] Based on this method, the tunnel width confidence level corresponding to the left and right tunnel boundary widths in each data frame acquired by the radar can be determined, and the theoretical lateral position of the tunnel at different distances can be calculated according to the above-mentioned lateral position calculation method. False targets can be deleted based on the theoretical lateral position of the tunnel and the tunnel width confidence level.

[0174] In one specific embodiment, the tunnel boundary width identification method includes the following steps:

[0175] Step 1: Calculate the tunnel curvature and compensate for the lateral position of the point cloud in the curved tunnel so that the point cloud distribution is vertical.

[0176] The turning radius (turnR) of the vehicle is used to roughly estimate the curvature radius (R) of the tunnel, and the calculation formula is shown in formula (1). In this embodiment, the maximum longitudinal distance for tunnel width recognition is set to 130 meters, and the width of each coarse search distance bin is 1 meter. Based on the method for determining the preset curvature radius threshold, the preset curvature radius threshold is set to 2625m. If the tunnel curvature radius (R) is less than the threshold, the entered tunnel is considered to be a curved tunnel.

[0177] If the tunnel is curved, further compensation of the stationary point cloud is achieved using formula (5), making the distribution profile of the stationary point cloud vertical. In formula (5), the tunnel curvature C is used... The calculated tunnel arc length is replaced by the radial distance Range of the point cloud. Therefore, formula (5) is transformed into... .

[0178] Specifically, considering that the turning radius of the vehicle cannot perfectly and accurately represent the curvature radius of the road (i.e., turnR cannot accurately represent R), the following method is used to optimize the statistical performance. Take R = (0.8~1.2) times turnR, and substitute it into... The horizontal position of the stationary point cloud is compensated, and then the number of peak point clouds in each distance bin under different coefficients is counted. The coefficient corresponding to the highest number of peak point clouds is taken as the final coefficient.

[0179] Step 2: Calculate the lateral position of the stationary point cloud of the tunnel and output the predicted width of the single-sided tunnel boundary of a single frame;

[0180] 1) Focus on targets within ±15 meters on either side of the radar. Divide the distance within ±15 meters on either side of the radar into 30 range bins with a step width of 1 meter. Then, a stationary point with a lateral position of PosY will be determined according to the formula... =ceil(PosY+15) determines the distance resolution cell where each element falls;

[0181] 2) Find the intermediate distance resolution unit bin where the peak number of stationary point clouds is located in the left and right distance bins respectively. The left distance bin is maxIdx_Left=max(bin[1~15]); the right distance bin is maxIdx_right=max(bin[16~30]).

[0182] 3) Divide the position widths represented by Bin[maxIdx_Left-1], Bin[maxIdx_Left], and Bin[maxIdx_Left+1] on the left into 10 distance bin2s on average, and count the number of point clouds falling into the distance bin2s again. Repeat the same operation on the right.

[0183] 4) Determine the left and right tunnel widths leftWidth and rightWidth by the bin where the peak value of the stationary point in bin2 is located in the left and right distances in step 3.

[0184] Step 3: Combine historical tunnel width information with filtering to update the tunnel width of the current frame;

[0185] Note that vehicles may change lanes while traveling in the tunnel, thus the widths of the left and right sides of the tunnel actually change. However, the total tunnel width generally remains constant, so a filter is applied to the total tunnel width. In this example, α is set to 0.5, and the formula described above is modified as follows: = +0.5 (abs(leftWidth)+abs(rightWidth)- Furthermore, the widths of the left and right tunnels were obtained respectively using the methods described above. , .

[0186] Step 4: Calculate the confidence level for tunnel width;

[0187] Left and right confidence levels are determined by the above. The calculation is performed using the formula. Where β is taken as 0.5. Take 5. Take 10, Take 5. Take 10, The value can be taken as 0.8, 0.9, or 1 depending on the vibration of the total tunnel width. This leads to the following calculation formula:

[0188] ;

[0189] Step 5: Calculation of the theoretical lateral position of the tunnel wall at different distances;

[0190] The theoretical lateral position of the tunnel wall at different distances is calculated based on the tunnel curvature C and the width of the tunnel wall (i.e., the filtered width of the tunnel boundary on one side), thereby removing false targets outside the tunnel wall.

[0191] In the aforementioned tunnel boundary width identification method, the lateral distance on both sides of the radar is divided into multiple distance bins. Then, stationary point clouds are assigned to the corresponding distance bins based on their lateral positions. Finally, the tunnel wall width is determined based on the number of point clouds in each distance bin. For roads with a radius of curvature smaller than a preset radius of curvature threshold, the lateral position of the stationary point clouds is compensated until the point cloud distribution is vertical before statistical analysis to ensure the accuracy of width identification. Based on this, the final output tunnel width is filtered to obtain the filtered single-sided tunnel width. Furthermore, the confidence level of the filtered tunnel width is calculated based on information such as the proportion of peak point clouds in the distance bins, tunnel width fluctuations, and the number of point clouds, thereby improving the accuracy of tunnel width identification.

[0192] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0193] Based on the same inventive concept, this application also provides a road boundary width recognition device for implementing the road boundary width recognition method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more road boundary width recognition device embodiments provided below can be found in the limitations of the road boundary width recognition method described above, and will not be repeated here.

[0194] In one exemplary embodiment, such as Figure 7 As shown, a road boundary width recognition device is provided, including: a data determination module 702, a position determination module 704, a road boundary width determination module 706, a filtering module 708, and a false target deletion module 710, wherein:

[0195] The data determination module 702 is used to determine the road curvature of the vehicle's driving path and to acquire point cloud information collected by the vehicle's radar. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud.

[0196] The position determination module 704 is used to divide all stationary point clouds into positions based on multiple range resolution units preset in multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, and to determine the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units.

[0197] The road boundary width determination module 706 is used to determine the single-sided road boundary measurement width of the current frame in the target detection direction based on the distance value corresponding to the target distance resolution unit for each target detection direction.

[0198] The filtering module 708 is used to obtain historical road width information and filter the measured width of the single-sided road boundary of the current frame according to the historical road width information to obtain the filtered single-sided road boundary width of the current frame.

[0199] The false target deletion module 710 is used to detect the first stationary point cloud distributed in the target detection direction based on the width of the single-sided road boundary, the road curvature and the distance of the current frame after filtering, and delete false targets in the first stationary point cloud.

[0200] The aforementioned road boundary width recognition device acquires all stationary point clouds collected by the vehicle radar. Based on multiple preset range resolution units in multiple target detection directions of the vehicle radar, it divides the positions of all collected stationary point clouds, determines the concentrated distribution area of ​​stationary point clouds in each target detection direction, determines the target range resolution unit based on the number of stationary point clouds in the concentrated distribution area of ​​stationary point clouds in each target detection direction, and then determines the single-sided road boundary measurement width of the current frame in each target detection direction based on the distance value corresponding to the target range resolution unit. The single-sided road boundary measurement width of the current frame is filtered to ensure the accuracy of the single-sided road boundary measurement width of the current frame. On this basis, based on the filtered single-sided road boundary width, road curvature, and distance of the current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets outside the road are deleted, reducing the false alarm rate of radar detection and thus improving the radar detection performance. In this method, the road boundary width in each target detection direction is determined based on all collected stationary point clouds, which weakens the influence of false point clouds on the road boundary width recognition result. Moreover, this method is not limited by environmental influences.

[0201] In an exemplary embodiment, the road boundary width recognition device further includes a compensation module, which is used to obtain the turning radius of the vehicle and determine the curvature radius of the road based on the turning radius; if the curvature radius is less than a preset curvature radius threshold, the lateral position of all stationary point clouds is compensated so that the distribution of the stationary point clouds reaches a vertical state, thereby obtaining the compensated lateral position of the stationary point clouds.

[0202] In an exemplary embodiment, the road boundary width recognition device further includes a preset curvature radius threshold determination module. The preset curvature radius threshold determination module is used to determine threshold constraints. The threshold constraints are used to ensure that, without lateral position compensation, at least half of the stationary point clouds collected by the vehicle radar belong to the same distance resolution unit in the lateral position. The preset curvature radius threshold is determined based on the maximum longitudinal distance that the vehicle can recognize based on the road width, the preset width of the distance resolution unit, and the threshold constraints.

[0203] In an exemplary embodiment, the position determination module 704 is used to divide all stationary point clouds into positions based on multiple preset range resolution units and the lateral position of each stationary point cloud in multiple target detection directions of the vehicle radar, and to determine the intermediate range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units.

[0204] For each target detection direction, multiple candidate range resolution units, including the intermediate range resolution unit, are determined from multiple range resolution units based on their respective intermediate range resolution units.

[0205] Multiple candidate range resolution units are divided into regions to obtain K sub-range resolution units. The target range resolution unit where the peak of the point cloud number is located is determined from the K sub-range resolution units.

[0206] In an exemplary embodiment, the road boundary width determination module 706 further acquires historical road width information; obtains the filtered total road width of the current frame based on the total road width of the previous frame, preset weights, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions in the historical road width information; for each target detection direction, determines the width change based on the road boundary width of the previous frame in the target detection direction, the filtered total road width of the current frame, the total road width of the previous frame, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions in the historical road width information; and obtains the filtered single-sided road boundary width of the current frame based on the road boundary width of the previous frame and the width change.

[0207] The false target removal module 710 is used to determine the theoretical lateral position corresponding to the width of the single-sided road boundary of the current frame based on the width of the single-sided road boundary, the road curvature, and the distance of the current frame after filtering.

[0208] Obtain the lateral position of the first stationary point cloud distributed along the target detection direction. Based on the theoretical lateral position and the lateral position of the first stationary point cloud, determine the false targets in the first stationary point cloud and delete the false targets in the first stationary point cloud.

[0209] The confidence calculation module uses jitter parameters, point cloud quantity threshold, and ratio threshold to obtain the total road width; for each target detection direction, it determines the first stationary point cloud distributed in the target detection direction, the first stationary point cloud quantity, and the ratio of the second stationary point cloud quantity in the target distance resolution unit to the total number of stationary point clouds. Based on the jitter parameters, point cloud quantity threshold, ratio threshold, first stationary point cloud quantity, and ratio, the road width confidence is determined.

[0210] In one exemplary embodiment, the false target deletion module 710 is used to delete false targets if the road width confidence level is greater than a preset confidence level.

[0211] Each module in the aforementioned road boundary width recognition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0212] In one exemplary embodiment, a vehicle is provided, which may be a terminal equipped with millimeter-wave radar. The internal structure diagram of the vehicle can be as follows: Figure 8 As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a road boundary width recognition method. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0213] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0214] In one embodiment, a radar is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0215] In one embodiment, a vehicle is also provided, including a memory and a processor, the memory storing a computer program, which the processor executes to implement the steps in the above method embodiments.

[0216] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0217] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0218] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0219] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0220] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0221] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying road boundary width, characterized in that, The method includes: The road curvature of the vehicle's driving path is determined and the point cloud information collected by the vehicle's radar is acquired. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud. Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all stationary point clouds is divided, and the target range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units. For each target detection direction, the single-sided road boundary measurement width of the current frame in the corresponding target detection direction is determined according to the distance value corresponding to the target distance resolution unit; Obtain historical road width information, and filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame. Based on the width of the single-sided road boundary, the road curvature, and the distance of the current frame, the first stationary point cloud distributed in the target detection direction is detected, and false targets in the first stationary point cloud are deleted.

2. The method according to claim 1, characterized in that, Before dividing all the stationary point clouds into positions based on the multiple range resolution units preset in the target detection direction of the vehicle radar and the lateral positions of each stationary point cloud, the method further includes: Obtain the turning radius of the vehicle, and determine the radius of curvature of the road based on the turning radius; If the radius of curvature is less than a preset radius of curvature threshold, the lateral position of all the stationary point clouds is compensated so that the distribution of the stationary point clouds reaches a vertical state, and the compensated lateral position of the stationary point clouds is obtained.

3. The method according to claim 2, characterized in that, The method for determining the preset radius of curvature threshold includes: A threshold constraint condition is determined, which is used to constrain that, without lateral position compensation, at least half of the stationary point clouds collected by the vehicle radar belong to the same range resolution unit in terms of lateral position. The preset radius of curvature threshold is determined based on the maximum longitudinal distance that the vehicle can reach based on the road width, the preset width of the distance resolution unit, and the threshold constraint conditions.

4. The method according to claim 1, characterized in that, The step of dividing all stationary point clouds into positions based on multiple preset range resolution units in multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, and determining the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units, includes: Based on the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, the position of all the stationary point clouds is divided, and the intermediate range resolution unit where the peak number of point clouds in each target detection direction is located is determined from the multiple range resolution units. For each of the target detection directions, multiple candidate distance resolution units, including the intermediate distance resolution unit, are determined from the multiple distance resolution units according to their respective intermediate distance resolution units. The candidate distance resolution units are divided into regions to obtain K sub-distance resolution units. The target distance resolution unit where the peak of the point cloud number is located is determined from the K sub-distance resolution units.

5. The method according to claim 4, characterized in that, The step of obtaining historical road width information and filtering the measured width of the single-sided road boundary of the corresponding current frame based on the historical road width information to obtain the filtered single-sided road boundary width of the current frame includes: Obtain historical road width information; Based on the total road width of the previous frame in the road width history information, the preset weight, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions, the filtered total road width of the current frame is obtained. For each of the target detection directions, the width change is determined based on the road boundary width of the previous frame in the target detection direction in the road width history information, the filtered total road width of the current frame, the total road width of the previous frame, and the measured width of the single-sided road boundary of the current frame in multiple target detection directions; The filtered single-sided road boundary width of the current frame is obtained based on the road boundary width of the previous frame and the width change.

6. The method according to claim 1, characterized in that, The step of detecting the first stationary point cloud distributed along the target detection direction based on the single-sided road boundary width, road curvature, and distance of the filtered current frame, and deleting false targets in the first stationary point cloud, includes: Based on the width of the road boundary on one side of the filtered current frame, the road curvature, and the distance, determine the theoretical lateral position corresponding to the width of the road boundary on one side of the filtered current frame. Obtain the lateral position of the first stationary point cloud distributed along the target detection direction. Based on the theoretical lateral position and the lateral position of the first stationary point cloud, determine the false targets in the first stationary point cloud and delete the false targets in the first stationary point cloud.

7. The method according to any one of claims 1 to 6, characterized in that, Before deleting false targets from the first static point cloud, the method further includes: Obtain the jitter parameters of the total road width, the point cloud number threshold, and the ratio threshold; For each target detection direction, determine the number of first stationary point clouds distributed in the first stationary point cloud in the target detection direction, and determine the ratio of the number of second stationary point clouds in the target range resolution unit to the total number of stationary point clouds; The road width confidence level is determined based on the jitter parameter, the point cloud quantity threshold, the ratio threshold, the first stationary point cloud quantity, and the ratio. If the confidence level of the road width is greater than the preset confidence level, then the false target is deleted.

8. A road boundary width recognition device, characterized in that, The device includes: The data determination module is used to determine the road curvature of the vehicle's driving path and acquire point cloud information collected by the vehicle's radar. The point cloud information includes all stationary point clouds and the lateral position of each stationary point cloud. The position determination module is used to divide the position of all the stationary point clouds according to the multiple range resolution units preset in the multiple target detection directions of the vehicle radar and the lateral position of each stationary point cloud, and determine the target range resolution unit where the peak number of point clouds in each target detection direction is located from the multiple range resolution units. The road boundary width determination module is used to determine the single-sided road boundary measurement width of the current frame in the target detection direction according to the distance value corresponding to the target distance resolution unit for each target detection direction; The filtering module is used to acquire historical road width information, and to filter the measured width of the single-sided road boundary of the current frame corresponding to the historical road width information to obtain the filtered single-sided road boundary width of the current frame. The false target deletion module is used to detect the first stationary point cloud distributed in the target detection direction based on the width of the single-sided road boundary, the road curvature, and the distance of the filtered current frame, and to delete false targets in the first stationary point cloud.

9. A radar comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images