Obstacle detection methods, devices, vehicles, storage media, and software products

Abstract

This disclosure relates to an obstacle detection method, apparatus, vehicle, storage medium, and program product. The obstacle detection method includes: acquiring a target point cloud of a region surrounding a vehicle detected by an onboard environmental sensor; determining multiple types of grids by performing grid projection on the target point cloud, wherein each type of grid includes a projected point cloud corresponding to the target point cloud, and each type of grid uses a different grid unit division method; and determining target obstacle information based on the projected point clouds included in each type of grid. This technical solution can avoid the problem of missed obstacle detection and improve the accuracy of obstacle detection.

Description

Technical Field

[0001] This disclosure relates to the field of vehicle technology, and in particular to obstacle detection methods, devices, vehicles, storage media, and program products. Background Technology

[0002] With the development of vehicle technology, vehicle intelligent perception technology has also developed accordingly. In vehicle intelligent perception scenarios, vehicles can detect obstacles around them using information detected by sensors. This detected obstacle information can be used in intelligent driving scenarios such as autonomous driving, planning, and control. Summary of the Invention

[0003] To overcome the problems existing in related technologies, this disclosure provides an obstacle detection method, device, vehicle, storage medium, and program product.

[0004] According to a first aspect of the present disclosure, an obstacle detection method is provided, comprising: acquiring a target point cloud of a vehicle surrounding area detected by an on-board environmental sensor; determining multiple types of grids by performing grid projection on the target point cloud, wherein each type of grid includes a projected point cloud corresponding to the target point cloud, and each type of grid adopts a different grid unit division method; and determining target obstacle information based on the projected point clouds included in each of the multiple types of grids.

[0005] Optionally, the target point cloud is a three-dimensional point cloud in a world coordinate system. The step of determining multiple types of grids by raster projection of the target point cloud includes: projecting the target point cloud onto a two-dimensional plane in the world coordinate system to obtain a projected point cloud corresponding to the target point cloud in the two-dimensional plane; determining a first type of grid based on a first type of grid unit in the two-dimensional plane that includes the projected point cloud; and determining a second type of grid based on a second type of grid unit in the two-dimensional plane that includes the projected point cloud. The second type of grid unit and the first type of grid unit are obtained by using different grid unit division methods in the two-dimensional plane.

[0006] Optionally, determining the target obstacle information based on the projected point clouds included in the various types of grids includes: for any one type of grid, determining the number of projected point clouds included in each grid cell of that grid; determining a target grid cell from the grid cells of that grid, based at least on the number of projected point clouds included in each grid cell of that grid, the target grid cell including the projected point cloud corresponding to the obstacle; determining the obstacle information corresponding to that grid based on the projected point cloud included in the target grid cell; and determining the target obstacle information based on the obstacle information corresponding to the various types of grids.

[0007] Optionally, determining the target obstacle information based on the projected point clouds included in the various types of grids includes: for any one type of grid, determining the number of projected point clouds included in each grid cell of that grid; determining a target grid cell from the grid cells of that grid, based at least on the number of projected point clouds included in each grid cell of that grid, the target grid cell including the projected point cloud corresponding to the obstacle; determining the obstacle information corresponding to that grid based on the projected point cloud included in the target grid cell; and determining the target obstacle information based on the obstacle information corresponding to the various types of grids.

[0008] Optionally, the historical projected point cloud information corresponding to each grid cell includes: the cumulative number of projected point clouds and the cumulative number of projected point clouds. The cumulative number of projected point clouds is obtained by accumulating the number of historical projected point clouds included in the grid cell, and the cumulative number of projected point clouds is obtained by accumulating the number of times the grid cell includes historical projected point clouds.

[0009] Optionally, determining the target grid cell from the grid cells of the grid based on the number of projected point clouds included in each grid cell of the grid and the corresponding historical projected point cloud information includes: updating the cumulative number of projected point clouds corresponding to each grid cell based on the number of projected point clouds included in each grid cell of the grid and the cumulative number of projected point clouds corresponding to each grid cell, to obtain an updated cumulative number of projected point clouds; updating the cumulative number of projected point clouds corresponding to each grid cell, to obtain an updated cumulative number of projected point clouds; determining the confidence level corresponding to each grid cell based on the updated cumulative number of projected point clouds corresponding to each grid cell; and determining the target grid cell from the grid cells of the grid based on the confidence level corresponding to each grid cell and the updated cumulative number of projected point clouds corresponding to each grid cell.

[0010] Optionally, the step of updating the cumulative number of projected point clouds corresponding to each grid cell based on the number of projected point clouds included in each grid cell of the grid and the cumulative number of projected point clouds corresponding to each grid cell to obtain the updated cumulative number of projected point clouds includes: determining the logarithm of the number of projected point clouds included in each grid cell; and adding the logarithm of the number of projected point clouds included in each grid cell to the cumulative number of projected point clouds corresponding to each grid cell to obtain the updated cumulative number of projected point clouds.

[0011] Optionally, the projected point cloud is a point cloud in world coordinates. Determining the obstacle information corresponding to this type of grid based on the projected point cloud included in the target grid unit includes: converting the projected point cloud included in the target grid unit from the world coordinate system to an obstacle point cloud in the vehicle coordinate system; determining the size information of the obstacle based on the obstacle point cloud; determining the relative position information between the obstacle and the vehicle based on the obstacle point cloud and the vehicle position information; and determining the obstacle information corresponding to this type of grid based on the size information and the relative position information.

[0012] Optionally, the target point cloud is the point cloud corresponding to a stationary object, and the detection method further includes: merging adjacent target grid cells that include the projected point clouds corresponding to the same stationary object to obtain merged target grid cells; correspondingly, converting the projected point cloud included in the target grid cell from the world coordinate system to an obstacle point cloud in the vehicle coordinate system includes: converting the projected point cloud included in the merged target grid cell from the world coordinate system to an obstacle point cloud in the vehicle coordinate system.

[0013] Optionally, acquiring the target point cloud of the area surrounding the vehicle detected by the vehicle-mounted environmental sensor includes: acquiring vehicle location information and multiple frames of original point cloud of the area surrounding the vehicle detected by the vehicle-mounted environmental sensor; determining the point cloud corresponding to the stationary object based on the multiple frames of original point cloud and the vehicle location information; and determining the point cloud corresponding to the stationary object as the target point cloud.

[0014] Optionally, determining the point cloud corresponding to the stationary object based on the multiple original point clouds and the vehicle position information includes: converting the multiple original point clouds into multiple point clouds in the vehicle coordinate system; filtering the multiple point clouds in the vehicle coordinate system to obtain multiple filtered point clouds; converting the multiple filtered point clouds into multiple point clouds in the world coordinate system based on the vehicle position information; clustering the multiple point clouds in the world coordinate system to obtain multiple clustered point clouds; and determining the point cloud corresponding to the stationary object based on the multiple clustered point clouds.

[0015] Optionally, the vehicle-mounted environmental sensor is a 4D millimeter-wave radar.

[0016] According to a second aspect of the present disclosure, an obstacle detection device is provided, comprising: an acquisition module configured to acquire a target point cloud of a vehicle surrounding area detected by an on-board environmental sensor; a determination module configured to determine multiple types of grids by performing grid projection on the target point cloud, wherein the multiple types of grids respectively include a projected point cloud corresponding to the target point cloud, and the multiple types of grids respectively employ different grid unit division methods; and a detection module configured to determine target obstacle information based on the projected point clouds included in the multiple types of grids.

[0017] According to a third aspect of the present disclosure, a vehicle is provided, comprising: a processor; a memory for storing processor-executable instructions; wherein the processor is configured to execute the executable instructions to implement the obstacle detection method as described in the first aspect of the present disclosure.

[0018] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided having computer program instructions stored thereon, which, when executed by a processor, implement the obstacle detection method as described in the first aspect of the present disclosure.

[0019] According to a fifth aspect of the present disclosure, a computer program product is provided, comprising: a computer program that, when executed by a processor, implements the obstacle detection method as described in the first aspect of the present disclosure.

[0020] The technical solutions provided by the embodiments of this disclosure may include the following beneficial effects: By projecting the target point cloud of the area surrounding the vehicle into a grid, multiple types of grids are identified. Each type of grid includes the projected point cloud corresponding to the target point cloud. Based on the projected point clouds included in each type of grid, target obstacle information is determined. Since the different grid types employ different grid cell division methods, the distribution of the projected point clouds within each type of grid will vary under different division methods. Compared to a single grid format, obstacle detection can be performed based on a wider range of projected point cloud distributions, resulting in more comprehensive obstacle detection information. This helps avoid missed obstacle detection and improves the accuracy of obstacle detection.

[0021] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0022] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0023] Figure 1This is a flowchart illustrating an obstacle detection method according to an exemplary embodiment.

[0024] Figure 2 This is a schematic diagram illustrating an obstacle detection process according to an exemplary embodiment.

[0025] Figure 3A This is a schematic diagram of a first type of grid according to an exemplary embodiment.

[0026] Figure 3B This is a schematic diagram of a second type of grid according to an exemplary embodiment.

[0027] Figure 4 This is a schematic diagram illustrating a general obstacle detection process according to an exemplary embodiment.

[0028] Figure 5 This is a block diagram of an obstacle detection device according to an exemplary embodiment.

[0029] Figure 6 This is a block diagram illustrating a vehicle according to an exemplary embodiment. Detailed Implementation

[0030] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0031] It should be noted that all actions involving the acquisition of signals, information, or data in this disclosure are carried out in compliance with the relevant data protection laws and policies of the country where the location is situated, and with authorization from the owner of the relevant device.

[0032] With the development of vehicle technology, vehicle intelligent perception technology has also developed accordingly. In vehicle intelligent perception scenarios, vehicles can detect obstacles around them using information detected by sensors. This detected obstacle information can be used in intelligent driving scenarios such as autonomous driving, planning, and control.

[0033] In intelligent driving scenarios, the accuracy of obstacle detection is crucial, as it can affect the reliability and stability of intelligent driving functions.

[0034] In some related technologies, obstacle detection is based on visual models. This detection scheme acquires image data through visual sensors and identifies obstacles based on that image data. However, this method can only identify target obstacles that have been centrally labeled in the dataset; it is a whitelist-based approach. This method cannot identify the diverse types of obstacles encountered in real-world scenarios, resulting in low obstacle detection accuracy.

[0035] Based on this, the present disclosure provides a technical solution in which grid projection is performed on the target point cloud in the area surrounding the vehicle to determine multiple types of grids. Each type of grid includes the projected point cloud corresponding to the target point cloud. Based on the projected point clouds included by each type of grid, target obstacle information is determined.

[0036] Because different types of grids employ different grid cell division methods, the distribution of projected point clouds within each type of grid will vary under these different methods. Compared to a single grid format, obstacle detection can be performed based on a wider range of projected point cloud distributions, resulting in more comprehensive obstacle information. This, in turn, avoids missed obstacle detection and improves the accuracy of obstacle detection.

[0037] The issue of missed obstacle detection can be caused by a variety of reasons. For example, in an unstable onboard environment, sensors may become unstable, leading to unstable point cloud data, which in turn can result in missed obstacle detection.

[0038] Figure 1 This is a flowchart illustrating an obstacle detection method according to an exemplary embodiment. The detection method can be applied to vehicles, and the detection method includes the following steps: Step S11: Obtain the target point cloud of the area around the vehicle detected by the on-board environmental sensor.

[0039] Step S12: By performing grid projection on the target point cloud, multiple types of grids are determined. Each type of grid includes the projected point cloud corresponding to the target point cloud. Each type of grid uses a different grid cell division method.

[0040] Step S13: Determine the target obstacle information based on the projected point clouds included in the various types of grids.

[0041] In the embodiments of this disclosure, the point cloud described can be understood as a collection of multiple points, and the processing performed on the point cloud can be understood as processing each point in the point cloud separately. For example, transforming the point cloud coordinates can be understood as performing coordinate transformation on each point in the point cloud separately; projecting the point cloud can be understood as projecting each point in the point cloud separately; and counting the number of points in the point cloud can be understood as counting the number of points included in the point cloud.

[0042] In step S11, the vehicle-mounted environmental sensor can be a sensor capable of detecting point cloud data.

[0043] In some embodiments, the vehicle-mounted environmental sensor can be a millimeter-wave radar. Millimeter-wave radar sensors operate on the principle of echo reflection, characterizing the reflection position through point cloud data. Unlike cameras, they have lower semantic relevance to the target object being detected, do not rely on whitelist annotations, and can achieve general obstacle detection.

[0044] In some embodiments, the vehicle environment sensor may be a 4D millimeter-wave radar. As an emerging vehicle sensor, 4D millimeter-wave radar has a higher number and density of point clouds and higher height characteristics than 3D millimeter-wave radar, achieving a performance level close to that of lidar with fewer lines.

[0045] In addition, since millimeter-wave radar only produces echo reflection point clouds from objects with strong reflections, it has far fewer ground reflection points than lidar point clouds, making its point cloud processing simpler than lidar.

[0046] Furthermore, 4D millimeter-wave radar is much cheaper than lidar, and it features low cost and all-weather operation, making it an ideal sensor for obstacle detection.

[0047] In some embodiments, obstacle detection schemes based on lidar are generally not applicable to obstacle detection scenarios based on millimeter-wave radar, while the technical solutions of the present disclosure embodiments can be well applied to obstacle detection scenarios based on 4D millimeter-wave radar.

[0048] In some embodiments, the vehicle can perform obstacle detection based on multiple frames of point cloud data, with each frame of point cloud data corresponding to the same obstacle detection method.

[0049] Therefore, in step S11, the target point cloud can be the Nth frame point cloud detected by the vehicle environment sensor, where N is an integer.

[0050] In some embodiments, the target point cloud may be a point cloud obtained by processing point cloud data detected by onboard environmental sensors, which may correspond to objects in the area surrounding the vehicle.

[0051] Correspondingly, the target obstacle information determined in step S13 can be object information in the area surrounding the vehicle.

[0052] In some embodiments, the target point cloud may include point clouds corresponding to multiple objects respectively, and the target obstacle information may include information of one or more of these objects.

[0053] In some embodiments, the target point cloud can be the point cloud corresponding to a stationary object. Therefore, based on point cloud data detected by onboard environmental sensors, the point cloud corresponding to the stationary object is determined, and this point cloud corresponding to the stationary object is then identified as the target point cloud.

[0054] Therefore, as an optional implementation, step S11 includes: acquiring vehicle location information and multiple frames of original point clouds of the area surrounding the vehicle detected by the vehicle environment sensor; determining the point cloud corresponding to the stationary object based on the multiple frames of original point clouds and the vehicle location information; and determining the point cloud corresponding to the stationary object as the target point cloud.

[0055] In some embodiments, since the vehicle may be in motion, it is necessary to combine the vehicle position information and multiple frames of raw point clouds of the area around the vehicle detected by the onboard environmental sensors to determine the point cloud corresponding to the stationary object.

[0056] In this implementation, it is necessary to combine multi-frame point cloud data and vehicle location information for analysis to determine the point cloud corresponding to the stationary object.

[0057] In some embodiments, vehicle location information can be obtained through DR (Dead Reckoning) positioning, GPS (Global Positioning System) positioning, etc. These positioning technologies are mature technologies in this field and will not be described in detail here.

[0058] In some embodiments, the acquisition time of vehicle location information can be consistent with the acquisition time of point cloud. For example, for the Nth frame point cloud, the vehicle location information with the nearest neighbor timestamp of the Nth frame point cloud is obtained.

[0059] In some embodiments, determining the point cloud corresponding to a stationary object based on multiple frames of original point clouds and the vehicle position information may include: converting the multiple frames of original point clouds into multiple frames of point clouds in the vehicle coordinate system; filtering the multiple frames of point clouds in the vehicle coordinate system to obtain multiple frames of filtered point clouds; converting the multiple frames of filtered point clouds into multiple frames of point clouds in the world coordinate system based on the vehicle position information; clustering the multiple frames of point clouds in the world coordinate system to obtain multiple frames of clustered point clouds; and determining the point cloud corresponding to the stationary object based on the multiple frames of clustered point clouds.

[0060] It is understandable that the coordinate system transformation of point clouds is achieved through the transformation of point cloud coordinates. The transformation of point cloud coordinates in different coordinate systems can be referred to mature technologies in this field, which will not be described in detail here.

[0061] In some embodiments, the target point cloud is a three-dimensional point cloud. Therefore, in the above processing, all point clouds involved are three-dimensional point clouds.

[0062] In this implementation, since the original point clouds of multiple frames do not belong to the point clouds in the vehicle coordinate system, it is necessary to first convert the original point clouds of multiple frames into point clouds in the vehicle coordinate system, so as to determine the coordinates of the original point clouds of multiple frames in the vehicle coordinate system.

[0063] After determining the coordinates of the original point clouds in the vehicle coordinate system, the point clouds can be filtered according to the specific values ​​of the coordinates.

[0064] In some embodiments, point clouds that meet certain rejection criteria can be removed based on their coordinates in the vehicle coordinate system, and the remaining point clouds can be used as the filtered point clouds. The rejection criteria can be configured according to different scenario requirements. For example, point clouds whose distance from the origin of the vehicle coordinate system is greater than a preset distance can be removed, and the preset distance can be configured according to scenario requirements.

[0065] After point cloud filtering, the multi-frame filtered point clouds can be converted into multi-frame point clouds in the world coordinate system based on the vehicle location information.

[0066] It is understandable that since the point cloud after multi-frame filtering is a point cloud in the vehicle coordinate system, it is necessary to combine it with vehicle position information in order to convert it into a point cloud in the world coordinate system.

[0067] Furthermore, after transforming to world coordinates, the point clouds from multiple frames are clustered separately to obtain clustered point clouds. The clustering of point clouds can be achieved using a clustering algorithm. For example, the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm can be used for clustering.

[0068] Furthermore, based on the point cloud obtained from multi-frame clustering, multi-frame tracking can be performed to obtain a stable tracking target that is continuous in time and space, while also distinguishing between moving and stationary targets. Thus, the point cloud corresponding to a stationary target can be determined.

[0069] In some embodiments, after determining the point cloud corresponding to a stationary target, the mapping relationship between the point cloud and the corresponding target can be saved, and this mapping relationship can be used for subsequent obstacle detection. For example, target A: point cloud one; target B: point cloud two, etc.

[0070] It's understandable that the mapping relationship saved at this point only represents the point cloud corresponding to the stationary target detected in the current frame, but it cannot determine the specific obstacle information. Specific obstacle information needs to be determined through subsequent obstacle detection. Obstacle information includes, for example, the obstacle's position relative to the vehicle and the obstacle's dimensions such as length, width, and height.

[0071] Regarding target tracking technology based on clustered point clouds, mature technologies in this field can be referenced, and will not be described in detail here.

[0072] After determining the point cloud corresponding to the stationary target, the point cloud can be used as the target point cloud, and obstacle detection can be performed using the technical solution of the present disclosure embodiment.

[0073] Figure 2 This is a schematic diagram illustrating an obstacle detection process according to an exemplary embodiment, such as... Figure 2 As shown, in this obstacle detection process, 4D millimeter-wave radar point cloud data and vehicle positioning information are first acquired.

[0074] Based on 4D millimeter-wave radar point cloud data and vehicle positioning information, point cloud preprocessing can be performed. This point cloud preprocessing includes, in sequence, vehicle coordinate system transformation, point cloud filtering (removal), and world coordinate system transformation.

[0075] Then, the point cloud is clustered, and multi-frame tracking is performed based on the clustering results to distinguish moving targets from stationary targets. Furthermore, the mapping relationship between the point cloud and the target is saved.

[0076] Furthermore, the general obstacle detection module performs obstacle detection based on this mapping relationship to obtain the final detected obstacle information.

[0077] Therefore, steps S11 to S13 above can be understood as... Figure 2 The diagram shows the obstacle detection process executed in the general obstacle detection module. Furthermore, this obstacle detection process can be executed for each frame of point cloud data, thereby achieving continuous detection of obstacle information in the area surrounding the vehicle.

[0078] In step S12, the target point cloud is raster projected to determine multiple types of graticles. These multiple types of graticles may belong to different graticle systems in the same plane. Different graticle systems use different graticle cell division methods, so the multiple types of graticles also use different graticle division methods.

[0079] As an optional implementation, the target point cloud is a three-dimensional point cloud in the world coordinate system. Step S12 includes: projecting the target point cloud onto a two-dimensional plane in the world coordinate system to obtain the projected point cloud corresponding to the target point cloud in the two-dimensional plane; determining a first type of grid based on a first type of grid unit including the projected point cloud in the two-dimensional plane; determining a second type of grid based on a second type of grid unit including the projected point cloud in the two-dimensional plane. The second type of grid unit and the first type of grid unit are obtained by using different grid unit division methods in the two-dimensional plane.

[0080] In this implementation, the target point cloud is projected onto a two-dimensional plane of the world coordinate system. This two-dimensional plane may include two types of grid cells, which employ different grid cell division methods.

[0081] Thus, the first type of raster units, including the projected point cloud, can be integrated into a first type of raster, and the second type of raster units, including the projected point cloud, can be integrated into a second type of raster, resulting in two types of raster.

[0082] In some embodiments, the first type of grid cell can be obtained by dividing the two-dimensional plane before the target point cloud is projected, and the second type of grid cell can be obtained by dividing the two-dimensional plane after the first type of grid cell is determined.

[0083] For example, in a two-dimensional plane, this includes the original raster cells obtained using the original raster cell partitioning method. After projecting the target point cloud onto the two-dimensional plane, the original raster is determined based on the original raster cells distributed in the projected point cloud. Next, based on the distribution of the projected point cloud in the original raster, the original raster cell partitioning method is modified to obtain redundant raster cells, thereby determining the redundant raster based on the redundant raster cells distributed in the point cloud.

[0084] The method of dividing the raster into cells determines the relative positions of the raster and the projected point cloud. For example, in the presence of multiple raster types, some projected points in the projected point cloud may reside in multiple different raster cells. Alternatively, multiple projected points in the projected point cloud may reside in the same raster cell under one raster cell division method, but in different raster cell division methods.

[0085] This implementation method can cover a wider range of projected point cloud grid distributions, thus addressing the issue of missed obstacle detection.

[0086] The inclusion of projected point clouds within a raster cell can be understood as the projection point cloud comprising the projected points / point clouds located within the corresponding raster cell. Alternatively, it can be understood as the projection point cloud being projected onto the corresponding raster cell.

[0087] Figure 3A This is a schematic diagram of a first type of grid according to an exemplary embodiment, such as... Figure 3A As shown, the grid includes multiple grid cells, each of which contains a projection point cloud / projection point.

[0088] Figure 3B This is a schematic diagram of a second type of grid according to an exemplary embodiment. Figure 3B There are two types of graticles: original graticles and redundant graticles. It can be seen that the graticle cells of a redundant graticle are offset by half a graticle from the graticle cells of the original graticle. That is, the graticle cells of the redundant graticle and the graticle cells of the original graticle use different graticle cell division methods in the same two-dimensional plane.

[0089] This results in the same projected point cloud having different distribution patterns in the original raster and the redundant raster. For example, some projected point clouds may not be located in the same raster cell in the original raster, but are located in the same raster cell in the redundant raster. Conversely, some projected point clouds may be located in the same raster cell in the original raster, but are not located in the same raster cell in the redundant raster.

[0090] Understandable. Figure 3B The grid cell division method shown is only an example. In different application scenarios, other different grid cell division methods can be adopted according to different needs.

[0091] It is understandable that in a two-dimensional plane, there are also grid cells that do not include the projected point cloud. These grid cells may not be used to determine the corresponding grid in the current frame, but may be used to determine the corresponding grid in the next frame because they include the projected point cloud.

[0092] Furthermore, since the target point cloud is the point cloud corresponding to a stationary target, the position of each grid cell in various types of grids may not change, but the projected point cloud included in each grid cell may change with the point cloud frame.

[0093] Furthermore, in step S13, target obstacle information can be determined based on the projected point clouds included in various types of grids.

[0094] In some embodiments, step S13 may include: for any type of grid, determining the number of projected point clouds included in each grid cell of that grid; determining a target grid cell from each grid cell of that grid, the target grid cell including the projected point cloud corresponding to an obstacle, based at least on the number of projected point clouds included in each grid cell of that grid; determining obstacle information corresponding to that grid based on the projected point cloud included in the target grid cell; and determining target obstacle information based on the obstacle information corresponding to each of the multiple types of grids.

[0095] In this implementation, target grid cells can be determined for each type of grid, and then the corresponding obstacle information can be determined based on the target grid cells. Then, the obstacle information corresponding to each type of grid can be integrated to obtain the target obstacle information.

[0096] It is understandable that the difference between various types of grids lies in the different ways of dividing grid cells, but the same method of determining obstacle information can be used.

[0097] In some embodiments, the number of projected point clouds included in each grid cell can be determined, and based on the number of projected point clouds, the grid cell occupied by the projected point cloud corresponding to the obstacle, i.e., the target grid cell, can be determined.

[0098] In some embodiments, the confidence level of each grid cell can be determined based on the number of projected point clouds included in each grid cell, and whether each grid cell is a target grid cell can be determined based on the confidence level of each grid cell.

[0099] It can be understood that the number of projected point clouds included in each grid cell can be the number of projected point clouds included in each grid cell in the current frame.

[0100] The correspondence between confidence level and the number of projected point clouds can be pre-configured. Based on the configured correspondence, the confidence level can be determined based on the number of projected point clouds.

[0101] For example, if the confidence level of a grid cell is higher than the preset confidence level, that grid cell is identified as the target grid cell. The preset confidence level can be configured in advance according to different application scenarios, and is not limited here.

[0102] This implementation of the target grid unit can be applied to situations where obstacle detection is performed based on the target point cloud of the previous few frames after obstacle detection has been initiated. As the number of target point cloud frames accumulates, more information from the grid units can be combined to determine the target grid unit.

[0103] It is understandable that since the number of projected point clouds included in each grid cell is the number of projected point clouds included in each grid cell in the current frame, and each grid cell may also include corresponding projected point clouds in historical frames, if the situation of projected point clouds included in previous frames is combined, a more reasonable and accurate determination of the target grid cell can be achieved.

[0104] Therefore, as an optional implementation, the detection method further includes: acquiring historical projected point cloud information corresponding to each grid cell of the grid. Correspondingly, determining the target grid cell from the grid cells of the grid, at least based on the number of projected point clouds included in each grid cell, includes: determining the target grid cell from the grid cells of the grid based on the number of projected point clouds included in each grid cell and the corresponding historical projected point cloud information.

[0105] Historical projection point cloud information can characterize the projection point cloud situation of each grid cell in the historical frame.

[0106] As an optional implementation, the historical projected point cloud information corresponding to each grid cell includes: the cumulative number of projected point clouds and the cumulative number of projected point clouds. The cumulative number of projected point clouds is obtained by accumulating the number of historical projected point clouds included in the grid cell, and the cumulative number of projected point clouds is obtained by accumulating the number of times the grid cell includes historical projected point clouds.

[0107] Historical projected point clouds can be understood as point clouds projected onto the corresponding raster cells in frames preceding the current frame. Since point cloud projection is required for every frame preceding the current frame, each raster cell in each frame before the current frame may contain projected point clouds. Therefore, the number of projections and the quantity of projections can be accumulated separately to determine the historical projected point cloud information.

[0108] Therefore, the cumulative number of projected point clouds can be the cumulative number of projected point clouds of the grid cell in the previous frame, and the cumulative number of projected point clouds can be the cumulative number of projected point clouds of the grid cell in the previous frame.

[0109] In some embodiments, when accumulating the number of projected point clouds, the number can be accumulated directly, or the number can be processed first and then accumulated.

[0110] In some embodiments, determining a target grid cell from the grid cells of the grid based on the number of projected point clouds included in each grid cell and the corresponding historical projected point cloud information may include: updating the cumulative number of projected point clouds corresponding to each grid cell based on the number of projected point clouds included in each grid cell and the cumulative number of projected point clouds corresponding to each grid cell, to obtain an updated cumulative number of projected point clouds; updating the cumulative number of projected point clouds corresponding to each grid cell, to obtain an updated cumulative number of projected point clouds; determining the confidence level corresponding to each grid cell based on the updated cumulative number of projected point clouds corresponding to each grid cell; and determining the target grid cell from the grid cells based on the confidence level corresponding to each grid cell and the updated cumulative number of projected point clouds corresponding to each grid cell.

[0111] In this implementation, the cumulative number of projected point clouds is constantly updated. Therefore, the cumulative number of projected point clouds corresponding to each grid can be updated first based on the number of projected point clouds included in each grid cell of the grid and the cumulative number of projected point clouds corresponding to each grid cell.

[0112] Updating the cumulative number of projected point clouds corresponding to each grid can also be understood as accumulating the number of projected point clouds included in each grid.

[0113] As an example, based on the number of projected point clouds included in each grid cell of the grid and the cumulative number of projected point clouds corresponding to each grid cell, the cumulative number of projected point clouds corresponding to each grid cell is updated to obtain the updated cumulative number of projected point clouds. This includes: determining the logarithm of the number of projected point clouds included in each grid cell; and adding the logarithm of the number of projected point clouds included in each grid cell to the cumulative number of projected point clouds corresponding to each grid cell to obtain the updated cumulative number of projected point clouds.

[0114] As an example, the cumulative number of projected point clouds can be represented as: odd current =odd previous +ln(point_num), where odd current The odd value represents the cumulative number of projected point clouds in the current frame, which is also the updated cumulative number of projected point clouds. previous `point_num` represents the cumulative number of projected point clouds in the previous frame, while `point_num` represents the number of projected point clouds in the current frame.

[0115] Therefore, for each grid cell, the cumulative number of projected point clouds in the current frame is accumulated based on the cumulative number of projected point clouds in the previous frame, resulting in the current cumulative number of projected point clouds in the current frame, which is the updated cumulative number of projected point clouds.

[0116] In some embodiments, updating the cumulative number of projected point clouds corresponding to each grid cell may include: incrementing the cumulative number of projected point clouds corresponding to each grid cell by 1 to obtain the updated cumulative number of projected point clouds.

[0117] Therefore, for each grid cell in the current frame, if the grid cell in the current frame includes a projected point cloud, the cumulative number of projected point clouds corresponding to the grid cell in the previous frame will be increased by 1 to obtain the cumulative number of projected point clouds corresponding to the grid cell in the current frame.

[0118] In some embodiments, for grid cells that do not include the projected point cloud in the current frame, the corresponding information can also be updated. However, in this obstacle detection, since the projected point cloud is not included, it can be directly determined that the grid cell does not belong to the target grid cell, so the corresponding information will not be applied. It may be applied in subsequent obstacle detection processes.

[0119] In some embodiments, if a grid cell does not include a corresponding projected point cloud in the current frame, the cumulative number of projected point clouds for that grid cell in the current frame can be accumulated using a preset number of projected point clouds. That is, the cumulative number of projected point clouds for that grid cell in the current frame is obtained by adding the logarithm of the preset number of projected point clouds to the cumulative number of projected point clouds in the previous frame. For example, the preset number of projected point clouds can be 3. Furthermore, the cumulative number of projected point clouds for that grid cell in the current frame remains unchanged.

[0120] It is understandable that each grid cell in a two-dimensional plane will determine the corresponding number of accumulated projected point clouds and the number of times the projected point clouds are accumulated after each frame.

[0121] Furthermore, the cumulative number of updated projected point clouds corresponding to each grid can be used to determine the confidence level of the grid cell.

[0122] As an example, confidence can be expressed as: confidence = e odd / (e odd +1), where confidence represents the confidence level and odd represents the cumulative number of times the projected point cloud has been updated, that is, the cumulative number of times the projected point cloud of this raster cell has been updated in the current frame.

[0123] The confidence level determination method described here is only an example. Other confidence level determination methods can be configured in different scenarios, and no limitation is made here.

[0124] Furthermore, the target grid cell can be determined based on the confidence level of each grid cell and the cumulative number of times the updated projected point cloud is accumulated for each grid cell.

[0125] In some embodiments, the target raster cell is a raster cell whose confidence level and the number of times the updated projected point cloud is accumulated satisfy the corresponding conditions.

[0126] As an example, the confidence level and cumulative number of projected point clouds that the target raster cell must meet can be preset. If the confidence level and cumulative number of projected point clouds of the raster cell meet the corresponding conditions, the raster cell can be determined as the target raster cell.

[0127] In some embodiments, the confidence level condition may include: a first confidence level condition and a second confidence level condition, and the cumulative number of times the projected point cloud is counted may include: a first cumulative number of times the projected point cloud is counted and a second cumulative number of times the projected point cloud is counted.

[0128] If the confidence level of a grid cell satisfies the first confidence level condition, and the sum of the grid cell's projected point cloud cumulative counts satisfies the first projected point cloud cumulative counts condition, or if the confidence level of a grid cell satisfies the second confidence level condition, and the sum of the grid cell's projected point cloud cumulative counts satisfies the second projected point cloud cumulative counts condition, then the grid cell is determined to be the target grid cell.

[0129] In the first confidence level condition, the confidence level is higher than the first preset confidence level; in the first cumulative number of projected point cloud counts condition, the cumulative number of projected point cloud counts is lower than the first preset number. In the second confidence level condition, the confidence level is lower than the second preset confidence level; in the second cumulative number of projected point cloud counts condition, the cumulative number of projected point cloud counts is higher than the second preset number.

[0130] The first preset confidence level is higher than the second preset confidence level, and the first preset number of times is less than the second preset number of times. Therefore, the target raster unit can be a raster unit with a high confidence level and a low cumulative number of times the projected point cloud has been generated, or it can be a raster unit with a low confidence level and a high cumulative number of times the projected point cloud has been generated.

[0131] By combining confidence level and cumulative number of projection point cloud iterations, more accurate determination of grid cells, including those corresponding to obstacles, can be achieved.

[0132] Furthermore, after determining the target grid cell, the corresponding obstacle information can be determined based on the projected point cloud included in the target grid cell.

[0133] As an optional implementation, the projected point cloud of the target grid cell is converted from the world coordinate system to an obstacle point cloud in the vehicle coordinate system; the size information of the obstacle is determined based on the obstacle point cloud; the relative position information of the obstacle and the vehicle is determined based on the obstacle point cloud and the vehicle position information; and the obstacle information corresponding to this grid is determined based on the size information and the relative position information.

[0134] In this implementation, the projected point cloud in the target grid cell is first converted back to the obstacle point cloud in the vehicle coordinate system. Then, the size information of the obstacle is calculated based on the obstacle point cloud in the vehicle coordinate system. The size information may include the length, width, and height of the obstacle.

[0135] In some embodiments, the size information of an obstacle can be determined by first generating a bounding box of the obstacle point cloud based on the distribution of the obstacle point cloud, and then determining the size information of the obstacle based on the size of the bounding box.

[0136] Furthermore, a reference point is determined from the obstacle point cloud, and the positional relationship between this reference point and the vehicle's position is calculated to obtain relative position information. This relative position information can include information such as direction and distance. It can be understood that both the obstacle point cloud and the vehicle position information belong to the vehicle coordinate system; relative position information can be determined based on different positions within the vehicle coordinate system.

[0137] Furthermore, size information and relative position information can be integrated into obstacle information.

[0138] In some embodiments, before determining obstacle information based on the projected point cloud of the target grid cell, merging of grid cells from the same source can be performed. That is, grid cells that include projected point clouds corresponding to the same target are merged to facilitate the determination of obstacle information.

[0139] Therefore, as an optional implementation, the detection method further includes: merging adjacent target grid cells that include the projected point cloud corresponding to the same stationary object to obtain merged target grid cells.

[0140] In this implementation, adjacent target raster cells of the same origin are merged, while target raster cells of the same origin that are far apart are not merged.

[0141] As described in the foregoing embodiments, the target point cloud also retains the mapping relationship between the point cloud and the target. Therefore, based on this mapping relationship, it can be determined whether adjacent target grid cells include the projected point cloud corresponding to the same target.

[0142] Therefore, after merging, the projected point cloud can be converted from the world coordinate system to the obstacle point cloud in the vehicle coordinate system.

[0143] In some embodiments, after determining the obstacle information corresponding to the various grids, the obstacle information can be deduplicated, and the target obstacle information can be determined based on the deduplicated obstacle information.

[0144] In some embodiments, after obtaining the deduplicated obstacle information, it can be determined whether the obstacle information obtained this time belongs to the pre-configured obstacle information. If it belongs, it is identified as the target obstacle information; if it does not belong, it can be regarded as irrelevant obstacle information.

[0145] The pre-configured obstacle information can be general obstacle information that needs to be considered based on the vehicle configuration.

[0146] For example, pre-configured obstacle information could include construction signs, two-wheeled vehicles covered with car covers, piles of stones, and other obstacles.

[0147] Therefore, obstacle information identified based on different types of grids can be filtered and deduplicated to obtain general obstacle information that requires attention. These general obstacles may be those that cannot be detected using conventional visual detection methods, i.e., obstacles not included in the visual detection whitelist.

[0148] In some embodiments, after determining the target obstacle information, the target obstacle information can be synchronized to the intelligent driving system, which will then execute the corresponding braking control.

[0149] In some embodiments, the intelligent driving system may combine the target obstacle information with obstacle information detected by other means (such as visual obstacle detection information) to determine whether to perform braking control.

[0150] In some embodiments, the target obstacle information can be used in many other ways, which are not limited here.

[0151] Figure 4 This is a schematic diagram illustrating a general obstacle detection process according to an exemplary embodiment, such as... Figure 4 As shown, the general obstacle detection process includes: First, obtain the point cloud corresponding to the stationary target, and the point cloud corresponding to the stationary target is the point cloud in the world coordinate system.

[0152] Next, based on the point cloud corresponding to the stationary target, projections are made in the two-dimensional plane of the world coordinate system to obtain two different types of raster.

[0153] Then, information is accumulated based on two different types of raster. The accumulated information includes: the number of accumulated projected point clouds and the number of times the projected point clouds are accumulated.

[0154] Next, based on the accumulated information, the grid cells occupied by obstacles are determined, and adjacent grid cells with the same source are merged.

[0155] Then, based on the projected point cloud in the merged grid, candidate obstacle information is determined by using bounding boxes and reference points.

[0156] Furthermore, by filtering and deduplicating the candidate obstacle information, the final general obstacle information is obtained.

[0157] The technical solution adopted in this disclosure is unaffected by lighting conditions. Furthermore, it has been verified that, relying solely on 4D millimeter-wave radar, it can identify non-visual whitelisted obstacle targets within a range of 70 to 90 meters, such as construction signs, two-wheeled vehicles covered by car covers, and piles of stones, and can achieve automatic braking control of obstacles at speeds up to 70 kph.

[0158] Figure 5 This is a block diagram illustrating an obstacle detection device according to an exemplary embodiment. (Refer to...) Figure 5 The device includes: The acquisition module 501 is configured to acquire a target point cloud of the area surrounding the vehicle detected by the onboard environmental sensors.

[0159] The determination module 502 is configured to determine multiple types of grids by performing grid projection on the target point cloud. The multiple types of grids include the projected point cloud corresponding to the target point cloud, and the multiple types of grids adopt different grid unit division methods.

[0160] The detection module 503 is configured to determine target obstacle information based on the projected point clouds included in the various types of grids.

[0161] Optionally, the target point cloud is a three-dimensional point cloud in the world coordinate system, and the determining module 502 is further configured to: project the target point cloud onto a two-dimensional plane in the world coordinate system to obtain a projected point cloud corresponding to the target point cloud in the two-dimensional plane; determine a first type of grid based on a first type of grid unit in the two-dimensional plane that includes the projected point cloud; and determine a second type of grid based on a second type of grid unit in the two-dimensional plane that includes the projected point cloud, wherein the second type of grid unit and the first type of grid unit are obtained by using different grid unit division methods in the two-dimensional plane.

[0162] Optionally, the detection module 503 is further configured to: for any one of the multiple types of grids, determine the number of projected point clouds included in each grid cell of that grid; determine a target grid cell from each grid cell of that grid based at least on the number of projected point clouds included in each grid cell of that grid, the target grid cell including the projected point cloud corresponding to an obstacle; determine the obstacle information corresponding to that grid based on the projected point cloud included in the target grid cell; and determine the target obstacle information based on the obstacle information corresponding to each of the multiple types of grids.

[0163] Optionally, the acquisition module 501 is further configured to: acquire the historical projection point cloud information corresponding to each grid cell of the grid; the detection module 503 is further configured to: determine the target grid cell from each grid cell of the grid based on the number of projection point clouds included in each grid cell of the grid and the corresponding historical projection point cloud information.

[0164] Optionally, the historical projected point cloud information corresponding to each grid cell includes: the cumulative number of projected point clouds and the cumulative number of projected point clouds. The cumulative number of projected point clouds is obtained by accumulating the number of historical projected point clouds included in the grid cell, and the cumulative number of projected point clouds is obtained by accumulating the number of times the grid cell includes historical projected point clouds.

[0165] Optionally, the detection module 503 is further configured to: update the cumulative number of projected point clouds corresponding to each grid cell based on the number of projected point clouds included in each grid cell of the grid and the cumulative number of projected point clouds corresponding to each grid cell, to obtain an updated cumulative number of projected point clouds; update the cumulative number of projected point clouds corresponding to each grid cell, to obtain an updated cumulative number of projected point clouds; determine the confidence level corresponding to each grid cell based on the updated cumulative number of projected point clouds corresponding to each grid cell; and determine the target grid cell from each grid cell of the grid based on the confidence level corresponding to each grid cell and the updated cumulative number of projected point clouds corresponding to each grid cell.

[0166] Optionally, the detection module 503 is further configured to: determine the logarithm of the number of projected point clouds included in each grid cell; and add the logarithm of the number of projected point clouds included in each grid cell to the cumulative number of projected point clouds corresponding to each grid cell to obtain the updated cumulative number of projected point clouds.

[0167] Optionally, the detection module 503 is further configured to: convert the projected point cloud of the target grid unit from the world coordinate system to an obstacle point cloud in the vehicle coordinate system; determine the size information of the obstacle based on the obstacle point cloud; determine the relative position information of the obstacle and the vehicle based on the obstacle point cloud and the vehicle position information; and determine the obstacle information corresponding to the grid based on the size information and the relative position information.

[0168] Optionally, the detection module 503 is further configured to: merge adjacent target grid cells that include the projected point clouds corresponding to the same stationary object to obtain merged target grid cells; and convert the projected point clouds included in the merged target grid cells from the world coordinate system to obstacle point clouds in the vehicle coordinate system.

[0169] Optionally, the acquisition module 501 is further configured to: acquire vehicle location information and multiple frames of original point clouds of the area surrounding the vehicle detected by the vehicle environment sensor; determine the point cloud corresponding to the stationary object based on the multiple frames of original point clouds and the vehicle location information; and determine the point cloud corresponding to the stationary object as the target point cloud.

[0170] Optionally, the acquisition module 501 is further configured to: convert the multiple frames of original point clouds into multiple frames of point clouds in the vehicle coordinate system; perform point cloud filtering on the multiple frames of point clouds in the vehicle coordinate system to obtain multiple frames of filtered point clouds; convert the multiple frames of filtered point clouds into multiple frames of point clouds in the world coordinate system according to the vehicle position information; perform clustering on the multiple frames of point clouds in the world coordinate system to obtain multiple frames of clustered point clouds; and determine the point cloud corresponding to the stationary object based on the multiple frames of clustered point clouds.

[0171] Optionally, the vehicle-mounted environmental sensor is a 4D millimeter-wave radar.

[0172] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0173] This disclosure also provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the steps of the obstacle detection method provided in this disclosure.

[0174] Figure 6 This is a block diagram illustrating a vehicle 600 according to an exemplary embodiment. For example, vehicle 600 can be a hybrid vehicle, a non-hybrid vehicle, an electric vehicle, a fuel cell vehicle, or other types of vehicle. Vehicle 600 can be an autonomous vehicle, a semi-autonomous vehicle, or a non-autonomous vehicle.

[0175] Reference Figure 6 The vehicle 600 may include various subsystems, such as an infotainment system 610, a perception system 620, a decision control system 630, a drive system 640, and a computing platform 650. The vehicle 600 may also include more or fewer subsystems, and each subsystem may include multiple components. Furthermore, each subsystem and each component of the vehicle 600 can be interconnected via wired or wireless means.

[0176] In some embodiments, the infotainment system 610 may include a communication system, an entertainment system, and a navigation system, etc.

[0177] The perception system 620 may include several sensors for sensing information about the environment surrounding the vehicle 600. For example, the perception system 620 may include a global positioning system (which may be GPS, BeiDou, or other positioning systems), an inertial measurement unit (IMU), lidar, millimeter-wave radar, ultrasonic radar, and a camera device.

[0178] The decision control system 630 may include a computing system, a vehicle controller, a steering system, a throttle, and a braking system.

[0179] The drive system 640 may include components that provide powered motion to the vehicle 600. In one embodiment, the drive system 640 may include an engine, an energy source, a transmission system, and wheels. The engine may be one or a combination of internal combustion engines, electric motors, and compressed air engines. The engine is capable of converting energy provided by the energy source into mechanical energy.

[0180] Some or all of the functions of vehicle 600 are controlled by computing platform 650. Computing platform 650 may include at least one processor 651 and memory 652, processor 651 can execute instructions 653 stored in memory 652.

[0181] Processor 651 can be any conventional processor, such as a commercially available CPU. Processors may also include graphics processing units (GPUs), field-programmable gate arrays (FPGAs), systems-on-chips (SoCs), application-specific integrated circuits (ASICs), or combinations thereof.

[0182] The memory 652 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0183] In addition to instruction 653, memory 652 can also store data, such as road maps, route information, vehicle position, direction, speed, and other data. The data stored in memory 652 can be used by computing platform 650.

[0184] In this embodiment of the present disclosure, the processor 651 may execute instructions 653 to complete the obstacle detection method described above.

[0185] In another exemplary embodiment, a computer program product is also provided, the computer program product comprising a computer program executable by a programmable device, the computer program having a code portion for performing the obstacle detection method described above when executed by the programmable device.

[0186] Those skilled in the art will also understand that the various illustrative logical blocks and steps listed in the embodiments of this application can be implemented by electronic hardware, computer software, or a combination of both. Whether such functionality is implemented through hardware or software depends on the specific application and the overall system design requirements. Those skilled in the art can implement the described functionality using various methods for each specific application, but such implementation should not be construed as exceeding the scope of protection of the embodiments of this application.

[0187] Furthermore, the term “exemplary” is used herein to mean serving as an example, instance, or illustration. Any aspect or design described herein as “exemplary” is not necessarily to be construed as advantageous compared to other aspects or designs. Rather, the use of the term “exemplary” is intended to present the concept in a concrete manner. As used herein, the term “or” is intended to mean an inclusive “or” rather than an exclusive “or.” That is, unless otherwise specified or clear from the context, “X applies A or B” is intended to mean any of the natural inclusive arrangements. That is, “X applies A or B” satisfies any of the foregoing instances if X applies A; X applies B; or both X applies A and B. Additionally, unless otherwise specified or clear from the context to refer to the singular form, the articles “a” and “an” as used in this application and the appended claims are generally understood to mean “one or more.”

[0188] Similarly, although this disclosure has been shown and described with respect to one or more implementations, equivalent variations and modifications will occur to those skilled in the art upon reading and understanding this specification and the accompanying drawings. This disclosure includes all such modifications and variations and is limited only by the scope of the claims. In particular, with respect to the various functions performed by the components described above (e.g., elements, resources, etc.), unless otherwise indicated, the terminology used to describe such components is intended to correspond to any component (functionally equivalent) that performs the specific function of the described component, even if structurally not equivalent to the disclosed structure. Furthermore, although specific features of this disclosure may have been disclosed with respect to only one of several implementations, such features may be combined with one or more other features of other implementations, as may be desired and advantageous to any given or particular application. Moreover, with regard to the terms “comprising,” “owning,” “having,” “having,” or variations thereof as used in the detailed description or claims, such terms are intended to be inclusive in a manner similar to the term “including.”

[0189] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

[0190] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A method for detecting obstacles, characterized in that, include: Acquire target point clouds of the area surrounding the vehicle detected by onboard environmental sensors; By performing grid projection on the target point cloud, multiple types of grids are determined. Each type of grid includes the projected point cloud corresponding to the target point cloud. Each type of grid uses a different grid unit division method. Based on the projected point clouds included in the various types of grids, the target obstacle information is determined.

2. The detection method according to claim 1, characterized in that, The target point cloud is a three-dimensional point cloud in the world coordinate system. The process involves performing a raster projection on the target point cloud to determine various types of raster data, including: The target point cloud is projected onto a two-dimensional plane in the world coordinate system to obtain the projected point cloud corresponding to the target point cloud in the two-dimensional plane. The first type of grid is determined based on the first type of grid unit in the two-dimensional plane that includes the projected point cloud; A second type of grid is determined based on the second type of grid cells in the two-dimensional plane that include the projected point cloud. The second type of grid cells and the first type of grid cells are obtained by using different grid cell division methods in the two-dimensional plane.

3. The detection method according to claim 1, characterized in that, The step of determining target obstacle information based on the projected point clouds included in the various types of grids includes: For any of the various types of graticles, determine the number of projected point clouds included in each graticle cell of that graticle. Based at least on the number of projected point clouds included in each grid cell of the grid, a target grid cell is determined from each grid cell of the grid, wherein the target grid cell includes the projected point cloud corresponding to the obstacle. Based on the projected point cloud included in the target grid unit, determine the obstacle information corresponding to this type of grid; Based on the obstacle information corresponding to the various types of grids, the target obstacle information is determined.

4. The detection method according to claim 3, characterized in that, The detection method further includes: Obtain the historical projected point cloud information corresponding to each grid cell of this type of raster; The step of determining the target grid cell from each grid cell of the grid, based at least on the number of projected point clouds included in each grid cell of the grid, includes: Based on the number of projected point clouds included in each grid cell of this type of grid and the corresponding historical projected point cloud information, the target grid cell is determined from each grid cell of this type of grid.

5. The detection method according to claim 4, characterized in that, The historical projection point cloud information corresponding to each grid cell includes: the cumulative number of projection point clouds and the cumulative number of projection point clouds. The cumulative number of projection point clouds is obtained by accumulating the number of historical projection point clouds included in the grid cell, and the cumulative number of projection point clouds is obtained by accumulating the number of times the grid cell includes historical projection point clouds.

6. The detection method according to claim 5, characterized in that, The step of determining the target grid cell from each grid cell of the grid based on the number of projected point clouds included in each grid cell and the corresponding historical projected point cloud information includes: Based on the number of projected point clouds included in each grid cell of the grid and the cumulative number of projected point clouds corresponding to each grid cell, the cumulative number of projected point clouds corresponding to each grid cell is updated to obtain the updated cumulative number of projected point clouds. Update the cumulative count of the projected point cloud corresponding to each grid cell to obtain the updated cumulative count of the projected point cloud. The confidence level of each grid cell is determined based on the cumulative number of the updated projected point cloud corresponding to each grid cell. The target raster cell is determined from the raster cells of this type of raster based on the confidence level of each raster cell and the cumulative number of times the updated projected point cloud is accumulated for each raster cell.

7. The detection method according to claim 6, characterized in that, The step of updating the cumulative number of projected point clouds for each grid cell based on the number of projected point clouds included in each grid cell and the cumulative number of projected point clouds corresponding to each grid cell, to obtain the updated cumulative number of projected point clouds, includes: Determine the logarithm of the number of projected point clouds included in each grid cell; The updated cumulative number of projected point clouds is obtained by adding the logarithm of the number of projected point clouds included in each grid cell to the cumulative number of projected point clouds corresponding to each grid cell.

8. The detection method according to claim 3, characterized in that, The projected point cloud is a point cloud in world coordinates. The step of determining the obstacle information corresponding to the grid based on the projected point cloud included in the target grid unit includes: The projected point cloud comprising the target grid cell is converted from the world coordinate system to the obstacle point cloud in the vehicle coordinate system; Based on the obstacle point cloud, determine the size information of the obstacle; Based on the obstacle point cloud and vehicle position information, the relative position information between the obstacle and the vehicle is determined; Based on the size information and the relative position information, the obstacle information corresponding to this type of grid is determined.

9. The detection method according to claim 8, characterized in that, The target point cloud is the point cloud corresponding to a static object, and the detection method further includes: The adjacent target raster cells, including the projected point clouds corresponding to the same static object, are merged to obtain the merged target raster cells. Accordingly, the step of converting the projected point cloud comprising the target grid cell from the world coordinate system to the obstacle point cloud in the vehicle coordinate system includes: The projected point cloud of the target grid cell after merging is converted from the world coordinate system to the obstacle point cloud in the vehicle coordinate system.

10. The detection method according to claim 1, characterized in that, The acquisition of the target point cloud of the area surrounding the vehicle detected by the onboard environmental sensors includes: Acquire vehicle location information and multiple frames of raw point cloud data of the area surrounding the vehicle detected by the on-board environmental sensors. Based on the multiple frames of original point cloud and the vehicle position information, determine the point cloud corresponding to the stationary object; The point cloud corresponding to the static object is determined as the target point cloud.

11. The detection method according to claim 10, characterized in that, The step of determining the point cloud corresponding to the stationary object based on the multiple frames of original point cloud and the vehicle position information includes: The original point clouds of the multiple frames are converted into point clouds of the multiple frames in the vehicle coordinate system; The point cloud of multiple frames in the vehicle coordinate system is filtered to obtain the filtered point cloud. Based on the vehicle location information, the multi-frame filtered point cloud is converted into a multi-frame point cloud in the world coordinate system. Clustering is performed on the point clouds of multiple frames in the world coordinate system to obtain the point clouds after multi-frame clustering; Based on the point cloud after multi-frame clustering, determine the point cloud corresponding to the static object.

12. The detection method according to claims 1-11, characterized in that, The vehicle-mounted environmental sensor is a 4D millimeter-wave radar.

13. An obstacle detection device, characterized in that, include: The acquisition module is configured to acquire target point clouds of the area surrounding the vehicle detected by onboard environmental sensors; The determination module is configured to determine multiple types of grids by performing grid projection on the target point cloud. The multiple types of grids respectively include the projected point cloud corresponding to the target point cloud, and the multiple types of grids adopt different grid unit division methods. The detection module is configured to determine target obstacle information based on the projected point clouds included in the various types of grids.

14. A vehicle, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to execute the executable instructions to implement the obstacle detection method as described in any one of claims 1 to 12.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the obstacle detection method as described in any one of claims 1 to 12.

16. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the obstacle detection method as described in any one of claims 1 to 12.

Images