Obstacle detection method and device, medium, equipment and vehicle

Abstract

The invention discloses an obstacle detection method and device, a medium, equipment and a vehicle, relates to the technical field of automatic driving, and can solve the problem that a traditional laser radar sensing scheme cannot detect static obstacles with the characteristics of being too small in size, low in occurrence frequency and the like. The method comprises the following steps: filtering laser radar point clouds outside a drivable area in original laser radar point clouds to obtain first laser radar point clouds; filtering ground point clouds in the first laser radar point clouds to obtain second laser radar point clouds; clustering the second laser radar point clouds to obtain bounding box size information of at least one suspected static obstacle; and determining a real static obstacle from at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with an obstacle geometric rule. According to the scheme, the static obstacle with the characteristics of being too small in size, too low in occurrence frequency and the like can be detected based on the laser radar.

Description

Technical Field

[0001] The present application relates to the field of autonomous driving technology, and more specifically, to an obstacle detection method, device, medium, equipment and vehicle. Background Art

[0002] LiDAR is still an indispensable sensor in the current L3 / L4 autonomous driving system. LiDAR provides us with high-resolution data by emitting thousands of laser signals. These laser signals are reflected back to the LiDAR by the object, and then the distance to the object can be determined by calculating the time required for the signal to return. Therefore, the perception algorithm based on LiDAR point cloud has become one of the focuses of autonomous driving research and development in recent years. However, static obstacles such as cones and warning signs are too small in size and have a low frequency of occurrence. In vehicle cruise assisted driving, the traditional LiDAR perception scheme mentioned above fails to detect static obstacles, resulting in no effective warnings and braking of the vehicle, which in turn brings panic and safety risks to the driver. Summary of the invention

[0003] The present application provides an obstacle detection method, device, medium, equipment and vehicle, which can detect static obstacles with characteristics such as being too small in size and too low in frequency based on laser radar.

[0004] The specific technical solutions are as follows:

[0005] In a first aspect, an embodiment of the present application provides an obstacle detection method, the method comprising:

[0006] Filtering the laser radar point cloud outside the drivable area in the original laser radar point cloud to obtain a first laser radar point cloud;

[0007] Filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud;

[0008] Clustering the second laser radar point cloud to obtain bounding box size information of at least one suspected static obstacle;

[0009] The real static obstacle is determined from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule.

[0010] It can be seen from the above scheme that compared with the traditional laser radar perception scheme that perceives obstacles through laser reflection signals, after obtaining the original laser radar point cloud, the embodiment of the present application can first filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud, and then filter the ground point cloud, thereby obtaining foreground information including only obstacle point clouds (called the second laser radar point cloud), and then clustering the second laser radar point cloud to obtain the bounding box size information of at least one suspected static obstacle, and finally by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule, determine the real static obstacle from at least one suspected static obstacle. It can be seen from this that the embodiment of the present application can detect each static obstacle contained in each frame of the original laser radar point cloud by clustering the second laser radar point cloud corresponding to the foreground information and matching the obstacle geometry rule, without being affected by factors such as size and low frequency of occurrence, thereby improving the accuracy of detecting static obstacles.

[0011] In a first possible implementation manner of the first aspect, filtering the lidar point cloud located outside the drivable area in the original lidar point cloud to obtain the first lidar point cloud includes:

[0012] Searching for a preset object located around the vehicle posture information in the high-precision map, and filtering the preset object in the original laser radar point cloud, wherein the original laser radar point cloud includes a laser radar point cloud located outside the drivable area and a laser radar point cloud located within the drivable area;

[0013] Performing plane fitting on the original laser radar point cloud after filtering the preset object to obtain a ground plane;

[0014] Determine a boundary line of the ground plane along the direction of vehicle movement, and obtain a laser radar point located on the boundary line as a road boundary candidate point;

[0015] Performing curve fitting on the candidate road boundary points to obtain a road boundary, and determining an area enclosed by the road boundary as the drivable area;

[0016] The laser radar point cloud located outside the drivable area in the original laser radar point cloud is filtered to obtain the first laser radar point cloud.

[0017] Through the above scheme, it can be known that the embodiment of the present application can first determine the road boundary based on the original laser radar point cloud, then determine the area enclosed by the road boundary as the drivable area, and finally filter the laser radar point cloud outside the drivable area in the original laser radar point cloud to obtain the first laser radar point cloud. Among them, when determining the road boundary, by first filtering the preset object in the original laser radar point cloud (the preset object is included in the high-precision map), the interference of the preset object in determining the road boundary and subsequent obstacle detection can be avoided.

[0018] In a second possible implementation manner of the first aspect, before filtering the ground point cloud in the first lidar point cloud to obtain the second lidar point cloud, the method further includes:

[0019] Based on the target detection model, dynamic obstacles in the first laser radar point cloud are detected, and the detected dynamic obstacles are filtered to obtain a third laser radar point cloud;

[0020] The filtering of the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud includes:

[0021] The ground point cloud in the third laser radar point cloud is filtered according to the laser radar ray method to obtain the second laser radar point cloud.

[0022] It can be seen from the above scheme that before filtering the ground point cloud in the first lidar point cloud to obtain the second lidar point cloud, the embodiment of the present application first detects dynamic obstacles in the first lidar point cloud based on the target detection model, and filters the detected dynamic obstacles to obtain the third lidar point cloud, thereby preventing dynamic obstacles such as vehicles and pedestrians from interfering with the subsequent detection of static obstacles.

[0023] In a third possible implementation manner of the first aspect, filtering the ground point cloud in the third lidar point cloud according to the lidar ray method to obtain the second lidar point cloud includes:

[0024] Obtaining a two-dimensional laser radar point cloud in the third laser radar point cloud;

[0025] Determine a two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to a laser radar ray method;

[0026] Filter the two-dimensional ground point cloud in the two-dimensional lidar point cloud, and superimpose the filtered two-dimensional lidar point cloud with the depth information of the filtered two-dimensional lidar point cloud recorded in the third lidar point cloud to obtain a second lidar point cloud, or superimpose the two-dimensional ground point cloud with the depth information corresponding to the two-dimensional ground point cloud in the third lidar point cloud to obtain a three-dimensional ground point cloud, and filter the three-dimensional ground point cloud in the third lidar point cloud to obtain the second lidar point cloud.

[0027] It can be seen from the above scheme that after the embodiment of the present application determines the two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to the laser radar ray method, the three-dimensional foreground information, i.e., the second laser radar point cloud, can be accurately and quickly determined by two methods. One method includes first filtering the two-dimensional ground point cloud to obtain two-dimensional foreground information, and then superimposing depth information on the two-dimensional foreground information to obtain three-dimensional foreground information. Another method includes first superimposing depth information on the two-dimensional ground point cloud to obtain a three-dimensional ground point cloud, and then filtering the three-dimensional ground point cloud to obtain three-dimensional foreground information.

[0028] In a fourth possible implementation manner of the first aspect, determining a real static obstacle from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with an obstacle geometry rule includes:

[0029] Matching the bounding box size information of each of the suspected static obstacles with the size ranges of different types of obstacles specified in the obstacle geometry rules;

[0030] When the size information of the bounding box of the suspected static obstacle matches the size range of a certain category of obstacles, determining that the suspected static obstacle is the real static obstacle;

[0031] When the bounding box size information of the suspected static obstacle does not match the size ranges of all categories of obstacles, it is determined that the suspected static obstacle is not the real static obstacle.

[0032] It can be seen from the above scheme that the embodiment of the present application accurately determines whether the suspected static obstacle is a real static obstacle by matching the bounding box size information of each suspected static obstacle with the size ranges of different categories of obstacles specified in the obstacle geometry rules, thereby improving the accuracy and efficiency of detecting real static obstacles.

[0033] In a second aspect, an embodiment of the present application provides an obstacle detection device, the device comprising:

[0034] A first filtering unit is used to filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud to obtain a first laser radar point cloud;

[0035] A second filtering unit, used for filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud;

[0036] A clustering unit, configured to cluster the second laser radar point cloud to obtain size information of a bounding box of at least one suspected static obstacle;

[0037] The matching determination unit is used to determine the real static obstacle from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule.

[0038] In a first possible implementation manner of the second aspect, the first filtering unit includes:

[0039] A search module is used to search for preset objects around the vehicle's posture information in the high-precision map;

[0040] A first filtering module, configured to filter the preset object in the original laser radar point cloud, wherein the original laser radar point cloud includes a laser radar point cloud located outside the drivable area and a laser radar point cloud located within the drivable area;

[0041] A plane fitting module, used for performing plane fitting on the original laser radar point cloud after filtering the preset object to obtain a ground plane;

[0042] A first determination module, configured to determine a boundary line of the ground plane along a moving direction of the vehicle;

[0043] A first acquisition module is used to acquire a laser radar point located on the boundary line as a road boundary candidate point;

[0044] A curve fitting module, used for performing curve fitting on the candidate road boundary points to obtain the road boundary;

[0045] The first determining module is further used to determine the area enclosed by the road boundary as the drivable area;

[0046] The first filtering module is also used to filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud to obtain the first laser radar point cloud.

[0047] In a second possible implementation manner of the second aspect, the apparatus further includes:

[0048] a detection unit, configured to detect dynamic obstacles in the first laser radar point cloud based on a target detection model before filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud;

[0049] A third filtering unit, used to filter the detected dynamic obstacles to obtain a third laser radar point cloud;

[0050] The second filtering unit is used to filter the ground point cloud in the third laser radar point cloud according to the laser radar ray method to obtain the second laser radar point cloud.

[0051] In a third possible implementation manner of the second aspect, the second filtering unit includes:

[0052] A second acquisition module, used to acquire a two-dimensional laser radar point cloud in the third laser radar point cloud;

[0053] A second determination module is used to determine a two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to a laser radar ray method;

[0054] The second filtering module is used to filter the two-dimensional ground point cloud in the two-dimensional laser radar point cloud, and superimpose the filtered two-dimensional laser radar point cloud with the depth information of the filtered two-dimensional laser radar point cloud recorded in the third laser radar point cloud to obtain a second laser radar point cloud, or superimpose the two-dimensional ground point cloud with the depth information corresponding to the two-dimensional ground point cloud in the third laser radar point cloud to obtain a three-dimensional ground point cloud, and filter the three-dimensional ground point cloud in the third laser radar point cloud to obtain the second laser radar point cloud.

[0055] In a fourth possible implementation manner of the second aspect, the matching determination unit includes:

[0056] A matching module, used to match the bounding box size information of each of the suspected static obstacles with the size ranges of different types of obstacles specified in the obstacle geometry rules;

[0057] The third determination module is configured to determine that the suspected static obstacle is the real static obstacle when the bounding box size information of the suspected static obstacle matches the size range of a certain category of obstacles; and to determine that the suspected static obstacle is not the real static obstacle when the bounding box size information of the suspected static obstacle does not match the size range of all categories of obstacles.

[0058] It can be seen from the above scheme that compared with the traditional laser radar perception scheme that perceives obstacles through laser reflection signals, after obtaining the original laser radar point cloud, the embodiment of the present application can first filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud, and then filter the ground point cloud, thereby obtaining foreground information including only obstacle point clouds (called the second laser radar point cloud), and then clustering the second laser radar point cloud to obtain the bounding box size information of at least one suspected static obstacle, and finally by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule, determine the real static obstacle from at least one suspected static obstacle. It can be seen from this that the embodiment of the present application can detect each static obstacle contained in each frame of the original laser radar point cloud by clustering the second laser radar point cloud corresponding to the foreground information and matching the obstacle geometry rule, without being affected by factors such as size and low frequency of occurrence, thereby improving the accuracy of detecting static obstacles.

[0059] In a third aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any possible implementation manner of the first aspect.

[0060] In a fourth aspect, an embodiment of the present application provides an electronic device, the electronic device comprising:

[0061] one or more processors;

[0062] The processor is coupled to a storage device, and the storage device is used to store one or more programs;

[0063] When one or more programs are executed by one or more processors, the electronic device implements the method described in any possible implementation manner of the first aspect.

[0064] In a fifth aspect, an embodiment of the present application provides a vehicle, the vehicle comprising a device as described in any possible implementation of the second aspect, or comprising an electronic device as described in the fourth aspect.

[0065] In a sixth aspect, an embodiment of the present application provides a computer program product, which includes instructions. When the instructions are executed on a computer or a processor, the computer or the processor executes the method described in any possible implementation of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art description are briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative work.

[0067] Figure 1 A flowchart of an obstacle detection method provided in an embodiment of the present application;

[0068] Figure 2 A schematic diagram of a longitudinal cross-section of a laser radar laser beam provided in an embodiment of the present application;

[0069] Figure 3 A schematic top view of a laser radar laser beam provided in an embodiment of the present application;

[0070] Figure 4 A block diagram of an obstacle detection device provided in an embodiment of the present application;

[0071] Figure 5 A schematic diagram of the structure of an electronic device or computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0072] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0073] It should be noted that, in the absence of conflict, the embodiments in the present application and the features in the embodiments can be combined with each other. The terms "including" and "having" in the embodiments of the present application and the accompanying drawings and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or devices.

[0074] Figure 1 The figure is a flowchart of an obstacle detection method. The method can be applied to electronic equipment or computer equipment, and can be specifically applied to vehicles or servers. The method may include the following steps:

[0075] S110: Filtering the laser radar point cloud outside the drivable area in the original laser radar point cloud to obtain a first laser radar point cloud.

[0076] The embodiment of the present application can execute steps S110-S140 for each frame of original laser radar point cloud, thereby detecting the real static obstacles contained in each frame of original laser radar point cloud.

[0077] The specific implementation method of this step S110 may include: first determining the road boundary based on the original lidar point cloud, then determining the area enclosed by the road boundary as the drivable area, and finally filtering the lidar point cloud outside the drivable area in the original lidar point cloud to obtain a first lidar point cloud.

[0078] The specific implementation method of determining the road boundary according to the original laser radar point cloud may include steps A1-A4:

[0079] A1. Search for preset objects around the vehicle posture information in the high-precision map, and filter the preset objects in the original LiDAR point cloud, where the original LiDAR point cloud includes LiDAR point clouds outside the drivable area and LiDAR point clouds within the drivable area.

[0080] Preset objects include traffic lights, street lights, road signs and other objects recorded in high-precision maps. Filtering the preset objects in the original LiDAR point cloud can avoid the interference of preset objects in determining road boundaries and subsequent obstacle detection.

[0081] The vehicle posture information includes the position and posture of the vehicle, and can be located using a positioning device such as GPS (Global Positioning System) or GNSS (Global Navigation Satellite System) to determine the vehicle's posture information.

[0082] A2. Perform plane fitting on the original lidar point cloud after filtering the preset objects to obtain the ground plane.

[0083] Perform plane fitting on the original LiDAR point cloud after filtering the preset objects to obtain the fitting plane. When there are multiple fitting planes, the distance from the vehicle posture to each fitting plane can be calculated separately, and the fitting plane with the smallest distance is selected as the ground plane.

[0084] A3. Determine the boundary line of the ground plane along the direction of vehicle movement, and obtain the laser radar points located on the boundary line as candidate road boundary points.

[0085] The vehicle moving direction can be determined according to the vehicle moving track. The ground plane includes at least four boundary lines, and the boundary line with the same direction as the vehicle moving direction can be determined as the boundary line including the road boundary candidate point.

[0086] A4. Perform curve fitting on the road boundary candidate points to obtain the road boundary.

[0087] S120: Filter the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud.

[0088] Specifically, the two-dimensional laser radar point cloud in the first laser radar point cloud can be obtained first, and then the two-dimensional ground point cloud in the two-dimensional laser radar point cloud can be determined according to the laser radar ray method, and finally the second laser radar point cloud can be obtained by any of the following methods:

[0089] Method 1: Filter the two-dimensional ground point cloud in the two-dimensional laser radar point cloud, and superimpose the filtered two-dimensional laser radar point cloud and the depth information of the filtered two-dimensional laser radar point cloud recorded in the first laser radar point cloud to obtain the second laser radar point cloud. That is, this method first filters the two-dimensional ground point cloud to obtain two-dimensional foreground information, and then superimposes the depth information on the two-dimensional foreground information to obtain three-dimensional foreground information, that is, to obtain the second laser radar point cloud.

[0090] Method 2: Superimpose the depth information corresponding to the two-dimensional ground point cloud with the two-dimensional ground point cloud in the first laser radar point cloud to obtain a three-dimensional ground point cloud, and filter the three-dimensional ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud. That is, this method first superimposes the depth information on the two-dimensional ground point cloud to obtain a three-dimensional ground point cloud, and then filters the three-dimensional ground point cloud to obtain three-dimensional foreground information, that is, to obtain the second laser radar point cloud.

[0091] The core of the laser radar ray method (Ray Ground Filter) is to organize the point cloud in the form of rays. It reduces the (x, y, z) three-dimensional space of the laser radar point cloud to the (x, y) plane, calculates the plane angle from each laser radar point to the positive direction of the vehicle (the x-axis of the vehicle body coordinate system), and differentiates 360 degrees at the angle resolution of the laser radar. The angle of each part is 0.18 degrees. The equal parts of this differentiation can be approximately regarded as a ray. Figure 2 and Figure 3They are respectively a schematic diagram of the longitudinal section and a top view of the laser radar beam. The laser radar has multiple lasers distributed from bottom to top, emitting radioactive laser beams. These laser beams are transformed into multiple rays on flat ground. An n-line laser radar at the same angle should have n such rays. In order to facilitate the processing of beams at the same angle, the original laser radar point cloud in the rectangular coordinate system should be converted into a point cloud data structure described by the cylindrical coordinate system. The points on the beams at the same angle are sorted according to the size of the radius. Whether the slope of the front and rear points is greater than the preset slope threshold is used to determine whether the laser radar point corresponding to the beam is a ground point. That is, when the slope of the front and rear points is greater than the preset slope threshold, the laser radar point corresponding to the beam is determined to be a ground point. Otherwise, the laser radar point corresponding to the beam is determined not to be a ground point.

[0092] S130: Clustering the second laser radar point cloud to obtain bounding box size information of at least one suspected static obstacle.

[0093] After obtaining the second lidar point cloud, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) clustering algorithm or other clustering algorithms can be used to cluster the second lidar point cloud to obtain the obstacle point clusters in the foreground information, thereby calculating the bounding box size information of each suspected static obstacle. The bounding box size information includes the length, width and height of the bounding box.

[0094] S140: Determine a real static obstacle from at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule.

[0095] Specifically, the bounding box size information of each suspected static obstacle can be matched with the size ranges of different categories of obstacles specified in the obstacle geometry rules; when the bounding box size information of the suspected static obstacle matches the size range of a certain category of obstacles, the suspected static obstacle is determined to be a real static obstacle; when the bounding box size information of the suspected static obstacle does not match the size ranges of all categories of obstacles, the suspected static obstacle is determined not to be a real static obstacle. The embodiment of the present application accurately determines whether the suspected static obstacle is a real static obstacle by matching the bounding box size information of each suspected static obstacle with the size ranges of different categories of obstacles specified in the obstacle geometry rules, thereby improving the accuracy and efficiency of detecting real static obstacles.

[0096] Among them, the size ranges of different types of obstacles include:

[0097] (1) Small cylindrical shape: Taking a small cylindrical column as an example, the size range includes length (0, 0.25m], width (0, 0.25m], height [0.35m, 0.85m];

[0098] (2) Small cube: size range includes length [0.25m, 1.0m], width [0.25m, 1.0m], height [0.25, 1.0m];

[0099] (3) Medium-sized spheres / cubes: Taking the spherical stone pier as an example, the size range includes length [0.4, 0.8m], width [0.4, 0.8m], and height [0.4, 0.8m];

[0100] (4) Large cube: Taking a trash can (but not limited to a trash can) as an example, the size range includes length [0.4, 1.0 m], width [0.4, 1.0 m], and height [0.8, 1.0 m].

[0101] In addition, when the bounding box size information of the suspected static obstacle to be matched successfully matches the size range of a certain category of obstacles, in addition to determining that the suspected static obstacle to be matched is a real static obstacle, it can also be determined that the category of the real static obstacle is the category corresponding to the matched size range. And after detecting the real static obstacle contained in each frame of the original LiDAR point cloud, it can also be determined whether the proportion of the same real static obstacle in multiple consecutive frames of the original LiDAR point cloud exceeds a preset proportion threshold. When it exceeds the preset proportion threshold, it is finally determined that the same real static obstacle is a real static obstacle.

[0102] Compared with the traditional laser radar sensing scheme that senses obstacles through laser reflection signals, the obstacle detection method provided in the embodiment of the present application can first filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud after obtaining the original laser radar point cloud, and then filter the ground point cloud, thereby obtaining foreground information including only obstacle point clouds (called the second laser radar point cloud), and then clustering the second laser radar point cloud to obtain the bounding box size information of at least one suspected static obstacle, and finally by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule, determine the real static obstacle from at least one suspected static obstacle. It can be seen from this that the embodiment of the present application can detect each static obstacle contained in each frame of the original laser radar point cloud by clustering the second laser radar point cloud corresponding to the foreground information and matching the obstacle geometry rule, without being affected by factors such as size and low frequency of occurrence, thereby improving the accuracy of detecting static obstacles.

[0103] In one embodiment, in order to prevent interference from dynamic obstacles such as vehicles and pedestrians, the embodiment of the present application can detect dynamic obstacles in the first lidar point cloud based on the target detection model before filtering the ground point cloud in the first lidar point cloud to obtain the second lidar point cloud, and filter the detected dynamic obstacles to obtain a third lidar point cloud, and then filter the ground point cloud in the third lidar point cloud to obtain the second lidar point cloud.

[0104] Among them, the target detection model can be a PointRCNN model or other models. The embodiment of the present application does not limit the model used for dynamic target detection.

[0105] The specific implementation method of filtering the ground point cloud in the third laser radar point cloud to obtain the second laser radar point cloud includes: obtaining a two-dimensional laser radar point cloud in the third laser radar point cloud; determining the two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to the laser radar ray method; filtering the two-dimensional ground point cloud in the two-dimensional laser radar point cloud, and superimposing the depth information of the filtered two-dimensional laser radar point cloud and the filtered two-dimensional laser radar point cloud recorded in the third laser radar point cloud to obtain the second laser radar point cloud, or superimposing the two-dimensional ground point cloud with the depth information corresponding to the two-dimensional ground point cloud in the third laser radar point cloud to obtain a three-dimensional ground point cloud, filtering the three-dimensional ground point cloud in the third laser radar point cloud to obtain the second laser radar point cloud.

[0106] It can be seen from the above scheme that after the embodiment of the present application determines the two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to the laser radar ray method, the three-dimensional foreground information, i.e., the second laser radar point cloud, can be accurately and quickly determined by two methods. One method includes first filtering the two-dimensional ground point cloud to obtain two-dimensional foreground information, and then superimposing depth information on the two-dimensional foreground information to obtain three-dimensional foreground information. Another method includes first superimposing depth information on the two-dimensional ground point cloud to obtain a three-dimensional ground point cloud, and then filtering the three-dimensional ground point cloud to obtain three-dimensional foreground information.

[0107] Corresponding to the above method embodiment, another embodiment of the present application provides an obstacle detection device, such as Figure 4 As shown, the device comprises:

[0108] A first filtering unit 210 is used to filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud to obtain a first laser radar point cloud;

[0109] A second filtering unit 220, configured to filter the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud;

[0110] A clustering unit 230, configured to cluster the second laser radar point cloud to obtain size information of a bounding box of at least one suspected static obstacle;

[0111] The matching determination unit 240 is configured to determine a real static obstacle from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule.

[0112] In a possible implementation, the first filtering unit 210 includes:

[0113] A search module is used to search for preset objects around the vehicle's posture information in the high-precision map;

[0114] A first filtering module, configured to filter the preset object in the original laser radar point cloud, wherein the original laser radar point cloud includes a laser radar point cloud located outside the drivable area and a laser radar point cloud located within the drivable area;

[0115] A plane fitting module, used for performing plane fitting on the original laser radar point cloud after filtering the preset object to obtain a ground plane;

[0116] A first determination module, configured to determine a boundary line of the ground plane along a moving direction of the vehicle;

[0117] A first acquisition module is used to acquire a laser radar point located on the boundary line as a road boundary candidate point;

[0118] A curve fitting module, used for performing curve fitting on the candidate road boundary points to obtain the road boundary;

[0119] The first determining module is further used to determine the area enclosed by the road boundary as the drivable area;

[0120] The first filtering module is also used to filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud to obtain the first laser radar point cloud.

[0121] In a possible implementation, the device further includes:

[0122] a detection unit, configured to detect dynamic obstacles in the first laser radar point cloud based on a target detection model before filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud;

[0123] A third filtering unit, used to filter the detected dynamic obstacles to obtain a third laser radar point cloud;

[0124] The second filtering unit 220 is used to filter the ground point cloud in the third lidar point cloud according to the lidar ray method to obtain the second lidar point cloud.

[0125] In a possible implementation manner, the second filtering unit 220 includes:

[0126] A second acquisition module, used to acquire a two-dimensional laser radar point cloud in the third laser radar point cloud;

[0127] A second determination module is used to determine a two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to the laser radar ray method;

[0128] The second filtering module is used to filter the two-dimensional ground point cloud in the two-dimensional laser radar point cloud, and superimpose the filtered two-dimensional laser radar point cloud with the depth information of the filtered two-dimensional laser radar point cloud recorded in the third laser radar point cloud to obtain a second laser radar point cloud, or superimpose the two-dimensional ground point cloud with the depth information corresponding to the two-dimensional ground point cloud in the third laser radar point cloud to obtain a three-dimensional ground point cloud, and filter the three-dimensional ground point cloud in the third laser radar point cloud to obtain the second laser radar point cloud.

[0129] In a possible implementation, the matching determination unit 240 includes:

[0130] A matching module, used to match the bounding box size information of each of the suspected static obstacles with the size ranges of different types of obstacles specified in the obstacle geometry rules;

[0131] The third determination module is configured to determine that the suspected static obstacle is the real static obstacle when the bounding box size information of the suspected static obstacle matches the size range of a certain category of obstacles; and to determine that the suspected static obstacle is not the real static obstacle when the bounding box size information of the suspected static obstacle does not match the size range of all categories of obstacles.

[0132] Compared with the traditional laser radar sensing scheme that senses obstacles through laser reflection signals, the obstacle detection device provided in the embodiment of the present application can first filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud after obtaining the original laser radar point cloud, and then filter the ground point cloud, thereby obtaining foreground information including only obstacle point clouds (called the second laser radar point cloud), and then clustering the second laser radar point cloud to obtain the bounding box size information of at least one suspected static obstacle, and finally by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule, determine the real static obstacle from at least one suspected static obstacle. It can be seen from this that the embodiment of the present application can detect each static obstacle contained in each frame of the original laser radar point cloud by clustering the second laser radar point cloud corresponding to the foreground information and matching the obstacle geometry rule, without being affected by factors such as size and low frequency of occurrence, thereby improving the accuracy of detecting static obstacles.

[0133] Based on the above method embodiments, another embodiment of the present application provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method described in any of the above embodiments is implemented.

[0134] Based on the above method embodiment, another embodiment of the present application provides an electronic device or a computer device, such as Figure 5 As shown, including:

[0135] One or more processors 310;

[0136] The processor 310 is coupled to a storage device 320, and the storage device 320 is used to store one or more programs;

[0137] When the one or more programs are executed by the one or more processors 310, the electronic device or computer device implements the method described in any of the above embodiments.

[0138] Based on the above method embodiments, another embodiment of the present application provides a vehicle, which includes the device as described in any of the above embodiments, or includes the electronic device as described above.

[0139] The vehicle includes a CPU (Central Processing Unit), a T-Box (Telematics Box), and a laser radar. Among them, the laser radar is used for the original laser radar point cloud. The CPU can execute the above obstacle detection method after obtaining the original laser radar point cloud transmitted by the laser radar. The CPU can also upload the original laser radar point cloud to the server through the T-Box, and the server executes the above obstacle detection method. The vehicle can also include a positioning device, such as a GPS device or a GNSS device. The positioning device can locate the vehicle's posture information. When the CPU or the server determines the road boundary, it can first search for preset objects around the vehicle's posture information in the high-precision map, and filter the preset objects in the original laser radar point cloud, so as to avoid the interference of the preset objects in determining the road boundary and subsequent obstacle detection.

[0140] Based on the above embodiments, another embodiment of the present application provides a computer program product, which includes instructions. When the instructions are executed on a computer or a processor, the computer or the processor executes the method described in any of the above embodiments.

[0141] The above device embodiment corresponds to the method embodiment and has the same technical effect as the method embodiment. For specific description, please refer to the method embodiment. The device embodiment is obtained based on the method embodiment. For specific description, please refer to the method embodiment part, which will not be repeated here. It can be understood by those skilled in the art that the accompanying drawings are only schematic diagrams of one embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present application.

[0142] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An obstacle detection method, characterized in that: The method comprises: Filtering the laser radar point cloud outside the drivable area in the original laser radar point cloud to obtain a first laser radar point cloud; Filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud; Clustering the second laser radar point cloud to obtain bounding box size information of at least one suspected static obstacle; The real static obstacle is determined from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule.

2. The method according to claim 1, characterized in that The filtering of the laser radar point cloud outside the drivable area in the original laser radar point cloud to obtain the first laser radar point cloud includes: Searching for a preset object located around the vehicle posture information in the high-precision map, and filtering the preset object in the original laser radar point cloud, wherein the original laser radar point cloud includes a laser radar point cloud located outside the drivable area and a laser radar point cloud located within the drivable area; Performing plane fitting on the original laser radar point cloud after filtering the preset object to obtain a ground plane; Determine a boundary line of the ground plane along the direction of vehicle movement, and obtain a laser radar point located on the boundary line as a road boundary candidate point; Performing curve fitting on the candidate road boundary points to obtain a road boundary, and determining an area enclosed by the road boundary as the drivable area; The laser radar point cloud located outside the drivable area in the original laser radar point cloud is filtered to obtain the first laser radar point cloud.

3. The method according to claim 1, characterized in that Before filtering the ground point cloud in the first laser radar point cloud to obtain the second laser radar point cloud, the method further includes: Based on the target detection model, dynamic obstacles in the first laser radar point cloud are detected, and the detected dynamic obstacles are filtered to obtain a third laser radar point cloud; The filtering of the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud includes: The ground point cloud in the third laser radar point cloud is filtered according to the laser radar ray method to obtain the second laser radar point cloud.

4. The method according to claim 3, characterized in that The filtering of the ground point cloud in the third laser radar point cloud to obtain the second laser radar point cloud includes: Obtaining a two-dimensional laser radar point cloud in the third laser radar point cloud; Determine a two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to a laser radar ray method; Filter the two-dimensional ground point cloud in the two-dimensional lidar point cloud, and superimpose the filtered two-dimensional lidar point cloud with the depth information of the filtered two-dimensional lidar point cloud recorded in the third lidar point cloud to obtain a second lidar point cloud, or superimpose the two-dimensional ground point cloud with the depth information corresponding to the two-dimensional ground point cloud in the third lidar point cloud to obtain a three-dimensional ground point cloud, and filter the three-dimensional ground point cloud in the third lidar point cloud to obtain the second lidar point cloud.

5. The method according to any one of claims 1 to 4, characterized in that The determining a real static obstacle from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule comprises: Matching the bounding box size information of each of the suspected static obstacles with the size ranges of different types of obstacles specified in the obstacle geometry rules; When the size information of the bounding box of the suspected static obstacle matches the size range of a certain category of obstacles, determining that the suspected static obstacle is the real static obstacle; When the bounding box size information of the suspected static obstacle does not match the size ranges of all categories of obstacles, it is determined that the suspected static obstacle is not the real static obstacle.

6. An obstacle detection device, characterized in that: The device comprises: A first filtering unit is used to filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud to obtain a first laser radar point cloud; A second filtering unit, used for filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud; A clustering unit, configured to cluster the second laser radar point cloud to obtain size information of a bounding box of at least one suspected static obstacle; The matching determination unit is used to determine the real static obstacle from the at least one suspected static obstacle by matching the bounding box size information of each suspected static obstacle with the obstacle geometry rule.

7. The device according to claim 6, characterized in that The first filtering unit comprises: A search module is used to search for preset objects around the vehicle's posture information in the high-precision map; A first filtering module, configured to filter the preset object in the original laser radar point cloud, wherein the original laser radar point cloud includes a laser radar point cloud located outside the drivable area and a laser radar point cloud located within the drivable area; A plane fitting module, used for performing plane fitting on the original laser radar point cloud after filtering the preset object to obtain a ground plane; A first determination module, configured to determine a boundary line of the ground plane along a moving direction of the vehicle; A first acquisition module is used to acquire a laser radar point located on the boundary line as a road boundary candidate point; A curve fitting module, used for performing curve fitting on the road boundary candidate points to obtain the road boundary; The first determining module is further used to determine the area enclosed by the road boundary as the drivable area; The first filtering module is also used to filter the laser radar point cloud located outside the drivable area in the original laser radar point cloud to obtain the first laser radar point cloud.

8. The device according to claim 6, characterized in that The device also includes: a detection unit, configured to detect dynamic obstacles in the first laser radar point cloud based on a target detection model before filtering the ground point cloud in the first laser radar point cloud to obtain a second laser radar point cloud; A third filtering unit, used to filter the detected dynamic obstacles to obtain a third laser radar point cloud; The second filtering unit is used to filter the ground point cloud in the third laser radar point cloud according to the laser radar ray method to obtain the second laser radar point cloud.

9. The device according to claim 8, characterized in that The second filtering unit comprises: A second acquisition module, used to acquire a two-dimensional laser radar point cloud in the third laser radar point cloud; A second determination module is used to determine a two-dimensional ground point cloud in the two-dimensional laser radar point cloud according to a laser radar ray method; The second filtering module is used to filter the two-dimensional ground point cloud in the two-dimensional laser radar point cloud, and superimpose the filtered two-dimensional laser radar point cloud with the depth information of the filtered two-dimensional laser radar point cloud recorded in the third laser radar point cloud to obtain a second laser radar point cloud, or superimpose the two-dimensional ground point cloud with the depth information corresponding to the two-dimensional ground point cloud in the third laser radar point cloud to obtain a three-dimensional ground point cloud, and filter the three-dimensional ground point cloud in the third laser radar point cloud to obtain the second laser radar point cloud.

10. The device according to any one of claims 6 to 9, characterized in that: The matching determination unit includes: A matching module, used to match the bounding box size information of each of the suspected static obstacles with the size ranges of different types of obstacles specified in the obstacle geometry rules; The third determination module is configured to determine that the suspected static obstacle is the real static obstacle when the bounding box size information of the suspected static obstacle matches the size range of a certain category of obstacles; and to determine that the suspected static obstacle is not the real static obstacle when the bounding box size information of the suspected static obstacle does not match the size range of all categories of obstacles.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

12. An electronic device, characterized in that: The electronic device comprises: one or more processors; The processor is coupled to a storage device, and the storage device is used to store one or more programs; When the one or more programs are executed by the one or more processors, the electronic device implements the method according to any one of claims 1 to 5.

13. A vehicle, characterized in that: The vehicle comprises the device according to any one of claims 6 to 10, or comprises the electronic device according to claim 12.

Images