An obstacle detection method, apparatus, and domain controller based on point clouds.

Abstract

This invention discloses a point cloud-based obstacle detection method, apparatus, and domain controller. The method includes: acquiring a reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path; using the projection of the minimum convex hull of each obstacle's point cloud onto the reference path, determining the minimum rectangular bounding box of each minimum convex hull whose central axis is in the same direction as the reference path; determining the obstacle bounding box of each obstacle based on the minimum rectangular bounding box of each point cloud's minimum convex hull, and using the obstacle bounding box to represent the obstacle. The technical solution provided by this invention improves the accuracy of generating obstacle bounding boxes.

Description

Technical Field

[0001] This invention relates to the field of point cloud data processing, and specifically to a point cloud-based obstacle detection method, apparatus, and domain controller. Background Technology

[0002] As the research, application, and commercialization of autonomous driving technology accelerate, the demand for driving safety technology is also increasing. More closed-scenario autonomous driving systems require the elimination of safety drivers to achieve true driverless operation. In autonomous driving systems, obstacle detection is primarily handled by sensors such as LiDAR, cameras, millimeter-wave radar, and hypersonic sensors. LiDAR obtains depth information about obstacles around the vehicle. The detection results are tracked and fused to obtain the final detection result, which is then sent to the prediction module. The prediction module predicts a future trajectory based on obstacle information on the road. To determine the shape, size, and location of the obstacle's convex hull, the common practice is to compress the obstacle point cloud from three dimensions into two dimensions, then frame it into a two-dimensional convex hull and publish it. The robot plans an obstacle avoidance path based on the obstacle's two-dimensional convex hull information.

[0003] The patent "A Method for Convex Hull Segmentation of Obstacle Point Clouds Based on RANSAC and Corner Extraction" (CN114119940A) proposes a minimum convex hull method to calculate the minimum polygon bounding box of an obstacle. The specific steps include: S1. Performing at least one downsampling and dimensionality reduction on the obstacle point cloud to obtain a preprocessed two-dimensional point cloud; S2. Performing multiple RANSAC line point cloud extractions and Euclidean clustering on the two-dimensional point cloud, traversing each point cloud obtained by Euclidean clustering, and extracting corner points; S3. Determining the centroid based on the corner points of the two-dimensional point cloud, simultaneously determining the polar angle of each corner point, and sorting the corner points clockwise / counterclockwise according to the corresponding polar angles; S4. Traversing the sorted corner points, judging concave points, and publishing the two-dimensional convex hull until all corner points have been traversed; S5. Performing convex hull segmentation based on concave and convex points to obtain the minimum convex hull of the obstacle's point cloud. S6. In order to be suitable for obstacle tracking and fusion in algorithms such as path planning, the minimum rectangle is calculated based on the minimum convex hull, the irregular minimum convex hull is converted into a regular rectangle, the obstacle bounding box is obtained, and the obstacle is represented by the obstacle bounding box.

[0004] However, point clouds do not completely enclose obstacles. The larger the obstacle, the more likely the point cloud will become sparse, resulting in the loss of some obstacle locations. Therefore, the larger the obstacle and the larger the convex hull size, the greater the error in calculating the minimum rectangle using the minimum convex hull. This leads to a larger error in enveloping the obstacle's convex hull using the minimum rectangle, resulting in a larger error in the obstacle's bounding box, and consequently, inaccurate trajectory planning for autonomous vehicles. Another strategy in the industry is to use deep learning methods to detect obstacle bounding boxes, but this method requires the collection and annotation of a large amount of data, and its detection performance is not ideal for some obstacle types that have not been collected. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a point cloud-based obstacle detection method, apparatus, and domain controller, which improves the accuracy of generating obstacle bounding boxes.

[0006] According to a first aspect, embodiments of the present invention provide a point cloud-based obstacle detection method, the method comprising: obtaining a reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path; using the projection of the minimum convex hull of the point cloud of each obstacle onto the reference path, determining the minimum rectangular border of each minimum convex hull of the point cloud with the same central axis as the direction of the reference path; determining the obstacle border of each obstacle based on the minimum rectangular border of each minimum convex hull of the point cloud, and using the obstacle border to represent the obstacle.

[0007] Optionally, determining the minimum rectangular bounding box of each point cloud minimum convex hull with the same central axis as the reference path by projecting the minimum convex hull of each obstacle onto the reference path includes: determining the path vector and path heading angle of the reference path based on the start coordinates and end coordinates of the reference path; constructing multiple convex hull vectors from the start point of the reference path to each convex hull vertex using the vertex coordinates of each current point cloud minimum convex hull and the start coordinates of the reference path; calculating the projection of each convex hull vector onto the path vector, and calculating the distance from each convex hull vertex to the path vector; determining the bounding box length and bounding box width based on the calculated projections and distances, wherein the bounding box length and bounding box width are used to characterize the planar length and planar width of the current minimum rectangular bounding box with the same central axis as the reference path direction, and the current minimum rectangular bounding box is the minimum rectangular bounding box of the current point cloud minimum convex hull; determining the center coordinates used to characterize the center of the current minimum rectangular bounding box based on the bounding box length, the bounding box width, the path heading angle, and the start coordinates of the reference path; and representing the current minimum rectangular bounding box of the current point cloud minimum convex hull using the bounding box length, the bounding box width, and the center coordinates.

[0008] Optionally, determining the border length and border width based on the calculated projection and distance includes: extracting the longest projection, shortest projection, longest distance, and shortest distance from the calculated projection and distance; determining the border length using the difference between the longest projection and the shortest projection; and determining the border width using the difference between the longest distance and the shortest distance.

[0009] Optionally, determining the center coordinates representing the center of the current smallest rectangular border based on the border length, the border width, the path heading angle, and the starting coordinates of the reference path includes:

[0010] The center coordinates of the current smallest rectangle's border are calculated using the following formula.

[0011]

[0012] In the formula, Wc_j and Lc_ are half the length and half the width of the current smallest rectangle, respectively, and x c_j , These are the X and Y coordinates of the center of the current smallest rectangular border on the ground plane, respectively. The center coordinates are determined by x... c_j , constitute, is the heading angle of the path, and x_start,_ are the X and Y coordinates of the starting point on the ground plane, respectively.

[0013] Optionally, determining the obstacle border of each obstacle based on the minimum rectangular border of the minimum convex hull of each point cloud includes: obtaining the coordinates of the highest and lowest points in space from the point cloud data of the current obstacle; determining the border height based on the difference between the coordinates of the highest and lowest points in space, wherein the border height is used to characterize the spatial height of the current obstacle border outside the ground; and determining the current obstacle border using the current minimum rectangular border and the border height.

[0014] Optionally, representing the current obstacle border using the current minimum rectangular border and the border height includes: determining the height of the center coordinates using the border height and the starting point coordinates; and representing the current obstacle border based on the determined center coordinates, the border length, the border width, and the border height.

[0015] Optionally, obtaining the minimum convex hull of the point cloud corresponding to each obstacle near the reference path includes: filtering the point cloud outside the reference path and downsampling the point cloud within the reference path using voxel filtering; segmenting the downsampled point cloud into ground point cloud and non-ground point cloud; clustering the segmented non-ground point cloud to obtain the point cloud corresponding to each obstacle; and calculating the minimum convex hull of the point cloud corresponding to each obstacle based on the Graham scan method.

[0016] According to a second aspect, embodiments of the present invention provide a point cloud-based obstacle detection device, the device comprising: a data acquisition module, configured to acquire a reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path; a rectangular border determination module, configured to determine the minimum rectangular border of each minimum convex hull of the point cloud with the same central axis as the reference path by using the projection of the minimum convex hull of the point cloud of each obstacle onto the reference path; and an obstacle border determination module, configured to determine the obstacle border of each obstacle based on the minimum rectangular border of the minimum convex hull of each point cloud, and to represent the obstacle using the obstacle border.

[0017] According to a third aspect, embodiments of the present invention provide a domain controller, including: a perception processing unit, a decision processing unit, a control processing unit, and a communication unit, wherein the perception processing unit, the decision processing unit, the control processing unit, and the communication unit are communicatively connected to each other, the perception processing unit stores computer instructions, and the perception processing unit executes the first aspect, or any optional embodiment of the first aspect, by executing the computer instructions.

[0018] According to a fourth aspect, embodiments of the present invention provide a computer-readable storage medium storing computer instructions for causing the computer to perform the method described in the first aspect, or any alternative embodiment of the first aspect.

[0019] The technical solution provided in this application has the following advantages:

[0020] The technical solution provided in this application obtains a reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path. Then, it calculates the projection of the minimum convex hull of each obstacle's point cloud onto the reference path, thereby generating the minimum rectangular bounding box of each point cloud's minimum convex hull based on the projection. The central axis of the generated minimum rectangular bounding box is in the same direction as the reference path, i.e., the minimum rectangular bounding box with the same orientation as the reference path is determined based on the reference path direction, thus improving the accuracy of the minimum rectangular bounding box of the point cloud's minimum convex hull. Then, based on the minimum rectangular bounding box of each point cloud's minimum convex hull, the obstacle bounding box of each obstacle is determined to improve the accuracy of representing obstacles using the obstacle bounding box. Attached Figure Description

[0021] The features and advantages of the invention will be more clearly understood by referring to the accompanying drawings, which are schematic and should not be construed as limiting the invention in any way. In the drawings:

[0022] Figure 1 A schematic diagram of the minimum convex hull of a point cloud in the prior art is shown;

[0023] Figure 2 A schematic diagram of the minimum rectangular bounding box of the minimum convex hull of a point cloud in the prior art is shown;

[0024] Figure 3 The diagram illustrates the steps of a point cloud-based obstacle detection method according to one embodiment of the present invention.

[0025] Figure 4 A schematic diagram of the minimum rectangular bounding box of a point cloud-based obstacle detection method according to one embodiment of the present invention is shown.

[0026] Figure 5 This diagram illustrates the principle of a point cloud-based obstacle detection method for generating a minimum rectangular bounding box in one embodiment of the present invention.

[0027] Figure 6 A schematic diagram of a point cloud-based obstacle detection device is shown in one embodiment of the present invention;

[0028] Figure 7 A schematic diagram of a domain controller according to one embodiment of the present invention is shown. Detailed Implementation

[0029] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] like Figure 1 As shown, in existing technologies, point cloud detection techniques may suffer from sparse point clouds. Figure 2 As shown, if the point cloud of the obstacle is sparse, the minimum convex hull is identified based on the point cloud, and then the minimum bounding box of the minimum convex hull is determined. This is not necessarily the actual obstacle bounding box, which is especially obvious when the obstacle is large.

[0031] Based on this, such as Figure 3 and Figure 4As shown, in one embodiment, this invention provides a point cloud-based obstacle detection method, the method comprising:

[0032] Step S101: Obtain the reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path.

[0033] Step S102: Using the projection of the minimum convex hull of the point cloud of each obstacle onto the reference path, determine the minimum rectangular bounding box of the minimum convex hull of each point cloud with the same central axis as the reference path direction.

[0034] Step S103: Determine the obstacle bounding box of each obstacle based on the minimum rectangular bounding box of the minimum convex hull of each point cloud, and use the obstacle bounding box to represent the obstacle.

[0035] Specifically, to reduce the errors caused by directly using the minimum rectangular bounding box of the minimum convex hull of the point cloud to represent obstacles, this embodiment of the invention first obtains the reference path of the vehicle and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path. Then, the minimum convex hull of the point cloud corresponding to each obstacle is projected onto the reference path. According to the projection direction, a minimum rectangular bounding box with the same direction as the reference path is created for each minimum convex hull of the point cloud. From a visual perspective, one of the central axes of this rectangular bounding box is parallel to the reference path. Since obstacles on both sides of the road are usually placed along the direction of the reference path, for example, an empty car without passengers will be parked on the side of the road along the direction of the reference path, based on this assumption, the generated minimum rectangular bounding box is more in line with the actual placement of obstacles. The minimum rectangular bounding box generated in this way is more likely to be more accurate than the minimum rectangular bounding box generated directly in the free direction, and can cover the obstacle outline as much as possible. Thus, the obstacle bounding box of each obstacle is determined by the minimum rectangular bounding box with the same central axis as the reference path, improving the obstacle detection accuracy and thus improving the accuracy of subsequent vehicle path planning.

[0036] In this embodiment of the invention, the minimum convex hull of the point cloud corresponding to each obstacle near the reference path can be obtained in the following way:

[0037] 1. Filter out point clouds outside the road area based on the high-precision map, and use voxel filtering to downsample the point cloud within the road area. Specifically, 3D point clouds often contain a large amount of redundant data, and direct processing is computationally intensive and time-consuming. Therefore, filtering out point clouds outside the road area and downsampling are important steps to improve the accuracy of the bounding boxes.

[0038] 2. The RANSAC method is used to segment the downsampled point cloud into ground-based point clouds and non-ground-based point clouds. Specifically, when planning a vehicle's path, the obstacles the vehicle avoids originate from space, rather than obstacles attached to the ground, thus separating the ground-based point cloud from the non-ground-based point cloud to further reduce redundant data.

[0039] 3. For the segmented non-ground point clouds, Euclidean clustering is used to obtain the corresponding point cloud for each obstacle. Specifically, the clustering method groups point cloud data that are close together to determine the approximate location of each obstacle.

[0040] 4. The minimum convex hull of the point cloud corresponding to each obstacle is calculated using the Graham scan method. By processing the point cloud corresponding to each obstacle using the Graham scan method, the minimum convex hull of the point cloud that more closely resembles the true contour and area of ​​the obstacle is obtained.

[0041] Since the area enclosed by the two-dimensional convex hull is closer to the actual area of ​​the obstacle, but it is not consistent with other obstacle detection methods, it is not suitable for obstacle tracking and fusion. Therefore, it is necessary to generate a more accurate minimum rectangular bounding box for the minimum convex hull of each point cloud based on the method provided in the embodiments of the present invention.

[0042] Specifically, in one embodiment, step S102 above includes the following steps:

[0043] Step 1: Determine the path vector and path heading angle of the reference path based on the starting and ending coordinates of the reference path.

[0044] Step 2: Construct multiple convex hull vectors from the starting point of the reference path to each convex hull vertex using the coordinates of each vertex of the current point cloud's minimum convex hull and the starting point coordinates of the reference path.

[0045] Step 3: Calculate the projection of each convex hull vector onto the path vector, and calculate the distance from each convex hull vertex to the path vector.

[0046] Step 4: Determine the bounding box length and width based on the calculated projection and distance. The bounding box length and width are used to characterize the planar length and planar width of the current minimum rectangular bounding box with the same direction as the central axis and the reference path, respectively. The current minimum rectangular bounding box is the minimum rectangular bounding box of the current point cloud minimum convex hull.

[0047] Step 5: Based on the border length, border width, path heading angle, and the starting coordinates of the reference path, determine the center coordinates used to represent the center of the current smallest rectangle border.

[0048] Step 6: Use the border length, border width, and center coordinates to represent the current minimum rectangular border of the minimum convex hull of the current point cloud.

[0049] Specifically, such as Figure 5 As shown, to calculate the minimum rectangular bounding box of the minimum convex hull of each obstacle point cloud, first determine the path vector and path heading angle of the reference path based on the starting and ending coordinates of the reference path. The steps are as follows:

[0050] The navigation module obtains the current reference path of the autonomous vehicle; the coordinates of the nearest point on the reference path from the current navigation position (x, y, z) of the autonomous vehicle are calculated, which is the starting point (x_start, y_start). Figure 5 As shown ( Figure 5 Only the X and Y coordinates in the geodetic coordinate system are retained, i.e., the top view of the reference path, to facilitate the calculation of the minimum rectangular bounding box of the minimum convex hull of the point cloud; the coordinate point (x_end, y_end) at a distance of a set threshold from the starting point in the forward direction of the reference path is calculated as the endpoint. If the threshold is greater than the length of the reference path, the last coordinate point on the reference path is taken; the vector v1 of the reference path is obtained as (x_end – x_start, y_end – y_start), and this vector is used to represent the reference path.

[0051] Calculate the heading angle of the reference path

[0052] Then, for each minimum convex hull of the point cloud, connecting the starting point of the reference path to each vertex of the minimum convex hull yields multiple convex hull vectors. Each convex hull vector represents a vertex of the convex hull, with the starting point of the reference path pointing to a vertex. Each convex hull vector is represented by subtracting the coordinates of the starting point of the reference path from the coordinates of the convex hull vertex. Taking a minimum convex hull of a point cloud as an example... Figure 5 The vector sp in the vector is a convex hull vector of the minimum convex hull of the current point cloud.

[0053] Then, calculate the projection of each convex hull vector of the current point cloud's minimum convex hull onto the path vector, and calculate the distance from each vertex of the current point cloud's minimum convex hull to the path vector. For example... Figure 5 In this context, L is the projection of a convex hull vector sp onto the reference path. Figure 5 In this context, W represents the distance from the vertex of the convex hull corresponding to the convex hull vector sp to the path vector. The specific calculation formula is as follows:

[0054]

[0055]

[0056] In the formula, i represents the i-th vertex of the minimum convex hull of the j-th point cloud, and L ij sp represents the projection of the i-th convex hull vector of the j-th minimum convex hull of the point cloud. ij Let sp be the vector of the i-th convex hull of the j-th point cloud with minimum convex hull. ij[a] a = 1 represents the first coordinate value of the vector, a = 2 represents the second coordinate value of the vector, v1 represents the reference path vector, v1[b], b = 1 represents the first coordinate value of the vector, b = 2 represents the second coordinate value of the vector, and W_ij represents the distance from the i-th vertex of the minimum convex hull of the j-th point cloud to the reference path.

[0057] To determine the projection of the current point cloud's minimum convex hull onto the reference path, this embodiment of the invention extracts the longest projection Lmax and the shortest projection Lmin from the calculated projections. Then, it calculates the difference between the longest and shortest projections. This difference represents the projection of the current point cloud's minimum convex hull onto the reference path, and its magnitude is the border length of the currently generated minimum rectangular bounding box. Similarly, this embodiment extracts the longest distance Wmax and the shortest distance Wmin from the calculated distances. Then, it calculates the difference between the longest and shortest distances. This difference represents the border width of the current point cloud's minimum convex hull, which generates the current minimum rectangular bounding box. The orientation of this border is consistent with the orientation of the reference path.

[0058] In addition, the center coordinates representing the center of the current smallest rectangle's border are calculated based on the border length, border width, path heading angle, and the starting coordinates of the reference path. The calculation formula is as follows:

[0059]

[0060] In the formula, Wc_j and Lc_ are half the length and half the width of the current smallest rectangle, respectively, and x c_j , These are the X and Y coordinates of the center of the current smallest rectangular bounding box on the ground plane, respectively, and the center coordinates are determined by x... c_j , Composed of, by path heading angle The above formula is used to calculate that x_start and _ are the X and Y coordinates of the starting point on the Earth's surface, respectively.

[0061] Finally, a minimum rectangular bounding box is represented by a set of center coordinates, bounding box length, and bounding box width. Using the minimum rectangular bounding box to ultimately determine the obstacle bounding box in space, or directly using the minimum rectangular bounding box as the obstacle bounding box in the top view, improves the accuracy of obstacle detection, and thus improves the accuracy of vehicle path planning.

[0062] Specifically, in one embodiment, step S103 above includes the following steps:

[0063] Step 7: Obtain the coordinates of the highest and lowest points in space from the point cloud data of the current obstacles.

[0064] Step 8: Determine the border height based on the difference between the coordinates of the highest point and the lowest point in space. The border height is used to characterize the spatial height of the current obstacle border outside the ground.

[0065] Step 9: Determine the current obstacle border using the current smallest rectangle border and border height.

[0066] Specifically, to further improve the accuracy of obstacle borders, considering that obstacles include not only the length and width of the ground plane but also the height in spatial dimension, this embodiment of the invention also obtains the coordinates of the highest and lowest points in space from the point cloud data of the current obstacle. The difference between the coordinates of the highest and lowest points in space is used to determine the border height, thus introducing height information to the obstacle border. Furthermore, in this embodiment, the center coordinates are updated using the border height and the starting point coordinates, introducing height to the center coordinates of the rectangular border. The calculation method is as follows:

[0067]

[0068] In the formula, z is the height component of the reference path starting point coordinates, and H_max_j and H_min_j represent the coordinates of the highest and lowest points in the current obstacle point cloud data, respectively. This represents half the border height.

[0069] Finally, the center coordinates of the determined height, border length, border width, and border height are used as a set of information to represent the current obstacle border. This set of information for each obstacle is then shared with the path planning module, enabling more accurate path planning in the future.

[0070] Through the above steps, the technical solution provided in this application obtains the reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path; then, it calculates the projection of the minimum convex hull of the point cloud of each obstacle onto the reference path, thereby generating the minimum rectangular bounding box of each point cloud minimum convex hull based on the projection. The central axis of the generated minimum rectangular bounding box is the same as the direction of the reference path, that is, the minimum rectangular bounding box with the same orientation as the reference path is determined based on the direction of the reference path, thereby improving the accuracy of the minimum rectangular bounding box of the point cloud minimum convex hull. Then, based on the minimum rectangular bounding box of each point cloud minimum convex hull, the obstacle bounding box of each obstacle is determined to improve the accuracy of representing obstacles using obstacle bounding boxes.

[0071] like Figure 6 As shown, this embodiment also provides a point cloud-based obstacle detection device, the device comprising:

[0072] The data acquisition module 101 is used to acquire the reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path. For details, please refer to the relevant description of step S101 in the above method embodiment, which will not be repeated here.

[0073] The rectangular bounding box determination module 102 is used to determine the minimum rectangular bounding box of the minimum convex hulls of each point cloud that are aligned with the central axis of the reference path by projecting the minimum convex hulls of each obstacle's point cloud onto the reference path. For details, please refer to the relevant description of step S102 in the above method embodiment, which will not be repeated here.

[0074] The obstacle bounding box determination module 103 is used to determine the obstacle bounding box of each obstacle based on the minimum rectangular bounding box of the minimum convex hull of each point cloud, and to represent the obstacle using the obstacle bounding box. For details, please refer to the relevant description of step S103 in the above method embodiment, which will not be repeated here.

[0075] The obstacle detection device based on point cloud provided in this embodiment of the invention is used to execute the obstacle detection method based on point cloud provided in the above embodiment. Its implementation method and principle are the same. For details, please refer to the relevant description of the above method embodiment, which will not be repeated here.

[0076] Through the collaborative operation of the aforementioned components, the technical solution provided in this application obtains the reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path. Then, it calculates the projection of the minimum convex hull of each obstacle's point cloud onto the reference path, thereby generating the minimum rectangular bounding box of each point cloud's minimum convex hull based on the projection. The central axis of the generated minimum rectangular bounding box is in the same direction as the reference path; that is, the minimum rectangular bounding box with the same orientation as the reference path is determined based on the reference path direction, thus improving the accuracy of the minimum rectangular bounding box of the point cloud's minimum convex hull. Then, based on the minimum rectangular bounding box of each point cloud's minimum convex hull, the obstacle bounding box of each obstacle is determined to improve the accuracy of representing obstacles using the obstacle bounding box.

[0077] Figure 7 This invention illustrates a domain controller according to an embodiment of the present invention. The domain controller includes at least a perception processing unit 901, a decision processing unit 902, a control processing unit 903, and a communication unit 904. The perception processing unit 901, the decision processing unit 902, the control processing unit 903, and the communication unit 904 can communicate with each other via a bus or other means. Figure 7 Taking the bus method as an example.

[0078] In this embodiment, the perception processing unit 901 and the decision processing unit 902 each include an independent processor. The perception processing unit 901 and the decision processing unit 902 may each include an independent memory, or they may use a shared memory.

[0079] In this embodiment of the invention, the perception processing unit 901 is mainly applied to engineering machinery scenarios. Its main function is to perform perception fusion processing on sensor data to obtain environmental information of the current environment of the engineering machinery, and then send the environmental information to the control processing unit 903 or the decision processing unit 902 according to the data type of the environmental information signal.

[0080] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs, non-transitory computer-executable programs, and modules, such as the program instructions / modules corresponding to the methods in the above method embodiments. The perception processing unit 901 executes various functional applications and data processing of the processor by running the non-transitory software programs, instructions, and modules stored in the memory, thereby implementing the methods in the above method embodiments.

[0081] The function of the decision processing unit 902 is to formulate driving or operation strategies by combining information such as the surrounding environment, operation scenario, and vehicle status, and finally issue control commands.

[0082] The main function of the control processing unit 903 is to convert between different types of signals, such as communication protocol conversion (CAN, Ethernet, LIN, etc.), AD conversion (sensor input), and DA conversion (control drive). For example, to convert the signal scanned by the LiDAR into point cloud data, the control processing unit 903 can be an MCU with chips such as the Texas Instruments (TI) TDA4VM, Mobileye's EyeQ series, Renesas' R-CAR H3, or Horizon Robotics' Journey series.

[0083] The main function of the communication unit 904 is to conduct wireless communication, including but not limited to 5G / 4G network communication, Wi-Fi communication, and satellite communication, and to communicate with the cloud server. Its main functions include uploading device-related status and information to the cloud service, requesting the cloud server to assist in calculation and processing, downloading data from the cloud server, and performing OTA software upgrades on the controller. It also communicates with nearby devices, receiving status information from other devices and collaboratively completing tasks. The communication unit 110 of the control module can be a 5G module, Wi-Fi module, Bluetooth module, etc.

[0084] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The implemented program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk drive (HDD), or solid-state drive (SSD), etc.; the storage medium can also include combinations of the above types of memory.

[0085] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A point cloud-based obstacle detection method, characterized in that, The method includes: Obtain the reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path; By using the projection of the minimum convex hull of the point cloud of each obstacle onto the reference path, the minimum rectangular bounding box of the minimum convex hull of each point cloud with the central axis in the same direction as the reference path is determined. The obstacle bounding box of each obstacle is determined based on the minimum rectangular bounding box of the minimum convex hull of each point cloud, and the obstacle is represented by the obstacle bounding box. The step of determining the minimum rectangular bounding box of each point cloud minimum convex hull with the same central axis as the reference path by projecting the minimum convex hull of each obstacle onto the reference path includes: The path vector and path heading angle of the reference path are determined based on the starting coordinates and ending coordinates of the reference path. Construct multiple convex hull vectors from the starting point of the reference path to each convex hull vertex using the coordinates of each vertex of the current minimum convex hull of the point cloud and the starting point coordinates of the reference path. Calculate the projection of each convex hull vector onto the path vector, and calculate the distance from each convex hull vertex to the path vector; The bounding box length and width are determined based on the calculated projection and distance. The bounding box length and width are used to characterize the planar length and planar width of the current minimum rectangular bounding box whose central axis is in the same direction as the reference path. The current minimum rectangular bounding box is the minimum rectangular bounding box of the current point cloud minimum convex hull. Based on the border length, the border width, the path heading angle, and the starting coordinates of the reference path, determine the center coordinates used to characterize the center of the current smallest rectangular border; The minimum rectangular border of the minimum convex hull of the current point cloud is represented by the border length, the border width, and the center coordinates.

2. The method according to claim 1, characterized in that, The determination of border length and border width based on calculated projection and distance includes: Extract the longest projection, shortest projection, longest distance, and shortest distance from the calculated projections and distances, respectively; The border length is determined by the difference between the longest projection and the shortest projection, and the border width is determined by the difference between the longest distance and the shortest distance.

3. The method according to claim 1, characterized in that, The determination of the center coordinates representing the center of the current smallest rectangular border based on the border length, the border width, the path heading angle, and the starting coordinates of the reference path includes: The center coordinates of the current smallest rectangle's border are calculated using the following formula. In the formula, These are half the border length and half the border width of the current smallest rectangle, respectively. , These are the X and Y coordinates of the center of the current smallest rectangular border on the ground plane, respectively. The center coordinates are determined by... , It is the heading angle of the stated path. , These are the X and Y coordinates of the starting point on the Earth's surface.

4. The method according to claim 3, characterized in that, The process of determining the obstacle bounding box of each obstacle based on the minimum rectangular bounding box of the minimum convex hull of each point cloud includes: Obtain the coordinates of the highest and lowest points in space from the point cloud data of the current obstacles; The border height is determined based on the difference between the coordinates of the highest point and the lowest point in the space. The border height is used to characterize the spatial height of the current obstacle border outside the ground. The current obstacle border is determined using the current minimum rectangle border and the border height.

5. The method according to claim 4, characterized in that, The step of representing the current obstacle border using the current minimum rectangular border and the border height includes: The height of the center coordinates is determined using the border height and the starting point coordinates; The current obstacle border is represented by the center coordinates of the determined height, the border length, the border width, and the border height.

6. The method according to claim 1, characterized in that, The step of obtaining the minimum convex hull of the point cloud corresponding to each obstacle near the reference path includes: Filter out point clouds outside the reference path, and use voxel filtering to downsample the point clouds within the reference path; The downsampled point cloud is segmented into ground point cloud and non-ground point cloud; Cluster the non-ground point clouds obtained from the segmentation to obtain the point cloud corresponding to each obstacle; The minimum convex hull of the point cloud corresponding to each obstacle is calculated based on the Graham scan method.

7. An obstacle detection device based on point clouds, characterized in that, The device includes: The data acquisition module is used to acquire the reference path and the minimum convex hull of the point cloud corresponding to each obstacle near the reference path; The rectangular bounding box determination module is used to determine the minimum rectangular bounding box of each point cloud minimum convex hull with the same central axis as the reference path by using the projection of the minimum convex hull of each obstacle's point cloud onto the reference path. The obstacle bounding box determination module is used to determine the obstacle bounding box of each obstacle based on the minimum rectangular bounding box of the minimum convex hull of each point cloud, and to represent the obstacle using the obstacle bounding box. The step of determining the minimum rectangular bounding box of each point cloud minimum convex hull with the same central axis as the reference path by projecting the minimum convex hull of each obstacle onto the reference path includes: The path vector and path heading angle of the reference path are determined based on the starting coordinates and ending coordinates of the reference path. Construct multiple convex hull vectors from the starting point of the reference path to each convex hull vertex using the coordinates of each vertex of the current minimum convex hull of the point cloud and the starting point coordinates of the reference path. Calculate the projection of each convex hull vector onto the path vector, and calculate the distance from each convex hull vertex to the path vector; The bounding box length and width are determined based on the calculated projection and distance. The bounding box length and width are used to characterize the planar length and planar width of the current minimum rectangular bounding box whose central axis is in the same direction as the reference path. The current minimum rectangular bounding box is the minimum rectangular bounding box of the current point cloud minimum convex hull. Based on the border length, the border width, the path heading angle, and the starting coordinates of the reference path, determine the center coordinates used to characterize the center of the current smallest rectangular border; The minimum rectangular border of the minimum convex hull of the current point cloud is represented by the border length, the border width, and the center coordinates.

8. A domain controller, characterized in that, include: The system comprises a perception processing unit, a decision processing unit, a control processing unit, and a communication unit, wherein the perception processing unit, the decision processing unit, the control processing unit, and the communication unit are interconnected and communicate with each other. The perception processing unit stores computer instructions, and the perception processing unit executes the computer instructions to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method as described in any one of claims 1-6.

Images