Obstacle detection methods, devices and storage media

Abstract

This application provides an obstacle detection method, apparatus, and storage medium. The method includes: acquiring point cloud data within the Nth detection sub-range of a lidar in a preset order; the detection range of the lidar in one detection cycle includes M detection sub-ranges; wherein the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M; calculating a confidence level that the Nth detection sub-range includes a preset target object based on the acquired point cloud data within the Nth detection sub-range; and outputting a preset target obstacle recognition result if the confidence level is greater than or equal to a preset threshold. The obstacle detection method provided in this application improves the real-time performance of lidar obstacle detection.

Description

Technical Field

[0001] This application relates to the field of intelligent driving technology, and in particular to an obstacle detection method, device and storage medium. Background Technology

[0002] In the field of artificial intelligence technology, target tracking is a crucial topic in Advanced Driving Assistance Systems (ADAS). Effectively improving the real-time performance of target perception can provide drivers with more timely and reliable information about target objects. This information can then be used to further enhance the safety of ADAS, ultimately achieving autonomous driving. However, current LiDAR systems are limited by scanning methods and cycles, making it difficult to improve real-time performance. Therefore, how to improve the real-time performance and accuracy of target perception is a pressing issue that needs to be addressed. Summary of the Invention

[0003] This application provides an obstacle detection method, device, and storage medium that are not limited by radar scanning methods or scanning cycles, thereby effectively improving the real-time performance of obstacle detection.

[0004] To solve the above technical problems, this application includes the following technical solutions:

[0005] In a first aspect, embodiments of this application provide an obstacle detection method, the method comprising:

[0006] Point cloud data within the Nth detection sub-range of the lidar are acquired according to a preset order; the detection range of one detection cycle of the lidar includes M detection sub-ranges; the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M;

[0007] The confidence level of the Nth detection sub-range including the preset target object is calculated based on the point cloud data obtained in the Nth detection sub-range; if the confidence level is greater than or equal to the preset threshold, the preset target obstacle recognition result is output.

[0008] Secondly, embodiments of this application provide an obstacle detection device, the device comprising:

[0009] The acquisition module is used to acquire point cloud data within the Nth detection sub-range of the lidar in a preset order; the detection range of one detection cycle of the lidar includes M detection sub-ranges; the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M;

[0010] The detection module is used to calculate the confidence level that the Nth detection sub-range includes a preset target object based on the point cloud data acquired in the Nth detection sub-range; if the confidence level is greater than or equal to a preset threshold, the module outputs the preset target obstacle recognition result.

[0011] Thirdly, embodiments of this application provide another obstacle detection device, which includes a processor, a memory, and a communication interface:

[0012] The processor is connected to the memory and the communication interface;

[0013] The memory is used to store executable program code;

[0014] The processor runs a program corresponding to the executable program code stored in the memory to execute the obstacle detection method provided in any of the first aspects.

[0015] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, characterized in that, when executed by a processor, the program implements the obstacle detection method provided in any of the first aspects above.

[0016] This application provides an obstacle detection method. By changing the output mode of the lidar, the detection range of the lidar is divided into multiple sub-ranges, and the multiple sub-ranges are detected in a preset order. The confidence of the multiple sub-ranges is acquired and accumulated in real time until the detection confidence exceeds the preset confidence, at which point the obstacle detection result is output. The obstacle detection method provided in this application shortens the detection time of each sub-unit by dividing the detection cycle, thereby improving the real-time performance of obstacle detection. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of an obstacle detection scenario provided in an embodiment of this application;

[0019] Figure 2 This is a schematic diagram of a lidar installation location provided in an embodiment of this application;

[0020] Figure 3 This is a schematic flowchart of an obstacle detection method provided in an embodiment of this application;

[0021] Figure 4 This is a schematic diagram of a solid-state lidar application provided in an embodiment of this application;

[0022] Figure 5 This is a schematic diagram of a rotating mechanical lidar application provided in an embodiment of this application;

[0023] Figure 6 This is a schematic diagram of point cloud data detected when there is an overlapping part in the detection sub-range, as provided in an embodiment of this application;

[0024] Figure 7 This is a schematic diagram of scanning point cloud data during obstacle detection using a solid-state lidar, as provided in an embodiment of this application.

[0025] Figure 8 This is a flowchart illustrating another obstacle detection method provided in an embodiment of this application;

[0026] Figure 9 This is a schematic diagram of an obstacle detection device provided in an embodiment of this application;

[0027] Figure 10 This is a schematic diagram of another obstacle detection device provided in an embodiment of this application. Detailed Implementation

[0028] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the specific embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0029] The terms "first," "second," "third," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or apparatuses.

[0030] Please refer to Figure 1 and Figure 2 As shown, Figure 1 This is a schematic diagram of an obstacle detection scenario provided in an embodiment of this application. The obstacle detection scenario includes an obstacle 10 and a vehicle 20. The vehicle 20 is equipped with a LiDAR 201, such as... Figure 2 As shown, Figure 2 This is a schematic diagram of a lidar installation location provided in an embodiment of this application.

[0031] Obstacle 10 may include, but is not limited to, pedestrians, vehicles, animals, trees, buildings, or objects with speed that obstruct the normal movement of the vehicle. In this embodiment, obstacle 10 can be detected by lidar 201, and the detection results can be obtained and displayed on the vehicle terminal. The detection results of obstacle 10 may include, but are not limited to, point cloud data of obstacle 10, obstacle category identification results, speed information, distance information, and warning images or sound information.

[0032] Vehicle 20 may include cars, buses, semi-trailers, off-road vehicles, special-purpose vehicles, freight trucks, tractor-trailers, dump trucks, or any other type of vehicle; there is no limitation on the type, model, or type of vehicle 20. In this embodiment, the vehicle can travel on roads that do not impede traffic, such as intersections, crossroads, and highways. Vehicle 20 can detect obstacles within a preset range using lidar 201 and display the detection results on the vehicle-mounted terminal.

[0033] The lidar 201 may include lidar of the Micro-Electro-Mechanical System (MEMS) solid-state type, rotating mechanical scanning lidar, etc., etc., in this embodiment of the application. The autonomous driving perception system may consist of a lidar 201, wherein, as... Figure 2 As shown, when the lidar 201 collects point cloud data, it can be installed at position A. It is understood that the lidar 201 can also be installed at position B. In this embodiment, the specific installation location of the lidar 201 is not limited. It is understood that in this embodiment, as... Figure 2 The autonomous driving perception system can also be composed of multiple LiDARs. When the autonomous driving perception system is composed of multiple LiDARs, the specific number of LiDARs included in the system is not limited. LiDAR 201 can be any single LiDAR in the autonomous driving perception system; this application does not limit the specific composition of the autonomous driving perception system.

[0034] Next, we will combine Figure 1 and Figure 2 The schematic diagram of the obstacle detection scenario and the schematic diagram of the lidar installation shown illustrate the obstacle detection method provided in the embodiments of this application.

[0035] Please refer to Figure 3 As shown, Figure 3 This is a flowchart illustrating an obstacle detection method according to an embodiment of this application. The method includes:

[0036] S301. Obtain point cloud data within the Nth detection sub-range of the lidar according to a preset order; the detection range of one detection cycle of the lidar includes M detection sub-ranges; the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M.

[0037] The preset order can be the scanning order of the lidar.

[0038] Specifically, different types of LiDARs use different scanning methods for point cloud data. MEMS LiDARs scan using the simple harmonic vibration of a galvanometer, so the optical path scanning sequence can be, for example, a slow axis from top to bottom and a fast axis from left to right, creating a scanning field of view. Another example is a mechanical LiDAR, which uses a mechanical drive to rotate the optical system 360 degrees to achieve scanning, creating a cylindrical detection area centered on the LiDAR. Different LiDARs acquire point cloud data from the Nth detection sub-range in a preset order. The detection range of one LiDAR detection cycle includes M detection sub-ranges. The Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M.

[0039] Furthermore, before the lidar acquires point cloud data within the Nth detection sub-range according to a preset order, the method further includes: determining the division method of the M detection sub-ranges; the division method includes: dividing according to the detection time corresponding to one detection cycle of the lidar, and dividing according to the detection range corresponding to one detection cycle of the lidar.

[0040] Specifically, when dividing the detection time according to one detection cycle of the LiDAR, before acquiring the point cloud data within the Nth detection sub-range of the LiDAR in a preset order, the method further includes: acquiring the detection time corresponding to one detection cycle of the LiDAR, and dividing the detection cycle into M detection sub-cycles according to a preset rule. It can be understood that the detection range corresponding to each sub-cycle is a detection sub-range, and the detection cycle is the detection time for one frame of data. Acquiring the point cloud data within the Nth detection sub-range of the LiDAR in a preset order includes: determining the detection time corresponding to each of the M detection sub-ranges based on the detection time corresponding to one detection cycle of the LiDAR; acquiring the first detection time corresponding to the Nth detection sub-range; and acquiring the point cloud data corresponding to the Nth detection sub-range when the LiDAR reaches the first detection time, according to the scanning order of the LiDAR. The preset rule can be a custom rule, such as an equal division rule, where the detection cycle can be divided into M equal parts. If the detection cycle of the LiDAR is T, then the point cloud data scanned by the LiDAR within T / M time is acquired every T / M time period. It is understood that the preset rules can also be arranged in ascending order of time. For example, if the operating cycle of a mechanical lidar is 100ms, and M is 5, then the operating times corresponding to the 5 detection sub-ranges could be, for example, 30, 25, 20, 15, and 10 seconds respectively. It is also understood that the preset rules can be arranged in ascending order of time. For example, if the operating cycle of a mechanical lidar is 100ms, and M is 5, then the operating times corresponding to the 5 detection sub-ranges could be, for example, 10, 15, 20, 25, and 30 seconds respectively.

[0041] Specifically, when dividing the detection range according to one detection cycle of the lidar, before acquiring the point cloud data within the Nth detection sub-range of the lidar in a preset order, the method further includes: acquiring the detection range corresponding to one detection cycle of the lidar, and dividing the detection range into M sub-detection ranges according to a preset rule. Acquiring the point cloud data within the Nth detection sub-range of the lidar in a preset order includes: determining the detection range corresponding to each of the M detection sub-ranges based on the detection range corresponding to one detection cycle of the lidar; acquiring the first detection range corresponding to the Nth detection sub-range; and acquiring the point cloud data corresponding to the Nth detection sub-range when the lidar scans the first detection range according to the lidar's scanning order. The preset rule may include an equal division rule. It is understood that the preset rule may also be an arrangement of detection ranges from smallest to largest. This application does not impose specific restrictions on the preset rule.

[0042] like Figure 4 The diagram shown Figure 4The vehicle shown is equipped with a MEMS solid-state LiDAR. When dividing the detection range of a MEMS LiDAR, because the fast axis is a reciprocating motion from one direction to another within a cycle, and the slow axis is a linear motion from one direction to another, the detection range is often divided based on the slow axis detection field of view. This LiDAR typically uses a top-to-bottom or bottom-to-top reciprocating scanning mode. Therefore, the detection range is usually divided based on the slow axis detection field of view, i.e., the vertical field of view. For example... Figure 4 As shown in Figure 40, where 40 represents the vertical field of view corresponding to the slow axis of the MEMS LiDAR, if this vertical field of view is -12.5° to 12.5°, and assuming M equals 5, then the MEMS LiDAR forms a detection sub-range every 5° scan along the slow axis. Figure 4 The angle corresponding to the detection sub-ranges of 401, 402, 403, 404, and 405 is 5°.

[0043] like Figure 5 The diagram shown Figure 5 The vehicle shown is equipped with a rotating mechanical scanning LiDAR. Rotating mechanical LiDARs typically transmit and receive multiple lines longitudinally while simultaneously rotating and scanning horizontally, generating parallel lines. For mechanical LiDARs, the detection range corresponding to one rotation cycle is usually divided into detection angles. If the detection range of this type of LiDAR is 360°, according to the equal division rule, every 60° rotation of the mechanical LiDAR forms a detection sub-range; that is, every 60° rotation of the LiDAR acquires point cloud data within one detection sub-range. Figure 5 As shown, the detection range of the lidar can be set to six sub-detection ranges as shown in Figures 501, 502, 503, 504, 505, and 506. At this time, the detection angle corresponding to each sub-detection range is 60°.

[0044] It is understood that, before acquiring the point cloud data within the Nth detection sub-range of the LiDAR according to a preset order, the method further includes: acquiring the environmental information of the LiDAR; determining the scene where the LiDAR is located based on the environmental information; and adjusting the number M of detection sub-ranges included in one detection cycle of the LiDAR based on the scene. The environmental information may include information on the number of traffic lights, intersections, pedestrian crossings and pedestrian numbers, lane lines, toll stations, etc. This application does not limit the content of the environmental information. It is understood that the scene includes intersection scenes, highway scenes, urban road scenes, etc. It is understood that the intersection scene has the highest real-time requirement for the target detection area; therefore, the number M of detection sub-ranges is the largest in the intersection scene, followed by the urban road scene, and lastly the highway scene. It is understood that when the LiDAR acquires point cloud information, it can determine the scene where the LiDAR is located based on the feature information extracted from the point cloud information, and adjust the number M of detection sub-ranges included in one detection cycle of the LiDAR based on the scene.

[0045] It is understood that adjusting the number M of the detection sub-ranges corresponding to one detection cycle of the LiDAR may further include: determining the scene complexity of the scene in which the LiDAR is located; determining whether the scene complexity is greater than a preset threshold; and when the scene complexity is greater than the preset threshold, adjusting the number M of the detection sub-ranges corresponding to one detection cycle of the LiDAR based on the value of the scene complexity. It is understood that after the LiDAR acquires point cloud information, it can determine the complexity of the scene in which the LiDAR is located based on the number of obstacles identified in the point cloud information, the position of the obstacles, the speed information of the obstacles, etc., and adjust the number M of the detection sub-ranges included in one detection cycle of the LiDAR based on the scene complexity.

[0046] It is understood that the lidar can adjust the number of detection sub-ranges M according to changes in the scene, changes in scene complexity, or simultaneous changes in both the scene and scene complexity, thereby improving the flexibility of obstacle detection and further ensuring the real-time performance and accuracy of obstacle detection. This application does not restrict the order of scene determination and scene complexity determination, or whether they are combined or independent in determining whether to adjust the number of detection sub-ranges M corresponding to a detection cycle.

[0047] It is understood that the higher the complexity of the scenario, the higher the real-time requirement for the target detection area, and the more sub-ranges of the detection in the preset rules.

[0048] Specifically, when the scene complexity within the detection range represented by environmental information exceeds a preset threshold, the number of detection sub-ranges is adjusted from M to P, with the adjusted sub-ranges being smaller than the original sub-ranges. Here, P is a positive integer greater than M. In practical applications, when the detected scene complexity exceeds the preset threshold, it indicates a complex driving environment, such as a large number of pedestrians or vehicles, which negatively impacts the user's driving. In this case, reducing each detection sub-range allows for a more refined division of the detection sub-ranges, facilitating more timely calculation and analysis of point cloud data, improving the real-time performance and accuracy of obstacle detection, and enhancing user driving safety. When the scene complexity within the detection range represented by environmental information is less than a preset threshold, the number of detection sub-ranges is adjusted from M to K, with the adjusted sub-ranges being larger than the original sub-ranges. When the scene complexity is less than the preset threshold, it indicates a simpler driving environment, such as fewer pedestrians or vehicles. In this case, increasing each detection sub-range reduces the power consumption of devices such as LiDAR and vehicle terminals, improving the user experience. Here, K is a positive integer greater than M. The preset threshold for scene complexity can be set manually. For example, if complexity is represented by a range of 0 to 1, and the preset threshold is set to 0.6, then when the scene complexity is greater than 0.6, the number of detection sub-ranges will be adjusted from M to P. It should be noted that this application does not limit the method of characterizing scene complexity within the detection range of the lidar for representing environmental information.

[0049] As an optional approach, the lidar can also divide the detection range within a detection cycle into a general detection range and a target detection range; wherein, the target detection range is a user-preset key detection area, i.e., the ROI region. The target detection region can be preset by the user.

[0050] Optionally, before sequentially acquiring the M detection sub-ranges of the lidar in a preset order, the method further includes: determining the general detection range and the target detection range within the lidar detection range; determining a preset division rule for the M detection sub-ranges based on the general detection range and the target detection range; and dividing the detection range within a detection cycle into M detection sub-ranges based on the preset division rule.

[0051] The method involves determining M sub-ranges of detection based on the general detection range and the target detection range using preset division rules. These rules include: dividing the general detection range into *a* sub-ranges according to a first division rule, and dividing the target detection range into *b* sub-ranges according to a second division rule; where *a* + *b* = *M*, and *a* < *b*. It is understood that the real-time requirements for detection within the target detection range are higher, therefore the division of the detection area within the target detection range is more refined, resulting in a greater number of sub-ranges. Generally, each sub-range within the target detection range is smaller than each sub-range within the general detection range.

[0052] It should be noted that, in the embodiments of this application, during the division of the detection range of the lidar, any detection sub-range can overlap with two adjacent detection sub-ranges in the preset detection sequence. Taking the division of the detection range according to one detection cycle of the lidar as an example, in the case of overlap, the detection angle corresponding to each detection sub-range is greater than the detection angle corresponding to it in the case of no overlap. Furthermore, this application does not limit the division method or the angle of overlap when adjacent detection sub-ranges overlap.

[0053] like Figure 6 The diagram shown is as follows: Figure 4 Taking the MEMS solid-state lidar shown as an example, the detection range of this solid-state lidar includes 5 detection sub-ranges. Figure 6 The diagram shows the detection areas for sub-ranges 401, 402, and 403 when they overlap. Specifically, when there is no overlap and the angles are evenly divided, the detection angle for each of sub-ranges 401, 402, and 403 is 5°. Figure 6 401a, 402a, and 403a are all 5°. When there is overlap and the angles are evenly divided, the overlapping portion of any detection sub-range is set to one-fifth of the angle corresponding to the previous detection sub-range in the preset sequence. Taking detection sub-range 402 as an example, in the case of overlap, the overlapping portion of detection sub-range 402 is one-fifth of the distance from the edge of detection sub-range 401 to 402. Figure 6 In this embodiment, α is equal to 1°. Therefore, in the case of overlapping portions, the detection angle corresponding to the detection sub-range 402 is α plus 402a, which is... Figure 6 In the example 402b, the angle is 6°. Similarly, the detection angle corresponding to the detection sub-range 403 is β plus 403a, which is... Figure 6In 403b, the angle is 6°. For detection sub-range 401, since 401 is located at the boundary of the lidar detection range, the detection range of 401 can remain the original detection range, or it can overlap with detection sub-range 402 according to preset conditions; this application does not limit this. It is understood that by setting the overlapping area between the detection sub-range and two adjacent detection sub-ranges in the preset detection sequence, the recognition confidence of objects located in the left and right boundary areas of the detection sub-range can be effectively improved, thereby outputting recognition results and further improving the real-time performance of target object detection.

[0054] S302. Calculate the confidence level that the Nth detection sub-range includes a preset target object based on the point cloud data obtained in the Nth detection sub-range; if the confidence level is greater than or equal to a preset threshold, output the preset target obstacle recognition result.

[0055] The preset target object can be a pedestrian, vehicle, animal, tree, building, or other object with speed that obstructs the normal movement of the vehicle. This application does not limit the specific type of the preset target object.

[0056] Specifically, after acquiring point cloud data for the Nth detection sub-range, the LiDAR performs detection on the point cloud data within the Nth detection sub-range, calculates the confidence level of the preset target object included in the point cloud data of that detection sub-range, and outputs the preset target obstacle recognition result if the confidence level exceeds a preset threshold. The obstacle recognition result can be presented in forms including, but not limited to, the obstacle's point cloud data, obstacle category result, speed information, distance information, and warning images or sound information.

[0057] Furthermore, if the confidence level of the LiDAR detecting a preset target object within the Nth detection sub-range is less than a preset threshold, then point cloud data from the (N+1)th detection sub-range is acquired, and the point cloud data from the two detection sub-ranges are fused to calculate the confidence level of the two detection sub-ranges containing the preset target object. If the confidence level is greater than or equal to the preset threshold, the obstacle recognition result is output. If the confidence level is less than the preset threshold, point cloud data from the (N+2)th detection sub-range is acquired, and the point cloud data from the (N+2)th detection sub-range, the Nth detection sub-range, and the (N+1)th detection sub-range are fused. Based on the fused point cloud data, the confidence level of the fused multiple detection sub-ranges containing the preset target object is calculated, where N+2 ≤ M. For example, if the confidence level of the preset target object detected from the 1st detection sub-range, the 2nd detection sub-range, ..., up to the (M-1)th detection sub-range does not reach the preset threshold, then a frame of data scanning can be completed and the detection result output.

[0058] like Figure 7The diagram shown illustrates a mechanical lidar as an example, whose detection range includes... Figure 4 The six detection sub-ranges, ordered chronologically, are 501, 502, 503, 504, and 505. The preset confidence threshold for the lidar is set to 0.7. Figure 7 As shown in Figure a, the lidar first acquires the point cloud data of detection sub-range 501 and detects that the confidence level of 501 is 0.4. At this time, the confidence level of detection sub-range 501 does not exceed the preset confidence threshold of 0.7. Then, the point cloud data of detection sub-range 502 is merged with the point cloud data of detection sub-range 501 to obtain the following result. Figure 7 The merged point cloud data is shown in b, and the merged point cloud data is then inspected to obtain the merged result. Figure 7 The confidence level of the point cloud data shown in b is 0.9. Since the detected confidence level of 0.9 exceeds the preset confidence threshold of 0.7, the obstacle detection result can be output and displayed on the vehicle terminal screen as obstacle point cloud data or obstacle warning information. In this embodiment, the obstacle is determined to be a vehicle based on the point cloud data; therefore, the detected vehicle point cloud data or vehicle warning information can be output and displayed on the vehicle terminal. The obstacle detection result may include, but is not limited to, the point cloud data of obstacle 10, obstacle category determination results, speed information, distance information, and warning images or sound information.

[0059] Further, calculating the confidence level that the Nth detection sub-range includes a preset target object based on the point cloud data within the acquired Nth detection sub-range includes: clustering the point cloud data within the acquired Nth detection sub-range; extracting features from the clustered point cloud data; and matching the extracted feature information with the feature information of the preset target object to obtain the confidence level that the Nth detection sub-range includes the preset target object. In this embodiment, the type of features extracted after clustering is not limited.

[0060] As an optional embodiment of this application, after clustering the point cloud data within the acquired Nth detection sub-range, the method further includes: determining whether the clustered point cloud data is located in the boundary region adjacent to the Nth detection sub-range and the (N-1)th or (N+1)th detection sub-range; if the clustered point cloud data is located in the boundary region, then, according to a preset rule, a portion or all of the point cloud data of the (N-1)th or (N+1)th detection sub-range is acquired, fused, and a recognition result is output. The preset rule, for example, can be based on the size of the (N-1)th or (N+1)th detection sub-range to determine whether to acquire a portion or all of the N-1th detection sub-ranges and fuse them with the Nth detection sub-range to output a recognition result. Taking the boundary region between the Nth and N-1th detection sub-ranges of the clustered point cloud data as an example, if the N-1th detection sub-range is smaller than a preset threshold (e.g., for a mechanical LiDAR, smaller than the detection range corresponding to a 10° field of view), then the entire N-1th detection sub-range is fused with the Nth detection sub-range to output the recognition result. If the N-1th detection sub-range is larger than the preset threshold (e.g., for a mechanical LiDAR, larger than the detection range corresponding to a 10° field of view), then the proportion of the fused portion of the N-1th detection sub-range to the whole is set according to a preset rule. The proportion of the fused portion to the whole is related to the size of the clustered point cloud and the size of the N-1th detection sub-range. It can be understood that the larger the size of the clustered point cloud and the smaller the size of the detection sub-range, the larger the proportion of the fused portion to the whole. Specific partial fusion scenarios are as follows... Figure 6 As shown, Figure 6 The diagram illustrates the detection regions corresponding to sub-ranges 401, 402, and 403 when there is partial fusion. Specifically, in the case of no overlap and evenly divided angles, the detection angles corresponding to sub-ranges 401, 402, and 403 are all 5°, as shown in Figures 401a, 402a, and 403a. In the case of overlap and evenly divided angles, the overlap portion of any detection sub-range is defined as one-fifth of the angle corresponding to the previous detection sub-range in a preset order. Taking detection sub-range 402 as an example, in the case of overlap, the overlap portion of detection sub-range 402 is one-fifth of the distance from the edge of detection sub-range 401 to 402, as shown in Figure α. In this embodiment, α equals 1°. By identifying objects at the boundaries of the detection sub-ranges and thus setting partially fused regions, the recognition confidence of objects in the left and right boundary regions of the detection sub-ranges can be effectively improved, thereby outputting recognition results and further improving the real-time performance of target object detection.

[0061] This application provides an obstacle detection method that, by changing the output mode of a lidar, divides the detection range of the lidar into multiple sub-ranges, and detects these sub-ranges in a preset order, acquiring and accumulating the confidence levels of the multiple sub-ranges in real time, and outputting the obstacle detection result when the detection confidence level exceeds a preset confidence level. The obstacle detection method provided in this application improves the efficiency and real-time performance of obstacle detection by serializing the detection and outputting data.

[0062] Please refer to Figure 8 As shown, Figure 8 This is a flowchart illustrating another obstacle detection method provided in an embodiment of this application. Figure 5 Taking the rotating mechanical lidar shown as an example, the method includes:

[0063] S801. The lidar determines the detection range corresponding to each of the M detection sub-ranges within a detection cycle; where M is a positive integer greater than or equal to 2.

[0064] Specifically, taking a mechanical rotating lidar as an example, the detection angle corresponding to one rotation cycle of the rotating mechanical lidar is 360°. If M=6, then the detection angles corresponding to the six detection sub-ranges are obtained in a preset order, each corresponding to 60°. The preset order may include, but is not limited to, the scanning order of the lidar. For the specific division method of the detection sub-ranges, please refer to the above embodiment; it will not be repeated in this embodiment.

[0065] S802: The LiDAR acquires the current point cloud data, determines the scene complexity of the LiDAR based on the acquired point cloud data, and judges whether the current scene complexity is greater than a preset threshold. If yes, execute S803; otherwise, execute S804.

[0066] The process of determining the scene complexity of the LiDAR based on the acquired point cloud data includes: acquiring the current environmental information of the LiDAR based on the point cloud data of the LiDAR; and determining the scene complexity of the LiDAR based on the current environmental information of the LiDAR.

[0067] The environmental information may include scene information and obstacle information. The scene information may include the number of traffic lights, intersection information, pedestrian crossing information, and lane markings. The obstacle information may include, for example, the number of obstacles, their locations, their speeds, and their orientation.

[0068] Specifically, determining the scene complexity of the lidar based on its current environmental information involves: determining the road information of the lidar based on the environmental information; wherein the road information includes the lane in which the lidar is located.

[0069] Based on the environmental information, obstacle information on the lane where the radar is located is determined, including the number of obstacles, their positions, speeds, and directions of movement.

[0070] The complexity of the scene in which the radar is located is determined based on the obstacle information on the lane in which the radar is located.

[0071] In an optional embodiment, determining the complexity of the scene where the lidar is located based on the current environmental information of the lidar can also be done by: determining the scene where the lidar is located based on the environmental information acquired by the lidar.

[0072] The number and speed of obstacles in the radar's environment are determined based on the environmental information.

[0073] The complexity of the scene in which the radar is located is determined based on the number and speed of obstacles in the scene.

[0074] Specifically, the LiDAR detects the current driving environment information, determines the scene in which the LiDAR is located based on the environment information, and determines the scene complexity. The current scene complexity is compared with a preset threshold. If the scene complexity is greater than the preset threshold, step S803 is executed; if the scene complexity is less than the preset threshold, step S804 is executed. For details regarding the driving environment information, driving scene, and scene complexity, please refer to the above embodiment; this embodiment will not repeat them here.

[0075] It is understood that the higher the complexity of the scenario, the higher the real-time requirement for the target detection area, and the more screening sub-ranges are required in the preset rules.

[0076] It is understood that the scenarios include intersection scenarios, highway scenarios, and urban road scenarios. It is also understood that the intersection scenario has the highest real-time requirements for the target detection area, therefore the intersection scenario has the largest number of target detection area segments, followed by the urban road scenario, and lastly the highway scenario.

[0077] S803, the lidar adjusts the number of detection sub-ranges from M to P; where the adjusted detection sub-range is smaller than the original detection sub-range, and P is a positive integer greater than M.

[0078] Specifically, if the scene complexity within the lidar detection range, as represented by the current environmental information, exceeds a preset threshold, the number of detection sub-ranges is adjusted from M to P. The adjusted detection sub-ranges are smaller than the original detection sub-ranges, meaning the detection angle corresponding to each sub-range is reduced, and the number of detection sub-ranges is increased. For specific adjustment methods, please refer to the above embodiments. The specific manifestation of scene complexity exceeding the preset threshold in practical applications is also described in the above embodiments, and will not be repeated here.

[0079] S804. The lidar adjusts the M detection sub-ranges to K detection sub-ranges; wherein the adjusted detection sub-ranges are larger than the original detection sub-ranges, and K is a positive integer less than M.

[0080] Specifically, if the scene complexity within the lidar detection range, as represented by the current environmental information, is less than a preset threshold, the number of detection sub-ranges is adjusted from M to K. The adjusted detection sub-ranges are larger than the original detection sub-ranges, meaning the detection angle corresponding to each sub-range is increased, and the number of sub-ranges is decreased. For specific adjustment methods, please refer to the above embodiments. The practical application of the scenario complexity being less than the preset threshold is also described in the above embodiments, and will not be repeated here.

[0081] For example, if the obstacle density within the detection range of the rotating mechanical lidar, as represented by the environmental information acquired by the lidar in this embodiment, exceeds a preset level, then S903 is executed to adjust the detection sub-ranges of the rotating mechanical lidar from 6 to 10, that is, the detection angle corresponding to each detection sub-range is 36°. It should be noted that in this embodiment, when the preset adjustment conditions are met, the number of N or K is not limited, and the adjustment standard can be set manually.

[0082] S805: The lidar acquires point cloud data within M detection sub-ranges based on the adjusted detection angle and preset sequence.

[0083] Specifically, the lidar acquires point cloud data within any detection sub-range based on the adjusted detection angle of 36° and the preset sequence.

[0084] S806. The lidar detects the point cloud data of the first detection sub-range in the preset sequence, calculates the confidence that the first detection sub-range includes the preset target object, and determines whether the confidence exceeds the preset threshold. If yes, execute S808; otherwise, execute S807.

[0085] Specifically, when the lidar detects the point cloud data in the first detection sub-range in the preset sequence, it clusters the point cloud data in the first detection sub-range, extracts features from the clustered point cloud data, such as calculating the polygonal envelope or rectangular box of the clustered point cloud data to estimate the feature information of the clustered point cloud data, and then matches the feature information obtained above with the feature information of the preset target object to obtain the confidence level that the point cloud data in the first detection sub-range includes the preset target object, and determines whether the confidence level exceeds the preset threshold. If yes, S808 is executed; if no, S807 is executed.

[0086] S807: The LiDAR accumulates the point cloud data of the second detection sub-range according to a preset order, and calculates the confidence level of the point cloud data of the two acquired detection sub-ranges. It determines whether the confidence level exceeds a preset threshold. If yes, proceed to S808. If no, continue to calculate the confidence level of the point cloud data after fusing the point cloud data of the next detection sub-range with the previously accumulated point cloud data, including the target obstacle. When the confidence level exceeds the threshold, output the obstacle recognition result. If no, continue to accumulate and fuse the point cloud data of the detection sub-range according to the preset order. If the point cloud data of the Mth detection sub-range is accumulated, output the fusion recognition result of one frame of point cloud data.

[0087] For specific methods of accumulating point cloud data and calculating confidence levels, please refer to the above embodiments; these will not be repeated in this embodiment.

[0088] S808. If the confidence level exceeds the preset threshold, output the obstacle detection result.

[0089] Specifically, when the confidence level exceeds a preset threshold, the obstacle detection result is output and displayed on the vehicle terminal. The specific content of the obstacle detection result is detailed in the above embodiment and will not be repeated here.

[0090] This application's embodiments increase the flexibility and accuracy of real-time obstacle detection by adjusting the number and segmentation method of detection sub-ranges through real-time monitoring of scene complexity. Simultaneously, the obstacle detection method provided in this application divides the entire frame of point cloud data into multiple detection sub-ranges. These sub-ranges are then detected and their confidence scores calculated according to a preset order. If the confidence score does not exceed a preset threshold, the point cloud data of the next detection sub-range is fused with the current point cloud data and detected, obtaining the fused confidence score. This process continues until the obtained confidence score exceeds a preset threshold, at which point the obstacle detection result is output and displayed on the vehicle terminal. This method effectively reduces the target obstacle detection time and improves real-time performance.

[0091] Please refer to Figure 9As shown, based on obstacle detection methods, Figure 9 This is a schematic diagram of an obstacle detection device provided in an embodiment of this application, including:

[0092] The first acquisition module 901 is used to acquire point cloud data in the Nth detection sub-range of the lidar according to a preset order; the detection range of one detection cycle of the lidar includes M detection sub-ranges; the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M;

[0093] The detection module 902 is used to calculate the confidence level that the Nth detection sub-range includes a preset target object based on the point cloud data acquired in the Nth detection sub-range; if the confidence level is greater than or equal to a preset threshold, the preset target obstacle recognition result is output.

[0094] In some embodiments, the apparatus further includes:

[0095] The first calculation module is used to acquire point cloud data of the N+1th detection sub-range when the confidence level of the Nth detection sub-range including the preset target object is less than a preset threshold, fuse the point cloud data of the two detection sub-ranges, and calculate the confidence level of the two detection sub-ranges including the preset target object.

[0096] The output module is used to output the obstacle recognition result if the confidence level is greater than or equal to a preset threshold.

[0097] The second calculation module is used to acquire point cloud data of N+2 detection sub-ranges if the confidence level is less than a preset threshold, fuse the point cloud data of the N+2 detection sub-ranges, the Nth detection sub-range, and the N+1th detection sub-range, and calculate the confidence level of the multiple detection sub-ranges including the preset target object based on the fused point cloud data, where N+2≤M.

[0098] In some embodiments, the apparatus further includes:

[0099] A partitioning module is used to determine the partitioning method of the M detection sub-ranges; the partitioning method includes: partitioning according to the detection time corresponding to one detection cycle of the lidar, and partitioning according to the detection range corresponding to one detection cycle of the lidar;

[0100] The first acquisition module 901 is specifically used for:

[0101] Based on the detection time corresponding to one detection cycle of the lidar, determine the detection time corresponding to each of the M detection sub-ranges; obtain the first detection time corresponding to the Nth detection sub-range; according to the scanning order of the lidar, when the lidar reaches the first detection time, obtain the point cloud data corresponding to the Nth detection sub-range;

[0102] And / or, based on the detection range corresponding to one detection cycle of the lidar, determine the detection range corresponding to each of the M detection sub-ranges; obtain the first detection range corresponding to the Nth detection sub-range; according to the scanning order of the lidar, when the lidar scans the first detection range, obtain the point cloud data corresponding to the Nth detection sub-range.

[0103] In some embodiments, the apparatus further includes:

[0104] The second acquisition module is used to acquire environmental information of the lidar before the first acquisition module 901 acquires the point cloud data within the Nth detection sub-range of the lidar in a preset order.

[0105] The first determining module is used to determine the scene where the lidar is located based on the environmental information;

[0106] The first adjustment module is used to adjust the number M of the detection sub-ranges included in one detection cycle of the lidar according to the scenario.

[0107] In some embodiments, the apparatus further includes:

[0108] The second determining module is used to determine the scene complexity of the scene where the lidar is located;

[0109] The judgment module is used to determine whether the complexity of the scene is greater than a preset threshold.

[0110] The second adjustment module is used to readjust the number M of the detection sub-ranges corresponding to one detection cycle of the lidar according to the value of the scene complexity when the scene complexity is greater than a preset threshold.

[0111] In some embodiments, the apparatus further includes:

[0112] The third determining module is used to determine the general detection range and the target detection range within the detection range of the lidar before the first acquiring module 901 acquires the point cloud data within the Nth detection sub-range of the lidar in a preset order; the target detection range is a user-preset key detection area.

[0113] The fourth determining module is used to determine the preset division rules of the M detection sub-ranges based on the general detection range and the target detection range.

[0114] In some embodiments, the detection module 902 includes:

[0115] Clustering unit, used to cluster the point cloud data within the acquired Nth detection sub-range;

[0116] The feature extraction unit is used to extract features from the clustered point cloud data;

[0117] A matching unit is used to match the extracted feature information with the feature information of the preset target object to obtain the confidence that the Nth detection subrange includes the preset target object.

[0118] Please refer to Figure 10 The diagram shows a structural schematic of another obstacle detection device provided in this embodiment. This obstacle detection device may include at least: at least one processor 1001, such as a CPU; at least one network interface 1004; a user interface 1003; a memory 1005; and at least one communication bus 1002. The communication bus 1002 is used to establish communication between these components. The user interface 1003 may include, but is not limited to, a camera, display, touchscreen, keyboard, mouse, joystick, etc. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a WIFI interface), through which a communication connection can be established with a server. The memory 1002 may be a high-speed RAM or non-volatile memory, such as at least one disk storage device. Figure 10 As shown, the memory 1005, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and program instructions.

[0119] It should be noted that the network interface 1004 can connect to a receiver, a transmitter, or other communication modules. Other communication modules may include, but are not limited to, WiFi modules, operator network communication modules, etc. It is understood that the obstacle detection device in this embodiment may also include a receiver, a transmitter, and other communication modules.

[0120] Processor 1001 can be used to call program instructions stored in memory 1005 and can execute the following methods:

[0121] Point cloud data within the Nth detection sub-range of the lidar are acquired in a preset order; the detection range of one detection cycle of the lidar includes the M detection sub-ranges; wherein, the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M;

[0122] The confidence level of the Nth detection sub-range including the preset target object is calculated based on the point cloud data obtained in the Nth detection sub-range; if the confidence level is greater than or equal to the preset threshold, the preset target obstacle recognition result is output.

[0123] The processor 1001 may also be used to perform:

[0124] If the confidence that the Nth detection sub-range includes the preset target object is less than the preset threshold, obtain the point cloud data of the N+1th detection sub-range, fuse the point cloud data of the two detection sub-ranges, and calculate the confidence that the two detection sub-ranges include the preset target object.

[0125] If the confidence level is greater than or equal to a preset threshold, the obstacle recognition result is output.

[0126] If the confidence level is less than a preset threshold, continue to acquire point cloud data of N+2 detection sub-ranges, fuse the point cloud data of N+2 detection sub-ranges, the Nth detection sub-range and the N+1th detection sub-range, and calculate the confidence level of the multiple detection sub-ranges including the preset target object based on the fused point cloud data, where N+2≤M.

[0127] Possibly, before the processor 1001 acquires the point cloud data within the Nth detection sub-range of the lidar in a preset order, it is also used to perform:

[0128] Determine the division method for the M detection sub-ranges; the division method includes: dividing according to the detection time corresponding to one detection cycle of the lidar, and dividing according to the detection range corresponding to one detection cycle of the lidar;

[0129] The processor 1001 acquires point cloud data within the Nth detection sub-range of the lidar according to a preset order, specifically executing:

[0130] Based on the detection time corresponding to one detection cycle of the lidar, determine the detection time corresponding to each of the M detection sub-ranges; obtain the first detection time corresponding to the Nth detection sub-range; according to the scanning order of the lidar, when the lidar reaches the first detection time, obtain the point cloud data corresponding to the Nth detection sub-range;

[0131] And / or, based on the detection range corresponding to one detection cycle of the lidar, determine the detection range corresponding to each of the M detection sub-ranges; obtain the first detection range corresponding to the Nth detection sub-range; according to the scanning order of the lidar, when the lidar scans the first detection range, obtain the point cloud data corresponding to the Nth detection sub-range.

[0132] Possibly, before the processor 1001 acquires the point cloud data within the Nth detection sub-range of the lidar in a preset order, it is also used to perform:

[0133] Obtain the environmental information of the lidar;

[0134] The scene in which the lidar is located is determined based on the environmental information;

[0135] The number M of the detection sub-ranges included in one detection cycle of the lidar is adjusted according to the scenario.

[0136] The processor 1001 may also be used to perform:

[0137] Determine the scene complexity of the scene where the lidar is located;

[0138] Determine whether the complexity of the scenario exceeds a preset threshold;

[0139] When the scene complexity exceeds a preset threshold, the number M of the detection sub-ranges corresponding to one detection cycle of the lidar is readjusted according to the value of the scene complexity.

[0140] Possibly, before the processor 1001 acquires the point cloud data within the Nth detection sub-range of the lidar in a preset order, it is also used to perform:

[0141] The general detection range and target detection range within the detection range of the lidar are determined; the target detection range is a user-preset key detection area.

[0142] The preset division rules for the M detection sub-ranges are determined based on the general detection range and the target detection range.

[0143] Possibly, the processor 1001 calculates the confidence level that the Nth detection sub-range includes a preset target object based on the point cloud data acquired within the Nth detection sub-range, specifically executing:

[0144] Cluster the point cloud data acquired within the Nth detection sub-range;

[0145] Feature extraction is performed on the clustered point cloud data;

[0146] The extracted feature information is matched with the feature information of the preset target object to obtain the confidence that the Nth detection subrange includes the preset target object.

[0147] This application also provides a computer-readable storage medium storing instructions that, when executed on a computer or processor, cause the computer or processor to perform one or more steps of any of the above methods. If the constituent modules of the obstacle detection device described above are implemented as software functional units and sold or used as independent products, they can be stored in the computer-readable storage medium.

[0148] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially as a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted through the computer-readable storage medium. The computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, Digital Subscriber Line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., Digital Video Discs, DVDs), or semiconductor media (e.g., solid-state disks, SSDs), etc.

[0149] 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. This program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the methods described above. The aforementioned storage medium includes various media capable of storing program code, such as read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks. Unless otherwise specified, the technical features of this embodiment and its implementation schemes can be combined arbitrarily.

[0150] The embodiments described above are merely preferred embodiments of this application and are not intended to limit the scope of this application. Any modifications and improvements made by those skilled in the art to the technical solutions of this application without departing from the spirit of this application should fall within the protection scope defined by the claims of this application.

Claims

1. An obstacle detection method, characterized in that, The method includes: Point cloud data of the Nth detection sub-range of the lidar is acquired in a preset order. The detection range of one detection cycle of the lidar includes M detection sub-ranges. The Nth detection sub-range is any one of the M detection sub-ranges, and N is less than or equal to M. Cluster the point cloud data within the Nth detection sub-range obtained; Feature extraction is performed on the clustered point cloud data, and the extracted feature information is matched with the feature information of a preset target object to obtain the confidence that the Nth detection subrange includes the preset target object. If the confidence level is greater than or equal to a preset threshold, a preset target obstacle recognition result is output. The preset target obstacle recognition result includes: point cloud data of the obstacle and obstacle category result. If the confidence level is less than the preset threshold, the point cloud data of the (N+1)th detection sub-range is fused with the accumulated point cloud data of the first N detection sub-ranges, and the confidence level of the fused point cloud data including the target obstacle is calculated. If the confidence level is greater than or equal to the preset threshold, the preset target obstacle recognition result is output. If the confidence level is still less than the preset threshold, the point cloud data of the detection sub-ranges is accumulated and fused according to the preset order. If the point cloud data of the Mth detection sub-range is accumulated, the preset target obstacle recognition result of one frame of point cloud data is output. After clustering the point cloud data within the Nth detection sub-range, the method further includes: Determine whether the clustered point cloud data is in the adjacent boundary region between the N detection sub-ranges and the (N-1)th or (N+1)th detection sub-ranges; If, in the adjacent boundary region between the Nth and N-1th detection sub-ranges, some or all of the point cloud data in the N-1th detection sub-range are obtained according to a preset rule, the obtained point cloud data in the N-1th detection sub-range is fused with the point cloud data in the Nth detection sub-range, and the preset target obstacle recognition result is output. If, in the adjacent boundary region between the Nth and N+1th detection sub-ranges, some or all of the point cloud data within the N+1th detection sub-range are acquired according to a preset rule, the acquired point cloud data within the N+1th detection sub-range is fused with the point cloud data within the Nth detection sub-range to output the preset target obstacle recognition result.

2. The method as described in claim 1, characterized in that, Before acquiring the point cloud data within the Nth detection sub-range of the lidar according to a preset order, the method further includes: The division method of the M detection sub-ranges is determined, and the division method includes: dividing according to the detection time corresponding to one detection cycle of the lidar; The division of detection time according to one detection cycle of the lidar includes: Based on the detection time corresponding to one detection cycle of the lidar, determine the detection time corresponding to each of the M detection sub-ranges; Obtain the first detection time corresponding to the Nth detection sub-range; According to the scanning sequence of the lidar, when the lidar reaches the first detection time, it acquires the point cloud data corresponding to the Nth detection sub-range.

3. The method as described in claim 1, characterized in that, Before acquiring the point cloud data within the Nth detection sub-range of the lidar according to a preset order, the method further includes: Obtain the environmental information of the lidar; The scene in which the lidar is located is determined based on the environmental information; The number M of the detection sub-ranges included in one detection cycle of the lidar is adjusted according to the scenario.

4. The method as described in claim 1, characterized in that, Any of the detection sub-ranges may overlap with two adjacent detection sub-ranges in the preset order; When the target detection sub-ranges overlap, the detection angle corresponding to the target detection sub-range is greater than the detection angle corresponding to the target detection sub-range when the target detection sub-ranges do not overlap.

5. The method as described in claim 1, characterized in that, Before acquiring the point cloud data within the Nth detection sub-range of the lidar according to a preset order, the method further includes: The general detection range and the target detection range within the detection range of the lidar are determined; the target detection range is the key detection area preset by the user, and the general detection range is the detection range of the lidar within one detection cycle, excluding the target detection range; Based on the general detection range and the target detection range, a preset division rule for the M detection sub-ranges is determined, and the detection range within a detection cycle is divided into M detection sub-ranges according to the preset division rule. The step of dividing the detection range within a detection cycle into M sub-ranges according to the preset division rule includes: According to the first division rule, the general detection range is divided into a detection sub-ranges, and according to the second division rule, the target detection range is divided into b detection sub-ranges; where a+b=M, a<b.

6. The method as described in claim 1, characterized in that, In the case of the clustered point cloud data within the adjacent boundary region between the Nth detection sub-range and the (N-1)th detection sub-range: If the size of the (N-1)th detection sub-range is less than a preset threshold, then all the (N-1)th detection sub-ranges are fused with the Nth detection sub-range to determine the recognition result; If the size of the (N-1)th detection sub-range is greater than a preset threshold, then according to the first preset rule, the proportion of the fused portion of the (N-1)th detection sub-range to the whole is set, and the recognition result is determined by fusing a portion of the (N-1)th detection sub-range with the Nth detection sub-range according to the proportion.

7. The method as described in claim 1, characterized in that, The method further includes: The complexity of the scene in which the lidar is located is determined by identifying only one of the obstacle information, such as the number of obstacles, the position of obstacles, and the speed of obstacles, from the point cloud information obtained by the lidar. The number M of the detection sub-ranges included in one detection cycle of the lidar is adjusted according to the complexity of the scene.

8. An obstacle detection device, characterized in that, The device includes: The acquisition module is used to acquire point cloud data within the Nth detection sub-range of the lidar in a preset order; the detection range of one detection cycle of the lidar includes M detection sub-ranges; the Nth detection sub-range is any one of the M detection sub-ranges; M is an integer greater than or equal to 2, and N is less than or equal to M; The detection module is used to cluster the point cloud data within the acquired Nth detection sub-range; extract features from the clustered point cloud data; match the extracted feature information with the feature information of a preset target object to obtain the confidence level that the Nth detection sub-range includes the preset target object; if the confidence level is greater than or equal to a preset threshold, output the preset target obstacle recognition result, which includes: point cloud data of the obstacle and obstacle category result; If the confidence level is less than the preset threshold, the point cloud data of the (N+1)th detection sub-range is fused with the accumulated point cloud data of the first N detection sub-ranges, and the confidence level of the fused point cloud data including the target obstacle is calculated. If the confidence level is greater than or equal to the preset threshold, the preset target obstacle recognition result is output. If the confidence level is still less than the preset threshold, the point cloud data of the detection sub-ranges is accumulated and fused according to the preset order. If the point cloud data of the Mth detection sub-range is accumulated, the preset target obstacle recognition result of one frame of point cloud data is output. After clustering the point cloud data within the Nth detection sub-range, the method further includes: determining whether the clustered point cloud data is located in the adjacent boundary region between the Nth and (N-1)th or (N+1)th detection sub-ranges; if it is located in the adjacent boundary region between the Nth and (N-1)th detection sub-ranges, acquiring part or all of the point cloud data within the (N-1)th detection sub-range according to a preset rule, fusing the acquired part or all of the point cloud data within the (N-1)th detection sub-range with the point cloud data within the Nth detection sub-range, and outputting the preset target obstacle recognition result; if it is located in the adjacent boundary region between the Nth and (N+1)th detection sub-ranges, acquiring part or all of the point cloud data within the (N+1)th detection sub-range according to a preset rule, fusing the acquired part or all of the point cloud data within the (N+1)th detection sub-range with the point cloud data within the Nth detection sub-range, and outputting the preset target obstacle recognition result.

9. An obstacle detection device, characterized in that, Includes processor, memory, and communication interface: The processor is connected to the memory and the communication interface; The memory is used to store executable program code; The processor runs a program corresponding to the executable program code stored in the memory to perform the obstacle detection method as described in any one of claims 1-7.

Images