Deceleration strip detection method, device and equipment, medium and vehicle

Abstract

The invention discloses a deceleration strip detection method, device and equipment, a medium and a vehicle, and the method comprises the steps: obtaining point cloud data, and extracting a ground point cloud from the point cloud data; the ground point cloud is mapped to a preset ray model to obtain a point cloud sequence in the ray direction, and the ray model is constructed based on scanning characteristics of the laser radar; determining ground convex points according to included angle characteristics between adjacent points in the point cloud sequence; according to the characteristics of the deceleration strip, deceleration strip point cloud is determined from the ground protruding points, and a detection result of the deceleration strip is obtained according to the deceleration strip point cloud. According to the invention, the accuracy and robustness of deceleration strip detection can be improved.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, specifically to a speed bump detection method, device, equipment, medium, and vehicle. Background Technology

[0002] Speed ​​bumps, as common traffic signs, are typically placed in parking areas and on road sections where speed reduction is required. Effective and accurate detection of speed bumps allows autonomous vehicles to receive early warnings and slow down, improving the comfort and safety of vehicle passage.

[0003] Currently, the main methods for detecting speed bumps include those based on image recognition, those based on the vehicle's own wheel state information, and those based on lidar point cloud features.

[0004] Among these methods, image recognition-based detection can effectively classify speed bump targets, but its recognition performance is easily affected by lighting conditions, making it difficult to meet the application requirements of autonomous driving in all scenarios. Detection methods based on the vehicle's own wheel state information calculate speed bumps by measuring the state changes of the wheels after passing the target and combining this with the travel distance. This method has relatively high accuracy, but it can only obtain information after the vehicle has passed the speed bump and cannot achieve prior perception, therefore it is not suitable for the chassis pre-aiming scenarios required by autonomous driving systems. Detection methods based on LiDAR point clouds typically identify speed bumps by processing the point cloud into a grid and combining elevation features and reflection intensity information. However, since speed bumps are usually long and narrow, the constraint relationships between different grids are not considered during the gridding process, which can easily lead to missed detections.

[0005] In summary, current speed bump detection methods still have shortcomings in terms of accuracy and robustness. Summary of the Invention

[0006] This application provides a method, apparatus, equipment, medium, and vehicle for detecting speed bumps, so as to improve the accuracy and robustness of speed bump detection.

[0007] Firstly, this application provides a method for detecting speed bumps, the method comprising:

[0008] Acquire point cloud data and extract ground point cloud from the point cloud data;

[0009] The ground point cloud is mapped onto a preset ray model to obtain a point cloud sequence along the ray direction. The ray model is constructed based on the scanning characteristics of the lidar.

[0010] Determine the ground convexity point based on the angle characteristics between adjacent points in the point cloud sequence;

[0011] Based on the characteristics of the speed bump, the speed bump point cloud is determined from the ground protrusion point, and the speed bump detection results are obtained based on the speed bump point cloud.

[0012] In some embodiments of this application, the ray model includes multiple rays, and mapping the ground point cloud onto the preset ray model to obtain a point cloud sequence along the ray direction includes:

[0013] Multiple scatter points are obtained from each ray;

[0014] Based on the distance relationship between the scattered points and the ground point cloud, the point cloud sequence corresponding to the ray is determined from the ground point cloud.

[0015] In some embodiments of this application, determining the ground convexity point based on the angle features between adjacent points in a point cloud sequence includes:

[0016] Obtain the angles between the midpoint of three adjacent points in a point cloud sequence and the other two points;

[0017] If both included angles satisfy the angle threshold, the midpoint of the three points is selected as the candidate convex point.

[0018] The ground convex point is determined from the candidate convex point based on the height information of the three points.

[0019] In some embodiments of this application, obtaining the angles between the midpoint of three adjacent points in a point cloud sequence and the other two points includes:

[0020] Along the ray direction, select the first, second, and third adjacent points from the point cloud sequence in sequence;

[0021] Calculate the angle between the first point and the second point, and the angle between the second point and the third point, respectively, to obtain the first angle and the second angle, with the second point being the midpoint;

[0022] If two adjacent included angles both satisfy the angle threshold, the midpoint of the three points is selected as a candidate convex point, including:

[0023] If both the first and second included angles satisfy the angle threshold, the second point is selected as a candidate convex starting point.

[0024] Based on the height information of the three corresponding points, the ground convex point is determined from the candidate convex point, including:

[0025] If the height of the second point is greater than the height of the first point and the height of the third point, the corresponding candidate convex point is determined as the ground convex point.

[0026] In some embodiments of this application, the ray model includes multiple rays, and the ray model is constructed in the following manner:

[0027] Within the field of view of the lidar, multiple rays originating from the lidar are obtained.

[0028] In some embodiments of this application, extracting ground point clouds from point cloud data includes:

[0029] Point cloud data is preprocessed to obtain the region of interest point cloud, which includes road surface information;

[0030] Based on the preset ground plane equation and thickness threshold, ground point cloud is extracted from the point cloud of the region of interest. The ground point cloud includes speed bump information.

[0031] Secondly, this application also provides a speed bump detection device, the device comprising:

[0032] The point cloud acquisition module is used to acquire point cloud data and extract ground point cloud from the point cloud data;

[0033] The sequence acquisition module is used to map the ground point cloud onto a preset ray model to obtain a point cloud sequence along the ray direction. The ray model is constructed based on the scanning characteristics of the lidar.

[0034] The convex point acquisition module is used to determine the ground convex point based on the angle features between adjacent points in the point cloud sequence.

[0035] The detection and acquisition module is used to determine the speed bump point cloud from the ground protrusion points according to the characteristics of the speed bump, and to obtain the detection result of the speed bump based on the speed bump point cloud.

[0036] Thirdly, this application also provides an electronic device, including a memory and a processor; the memory stores a computer program, and the processor is used to run the computer program in the memory to perform the operations in the speed bump detection method provided in the first aspect.

[0037] Fourthly, this application also provides a storage medium storing a plurality of instructions adapted for loading by a processor to execute the steps in the speed bump detection method provided in the first aspect.

[0038] Fifthly, this application also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps in the speed bump detection method provided in the first aspect.

[0039] Sixthly, this application also provides a vehicle equipped with at least one of the following: the speed bump detection device provided in the second aspect, the electronic device provided in the third aspect, the storage medium provided in the fourth aspect, and the computer program product provided in the fifth aspect.

[0040] Through one or more embodiments of the above embodiments in this application, at least the following technical effects can be achieved:

[0041] In the speed bump detection method provided in this application, the ground point cloud is mapped onto a preset ray model to obtain a point cloud sequence along the ray direction, thereby achieving an ordered arrangement of disordered point cloud data, facilitating subsequent angular feature analysis based on spatial positional relationships. Ground protrusion points are determined based on the angular features between adjacent points in the point cloud sequence, thus achieving precise positioning of road surface height changes, effectively identifying potential speed bumps or other protruding structures, while avoiding interference from points on smooth road surfaces. Based on the characteristics of the speed bump, the speed bump point cloud is determined from the ground protrusion points, thereby distinguishing speed bumps from other protrusions and ensuring that only the set of protrusion points that meet the characteristics of height, width, length, and morphological continuity is retained, improving the accuracy of speed bump identification. The detection result of the speed bump is obtained from the speed bump point cloud. Therefore, the speed bump detection method provided in this application, through ray mapping and angular feature analysis, fully preserves the three-dimensional details of the point cloud, and combines ground segmentation to pre-emptively eliminate non-ground obstacle point clouds, thereby improving the accuracy and robustness of speed bump detection. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 This is one of the flowcharts illustrating a speed bump detection method provided in this application embodiment;

[0044] Figure 2 This is a second schematic flowchart of a speed bump detection method provided in the embodiments of this application;

[0045] Figure 3 This is a schematic diagram of the speed bump detection device provided in the embodiments of this application;

[0046] Figure 4 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0047] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0048] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Therefore, features defined as "first" or "second" may explicitly or implicitly include one or more features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0049] "A and / or B" includes the following three combinations: A only, B only, and a combination of A and B.

[0050] The use of "applies to" or "configured to" in this application implies open and inclusive language, which does not preclude applicability to or configuration to devices performing additional tasks or steps. Furthermore, the use of "based on" implies openness and inclusivity, because processes, steps, calculations, or other actions "based on" one or more conditions or values ​​may in practice be based on additional conditions or values ​​beyond those conditions.

[0051] In this application, the term "exemplary" is used to mean "used as an example, illustration, or description." Any embodiment described as "exemplary" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use this application. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that this application can be made without using these specific details. In other instances, well-known structures and processes are not described in detail to avoid obscuring the description of this application with unnecessary detail. Therefore, this application is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0052] Currently, speed bump detection methods mainly include those based on image recognition, those based on wheel state information, and those based on lidar point cloud features.

[0053] Image recognition-based speed bump detection methods typically rely on pre-trained models. This involves collecting and labeling a large amount of speed bump data, then training a classification model using a neural network to extract speed bump features from the image and perform identification. While neural network-based speed bump detection methods achieve high accuracy in well-lit environments, they generally depend on depth estimation algorithms to calculate target distance, resulting in inaccurate location information. Furthermore, in the full-scenario applications of autonomous driving, visual recognition accuracy drops significantly at night or in low-light conditions, especially for black targets like speed bumps, making reliable perception impossible.

[0054] Speed ​​bump detection methods based on wheel state information compare the vehicle's driving state before and after passing an obstacle with a preset threshold to determine whether the vehicle has passed a speed bump. While this method can reflect the presence of a speed bump, the detection occurs after the vehicle has passed the target, making it unsuitable for autonomous driving systems that require anticipating road conditions ahead.

[0055] Speed ​​bump detection methods based on lidar point cloud features typically rely on point cloud rasterization, combining height constraint features and reflection intensity information within the raster for comprehensive judgment. However, speed bumps often exhibit elongated, strip-like structures, and these methods fail to consider the constraint relationships between different raster cells, easily leading to missed detections. Furthermore, the height constraint information relied upon by this method is susceptible to interference from other road obstacles, while the reflection intensity feature depends on the presence of high-reflectivity stripes on the speed bump surface. In the absence of obvious reflection features, it is difficult to accurately identify speed bump targets in lidar scan results.

[0056] In summary, current speed bump detection methods still have shortcomings in terms of accuracy and applicability to all scenarios of autonomous driving.

[0057] To address the aforementioned problems, embodiments of this application provide a method, apparatus, equipment, medium, and vehicle for detecting speed bumps. The following description, in conjunction with the accompanying drawings, details the method, apparatus, equipment, medium, and vehicle provided in this application for detecting speed bumps.

[0058] like Figure 1 As shown in the figure, this application provides a method for detecting speed bumps, which includes the following steps:

[0059] S101, acquire point cloud data, and extract ground point cloud from the point cloud data.

[0060] Schematic illustration: A LiDAR scanner scans a target detection area to obtain point cloud data. This detection area can be a scene such as an urban road, a residential road, or a parking lot driveway. During operation, the LiDAR emits a laser beam and receives its reflected signal, enabling it to capture the three-dimensional spatial information of objects within the scanned area with high precision, thus generating point cloud data covering that area. This point cloud data includes not only ground information but also information about objects above ground, such as trees, vehicles, and buildings. Since speed bumps are typically located on the ground, their three-dimensional contour features naturally blend into the collected ground point cloud data; therefore, the presence of speed bumps can be identified by analyzing the ground point cloud.

[0061] After acquiring complete point cloud data, it is necessary to preprocess the point cloud to extract the ground point cloud. Illustratively, a ground segmentation algorithm is used to extract the ground point cloud from the point cloud data, effectively removing non-ground points such as roadside trees, buildings on both sides of the road, parked or moving vehicles, etc., retaining only the point cloud set reflecting the ground height and shape.

[0062] S102, map the ground point cloud onto a preset ray model to obtain a point cloud sequence along the ray direction.

[0063] The ray model is constructed based on the scanning characteristics of lidar. Understandably, lidar typically emits laser beams at fixed angular intervals during operation, exhibiting a clear regularity in its scanning pattern: the lidar scans horizontally at a set horizontal angular resolution and vertically at preset vertical angular intervals, progressively scanning the field of view. In this way, lidar can cover the target detection area with a certain spatial resolution and generate three-dimensional point cloud data. Since each point cloud originates from a laser ray originating from the lidar's emission center and extending along specific horizontal and vertical angular directions, the point cloud data naturally possesses "ray correlation."

[0064] Based on this, mapping the ground point cloud to a pre-defined ray model can fully utilize the aforementioned "ray correlation" to achieve ordered processing of the ground point cloud. By introducing the ray model, each ground point cloud is assigned to its corresponding ray, thereby transforming the disordered point cloud set into a set of point clouds on several rays.

[0065] S103, determine the ground convexity point based on the angle characteristics between adjacent points in the point cloud sequence.

[0066] The included angle between adjacent points refers to the angle formed by the straight line connecting any two adjacent points in the point cloud sequence and the horizontal plane. Angle characteristics refer to the pattern exhibited by the included angles between adjacent points within the same point cloud sequence as the sequence changes. The size of the included angle directly reflects the slope change of the ground along the ray direction: when the road surface is smooth, the included angle between adjacent points is usually small and changes smoothly throughout the sequence. When there is a significant convexity on the road surface, the height difference between adjacent points suddenly increases, causing the corresponding included angle to rise significantly in a local location, thus exhibiting a convexity characteristic. Conversely, when the road surface is sunken or has potholes, the height difference between adjacent points abruptly changes in the opposite direction, causing the corresponding included angle to decrease significantly in a local location, thus exhibiting a concave characteristic. Based on the aforementioned convexity characteristics, ground convexity points can be identified from the point cloud sequence.

[0067] S104, determine the speed bump point cloud from the ground protrusion points according to the characteristics of the speed bump, and obtain the speed bump detection results based on the speed bump point cloud.

[0068] Understandably, the three-dimensional geometry of a speed bump is characterized as a "continuous protrusion extending laterally along the road with a fixed width and height." Based on the geometric characteristics of this speed bump, the speed bump point cloud is determined from the protrusion points on the ground, thereby distinguishing it from scattered and isolated protrusion points caused by random obstacles such as stones and gravel, thus improving the detection accuracy of speed bumps.

[0069] After obtaining the speed bump point cloud, the position and size information of the speed bump are determined based on the speed bump point cloud, and the position and size information are output as the detection results.

[0070] The speed bump detection method provided in this application maps ground point clouds onto a preset ray model to obtain a point cloud sequence along the ray direction, thereby achieving an ordered arrangement of disordered point cloud data, facilitating subsequent angular feature analysis based on spatial positional relationships. Ground protrusion points are determined based on the angular features between adjacent points in the point cloud sequence, thus achieving precise positioning of road surface height changes, effectively identifying potential speed bumps or other protruding structures, while avoiding interference from points on smooth road surfaces. Speed ​​bump point clouds are determined from ground protrusion points based on the characteristics of speed bumps, thereby distinguishing speed bumps from other protrusions and ensuring that only a set of protrusion points meeting characteristics such as height, width, length, and morphological continuity is retained, improving the accuracy of speed bump identification. The detection results of the speed bumps are obtained from the speed bump point clouds. Therefore, the speed bump detection method provided in this application, through ray mapping and angular feature analysis, fully preserves the three-dimensional details of the point cloud, and combines ground segmentation to pre-emptively eliminate non-ground obstacle point clouds, thereby improving the accuracy and robustness of speed bump detection.

[0071] It should be noted that, since the speed bump detection method provided in this application embodiment is directly based on point cloud data and does not rely on the texture or color information of visual images, it can be used stably in environments such as strong light, heavy rain, dense fog, nighttime, or tunnels. Furthermore, as a "pre-perception" technology, it can provide sufficient reaction time for the autonomous driving system, aiding in speed control, braking, or steering operations and improving driving safety. In addition, the speed bump detection method provided in this application embodiment does not require point cloud rasterization, avoiding the loss of point cloud details and interference from surrounding obstacles caused by rasterization.

[0072] In some embodiments of this application, the ray model includes multiple rays, and mapping the ground point cloud onto the preset ray model to obtain a point cloud sequence along the ray direction includes:

[0073] Multiple scatter points are obtained from each ray.

[0074] Based on the distance relationship between the scattered points and the ground point cloud, the point cloud sequence corresponding to the ray is determined from the ground point cloud.

[0075] Schematic, with a ray L i Taking the lidar emission center as an example, multiple scattered points are generated outward along the ray direction with a certain step size, forming a scattered point array R for that ray. The farthest point of the scattered point array can be set according to the actual scene requirements, such as 40m. The step size can be fixed, such as 0.2m, or it can be adaptively adjusted according to scene requirements or point cloud density. In some examples, for sparse point clouds, a smaller step size can be selected to increase the number of sampling points and improve the resolution of ground elevation features; while for dense point clouds, a larger step size can be selected to reduce redundant points and improve computational efficiency. By reasonably setting the step size and point selection parameters, efficient ray point cloud generation can be achieved while ensuring the integrity of ground elevation features.

[0076] Iterate through each scatter point p in array R i (x i ,y i ,z i Using nearest neighbor algorithms (such as k-nearest neighbor search or sphere neighborhood search) in the ground point cloud P ground Find the nearest actual point cloud to the current scatter point, and update the coordinates of that actual point to the location of the scatter point, thus mapping the discrete scatter points to the corresponding actual point cloud p. i Location. By traversing all points in the scatter array R, an ordered point cloud sequence P along the ray direction is obtained. L .

[0077] In some embodiments of this application, determining the ground convexity point based on the angle features between adjacent points in a point cloud sequence includes:

[0078] Obtain the angle between the middle point of three adjacent points in a point cloud sequence and the other two points.

[0079] If both included angles satisfy the angle threshold, the midpoint of the three points is selected as the candidate convex point.

[0080] The ground convex point is determined from the candidate convex point based on the height information of the three points.

[0081] In some examples, along the ray direction, the first adjacent point p is selected sequentially from the point cloud sequence. i Second point p j And the third point p k .

[0082] Calculate the first point p respectively i With the second point p j The angle between them, the second point p j With the third point p kThe angle between them corresponds to the first angle θ. ij The second included angle θ jk The second point is the midpoint. (Illustratively) Among them, (x i ,y i ,z i (p) is the first point i Location information, (x j ,y j ,z j ) is the second point p j Location information, (x k ,y k ,z k (p) is the third point k Location information.

[0083] First included angle θ ij The second included angle θ jk If all angles meet the angle threshold, the second point is selected as a candidate convex starting point. It is understandable that when the included angles of adjacent points all meet the angle threshold, it indicates a significant height variation along the ray direction, potentially corresponding to a convexity or depression in the ground plane.

[0084] Furthermore, at the second point p j height z j Greater than the first point p i height z i With the third point p k height z k In the case of z, the corresponding candidate convex point is determined as the ground convex point. That is, when z j Greater than z i and z k At that time, determine the second point p j Located in a locally raised area, it distinguishes between a protrusion and a depression.

[0085] By sequentially traversing all rays and performing the aforementioned angle threshold judgment and height comparison on the point cloud sequence corresponding to each ray, all ground plane convex points on the entire ray can be obtained.

[0086] It should be noted that the spacing between adjacent scanning points of the lidar on the ground is determined by the installation height of the lidar on the vehicle and the angle between adjacent rays. Since this spacing is limited by the installation height and the angle between adjacent rays, and also by the width of the speed bump, in reality, three adjacent points on the same ray will almost never hit the same speed bump at the same time, resulting in a difference in the vertical angle between the three adjacent points.

[0087] In other embodiments of this application, ground protrusion points can also be determined based on the height information z of each point in the point cloud sequence. Indicatively, points where z exceeds a preset threshold are identified as protrusion points. The preset threshold can be dynamically calculated based on the average height of the road reference surface, smoothness parameters, or the height difference between adjacent points, thereby achieving effective identification of abnormal ground protrusions.

[0088] The speed bump detection method provided in this application uses an angle threshold to filter candidate convex points in areas of significant geometric change, effectively avoiding misidentification of smooth road surfaces or normal slope changes as convex or concave points. Furthermore, it combines height difference relationships to discriminate candidate points, identifying those significantly higher than adjacent points as convex points and those lower as concave points. The speed bump detection method provided in this application not only improves recognition accuracy but also effectively suppresses false detection rates caused by noise interference or uneven local density distribution of point clouds.

[0089] In some embodiments of this application, the ray model includes multiple rays, and the ray model is constructed in the following manner:

[0090] Within the field of view of the lidar, multiple rays originating from the lidar are obtained.

[0091] In some examples, a ray model is constructed based on the ground plane S within the field of view (FOV) of the lidar, centered on the lidar's own location. Schematic, rays are drawn sequentially to the left and right sides at fixed angular intervals, with the lidar's X-axis as the reference direction. For example, rays are evenly distributed throughout the entire FOV with an angular step size of 1°, thus forming a series of ray sets with equal angular intervals.

[0092] In some embodiments of this application, extracting ground point clouds from point cloud data includes:

[0093] For point cloud data P raw Preprocessing is performed to obtain the point cloud P of the region of interest. roi The region of interest (ROI) point cloud includes road surface information. In some examples, the obtained raw point cloud data P... raw Preprocessing such as denoising, filtering, and smoothing is performed to reduce the interference of random noise points and isolated points, improve the overall continuity and reliability of the point cloud, and obtain the region of interest point cloud, which includes road surface information, by using algorithms such as region growing, clustering segmentation, or ground segmentation.

[0094] Based on the preset ground plane equation and thickness threshold, ground point cloud is extracted from the point cloud of the region of interest. The ground point cloud includes speed bump information.

[0095] In some examples, the ground plane equation is constructed based on the Random Sample Consensus (RANSAC) algorithm. Schematably, the ground plane equation S: Ax + By + CZ + D = 0 is constructed using RANSAC, where the plane normal vector is a unit vector n1 = [A, B, C] composed of normalized coefficients, and the reference virtual ground plane normal vector is n2 = [0, 0, 1]. The angle θ between n1 and n2 is calculated. When θ is greater than a preset threshold, it indicates a low probability of the plane fitting the ground plane. In this case, a comprehensive evaluation is performed by combining ground plane parameters extracted from point clouds in adjacent frames to improve the stability and effectiveness of ground plane detection.

[0096] After obtaining a reliable ground plane equation, a thickness threshold T is set. thick Further filtering is performed on the point cloud of the region of interest. Specifically, for any point p in the point cloud of the region of interest... i (x i ,y i ,z i ), calculate its distance to the equation S of the ground plane. When the distance d i <T thick If the point is not specified, then retain that point to obtain the ground point cloud P containing the speed bump target. ground .

[0097] In other embodiments of this application, the ground plane equation can also be constructed using other plane fitting methods such as least squares fitting, PCA principal direction analysis, and region growing. It should be noted that although there will be some fitting error in constructing the ground plane equation using the RANSAC algorithm, this application assumes that a certain degree of error in the ground plane fitting result will not affect the detection of speed bumps because the speed bump target itself has certain thickness variations. On the contrary, this allowable error setting avoids the additional iterative calculations and time overhead brought about by pursuing high-precision fitting, thereby reducing computational complexity while ensuring the effectiveness and robustness of speed bump detection.

[0098] In some embodiments of this application, obtaining the detection result of the speed bump based on the speed bump point cloud includes:

[0099] Linear fitting is performed on the point cloud of the speed bump to obtain its lines. Based on these lines, the location and size of the speed bump are determined, and a bounding box is formed. All these elements are then integrated into the output speed bump detection results.

[0100] In some examples, the speed bump point cloud P is obtained using a line fitting algorithm (such as the RANSAC algorithm). S To obtain a straight line from the point cloud of the speed bump, the specific implementation is as follows: First, start from the point cloud P of the speed bump. S Two points p are randomly selected from the middle.a and p b Only point p is used a and p b The x and y information is used to generate a candidate straight line on the XOY plane of the radar coordinate system. Random point p a and p b A straight line L can be generated: Ex + Fy + G = 0, then P S Points with a distance L less than a threshold ∈ are designated as interior points, and the rest as exterior points. The proportion of interior points among all points is calculated as the score. This process is iterated multiple times to find lines with a score exceeding a certain threshold. Lines that are spatially close to or overlap each other are merged into a single target line, thus extracting the target line features corresponding to the speed bump. Further calculation of the speed bump point cloud P is then performed. O The boundaries of each axis are defined, and finally a 3D target box (3DBox) describing the speed bump area is generated.

[0101] The speed bump detection method provided in this application embodiment performs linear fitting on the speed bump point cloud, which fully utilizes the elongated geometric characteristics of the speed bump target to effectively extract the linear features of the speed bump in space. Furthermore, since each ray typically contains speed bump points, linear fitting can filter out noise and outliers, retaining only the point cloud belonging to the speed bump, thereby significantly improving the accuracy of identifying the speed bump's position, direction, and length. In addition, by spatially merging and boundary calculation of the extracted target lines, an accurate 3D target box (3DBox) can be generated, achieving a complete description of the speed bump's position, size, and spatial range.

[0102] like Figure 2 As shown in the embodiments of this application, a method for detecting speed bumps is also provided, which includes the following steps:

[0103] S201, Obtain the raw point cloud data P raw .

[0104] S202, Filter the original point cloud data to obtain the region of interest point cloud P containing the speed bump target. roi To illustrate, the raw point cloud data can be filtered using pass-through filtering along the XYZ axes.

[0105] S203, the ground plane equation S is obtained through a plane fitting algorithm, and based on this ground plane equation S and a thickness threshold, the point cloud P of the region of interest is extracted. roi Obtain the ground point cloud P containing the speed bump target. ground .

[0106] S204. Construct a ray model, which is an array of rays. The rays are drawn outwards from the LiDAR coordinates. The angle between adjacent rays can be a fixed angle, such as 1°, or it can be gradually varied according to requirements.

[0107] S205 involves taking multiple scattered points along the ray. In some examples, the origin is the lidar, and scattered points are taken along the ray within a preset length range with a certain step size. The preset length range can be a fixed value, such as 0-40m, or it can be continuously adjusted according to requirements. It should be noted that the scattered points on each ray form a scattered point array R.

[0108] S206, iterate through the scattered points in each group, and calculate the data based on the scattered points and the ground point cloud P. ground The distance between the virtual scattered points and the actual point cloud is used to establish a mapping relationship between the virtual scattered points and the actual point cloud. The position information of the scattered points is updated using the position information of the actual point cloud to obtain the ray point cloud P. L That is, a point cloud sequence along the ray direction.

[0109] S207, calculate the included angle between adjacent points in each group of ray point clouds. Illustratively, the distance R between adjacent points is calculated based on their X and Y coordinates, and then the included angle between them is calculated based on the distance R and their Z coordinates. It can be understood that each group of ray point clouds corresponds to an array of included angles.

[0110] S208: For each array of included angles, count the included angles that are greater than the angle threshold, and determine the corresponding points as ground convexity points.

[0111] S209, a straight line is found among the ground protrusions using a straight line fitting algorithm, and the speed bump point cloud P is determined based on this straight line. S .

[0112] S210, calculate the speed bump point cloud P in the XYZ direction. S The boundary points are used to determine the 3D target box (3DBox) of the speed bump.

[0113] The speed bump detection method provided in this application constructs a ray model and collects scattered points along the rays. It samples and records the elevation features of the point cloud along each ray direction, achieving fine point cloud analysis along the ray direction. By setting the ground plane equation and thickness threshold, it can quickly acquire the ground point cloud along the ray direction and identify ground protrusions through angle statistics, thereby efficiently filtering out large ground targets that are not speed bumps. A straight-line fitting algorithm is used to extract the elongated features of the speed bump from the ground protrusions, enabling accurate identification and location of the speed bump point cloud. In summary, the speed bump detection method provided in this application, through a ray-based point search strategy combined with angle features, height feature analysis, and straight-line fitting, not only reduces computational load and improves real-time performance but also fully utilizes the elongated geometric characteristics of speed bumps, achieving high-precision and highly robust speed bump detection.

[0114] like Figure 3 As shown in the figure, this application embodiment also provides a speed bump detection device, the device comprising:

[0115] The point cloud acquisition module 301 is used to acquire point cloud data and extract ground point cloud from the point cloud data.

[0116] The sequence acquisition module 302 is used to map the ground point cloud onto a preset ray model to obtain a point cloud sequence along the ray direction. The ray model is constructed based on the scanning characteristics of the lidar.

[0117] The convex point acquisition module 303 is used to determine the ground convex point based on the angle features between adjacent points in the point cloud sequence.

[0118] The detection and acquisition module 304 is used to determine the speed bump point cloud based on the ground protrusion points and obtain the speed bump detection results based on the speed bump point cloud.

[0119] The speed bump detection device provided in this application corresponds to the speed bump detection method provided in any of the above embodiments, and will not be described again here.

[0120] Based on any of the above embodiments, another embodiment of this application also provides an electronic device, such as... Figure 4 As shown, the electronic device may include a processor 410, a communication interface 420, a memory 430, and a communication bus 440. The processor 410, communication interface 420, and memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions stored in the memory 430 to execute the aforementioned speed bump detection method.

[0121] Furthermore, the logical instructions in the aforementioned memory 430 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0122] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0123] On the other hand, embodiments of this application also provide a storage medium storing a plurality of instructions adapted for loading by a processor to execute the speed bump detection method provided in the above embodiments.

[0124] On the other hand, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the speed bump detection method described above.

[0125] On the other hand, this application also provides a vehicle equipped with at least one of the above-mentioned speed bump detection device, electronic device, storage medium and computer program product.

[0126] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0127] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of various embodiments or some parts of embodiments.

[0128] The above provides a detailed description of a speed bump detection method, apparatus, device, medium, and vehicle provided in the embodiments of this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for detecting speed bumps, characterized in that, The methods include: Acquire point cloud data and extract ground point cloud from the point cloud data; The ground point cloud is mapped onto a preset ray model to obtain a point cloud sequence along the ray direction. The ray model is constructed based on the scanning characteristics of the lidar. The ground convexity point is determined based on the angle characteristics between adjacent points in the point cloud sequence; Based on the characteristics of the speed bump, a speed bump point cloud is determined from the ground protrusion point, and the speed bump detection result is obtained based on the speed bump point cloud.

2. The speed bump detection method according to claim 1, characterized in that, The ray model includes multiple rays, and mapping the ground point cloud onto the preset ray model to obtain a point cloud sequence along the ray direction includes: Multiple scatter points are obtained from each of the aforementioned rays; Based on the distance relationship between the scattered points and the ground point cloud, the point cloud sequence corresponding to the ray is determined from the ground point cloud.

3. The speed bump detection method according to claim 1, characterized in that, The step of determining the ground convexity point based on the angle features between adjacent points in the point cloud sequence includes: Obtain the angles between the midpoint of three adjacent points in the point cloud sequence and the other two points. If both included angles satisfy the angle threshold, the midpoint of the three points is selected as the candidate convex point. The ground protrusion point is determined from the candidate protrusion points based on the height information of the three points.

4. The speed bump detection method according to claim 3, characterized in that, The step of obtaining the angles between the midpoint of three adjacent points in the point cloud sequence and the other two points includes: Along the ray direction, sequentially select adjacent first, second, and third points from the point cloud sequence; Calculate the angle between the first point and the second point, and the angle between the second point and the third point, respectively, to obtain the first angle and the second angle, where the second point is the midpoint; When two adjacent included angles both satisfy the angle threshold, the midpoint of the three points is selected as a candidate convex point, including: If both the first included angle and the second included angle satisfy the angle threshold, the second point is taken as the candidate convex point. The step of determining the ground convex point from the candidate convex point based on the height information of the corresponding three points includes: If the height of the second point is greater than the height of the first point and the height of the third point, the corresponding candidate protrusion point is determined as the ground protrusion point.

5. The speed bump detection method according to claim 1, characterized in that, The ray model comprises multiple rays, and is constructed as follows: Within the field of view of the lidar, multiple rays with the lidar as the origin are obtained.

6. The speed bump detection method according to any one of claims 1 to 5, characterized in that, Extracting ground point cloud from the point cloud data includes: The point cloud data is preprocessed to obtain a region of interest point cloud, which includes road surface information; Based on a preset ground plane equation and thickness threshold, the ground point cloud is extracted from the point cloud of the region of interest, and the ground point cloud includes speed bump information.

7. A speed bump detection device, characterized in that, The device includes: The point cloud acquisition module is used to acquire point cloud data and extract ground point cloud from the point cloud data; The sequence acquisition module is used to map the ground point cloud onto a preset ray model to obtain a point cloud sequence along the ray direction. The ray model is constructed based on the scanning characteristics of the lidar. A convex point acquisition module is used to determine ground convex points based on the angle features between adjacent points in the point cloud sequence. The detection and acquisition module is used to determine the speed bump point cloud from the ground protrusion point according to the characteristics of the speed bump, and to obtain the detection result of the speed bump based on the speed bump point cloud.

8. An electronic device, characterized in that, It includes a memory and a processor; the memory stores a computer program, and the processor runs the computer program in the memory to perform the steps of the speed bump detection method according to any one of claims 1 to 6.

9. A storage medium, characterized in that, The storage medium stores a plurality of instructions adapted for loading by a processor to execute the steps of the speed bump detection method according to any one of claims 1 to 6.

10. A vehicle, characterized in that, The vehicle is equipped with at least one of the speed bump detection device of claim 7, the electronic device of claim 8, and the storage medium of claim 9.

Images