Target detection methods

Abstract

This invention discloses a target detection method. The method belongs to the field of lidar and includes: acquiring target point cloud data of a region to be detected, and scene information of the region to be detected, wherein the region to be detected contains a target to be detected; determining gradient information of multiple points based on the target point cloud data, wherein the gradient information includes velocity gradients and / or distance gradients of multiple points; detecting the region to be detected based on the gradient information and scene information to obtain initial contour information of the target to be detected; and performing target detection on multiple points based on the initial contour information to obtain the target type of the target to be detected. This invention solves the technical problem of low efficiency in target detection in related technologies.

Description

Technical Field

[0001] This invention relates to the field of lidar, and more specifically, to a target detection method. Background Technology

[0002] Frequency Modulated Continuous Wave (FMCW) laser radar is an advanced laser radar technology. Compared to traditional pulsed laser radar, it uses frequency modulation over time when transmitting signals, acquiring distance information by detecting the frequency difference between the transmitted and received waves. This method not only improves the accuracy of distance measurement but also simultaneously acquires the velocity information of the target object, making it a key technology for achieving four-dimensional imaging (distance, angle, velocity, and intensity). In autonomous vehicles, acquiring velocity information is of great value for predicting object dynamics and optimizing path planning.

[0003] Point cloud data processing is a core step in FMCW LiDAR applications, including point cloud classification, clustering, feature extraction, and boundary detection. Traditional point cloud processing algorithms have high computational complexity and relatively low detection efficiency in target detection scenarios, making it difficult to meet the real-time and low-power requirements of autonomous driving systems.

[0004] There is currently no effective solution to the above problems. Summary of the Invention

[0005] This invention provides a target detection method to at least address the technical problem of low efficiency in target detection in related technologies.

[0006] According to one aspect of the present invention, a target detection method is provided, comprising: acquiring target point cloud data of a region to be detected and scene information of the region to be detected, wherein the region to be detected contains a target to be detected; determining gradient information of multiple points based on the target point cloud data, wherein the gradient information includes velocity gradients and / or distance gradients of multiple points; detecting the region to be detected based on the gradient information and the scene information to obtain initial contour information of the target to be detected; and performing target detection on multiple points based on the initial contour information to obtain the target type of the target to be detected.

[0007] Further, acquiring target point cloud data of the region to be detected includes: acquiring initial point cloud data of the region to be detected; dividing the region to be detected into multiple sub-regions; and sorting the initial point cloud data based on the multiple sub-regions to obtain target point cloud data.

[0008] Further, the initial point cloud data is sorted based on multiple sub-regions to obtain target point cloud data, including: allocating the initial point cloud data to multiple sub-regions according to the spatial location of the initial point cloud data to obtain an allocation result, wherein the allocation result is used to represent the sub-point cloud data allocated to multiple sub-regions; sorting the sub-point cloud data of multiple sub-regions respectively to obtain a first sorting result; and filtering the sub-point cloud data of multiple sub-regions based on the first sorting result and the allocation result to obtain target point cloud data.

[0009] Furthermore, the gradient information of multiple points is determined based on the target point cloud data, including: determining the velocity of multiple points based on the target point cloud data, and determining the velocity gradient of multiple points based on the velocity of multiple points; determining the detection distance of multiple points based on the target point cloud data, and determining the distance gradient of multiple points based on the detection distance of multiple points; and determining the gradient information based on the velocity gradient and / or distance gradient.

[0010] Further, determining the velocity gradient of multiple points based on the velocities of multiple points includes: determining the target velocity of a target point among the multiple points, wherein the target point is a point among the multiple points whose velocity is within a preset velocity range; sorting the multiple points based on the target velocity of the target point and the velocities of the multiple points to obtain a second sorting result; and determining the velocity gradient based on the second sorting result.

[0011] Furthermore, based on the target speed of the target point and the speeds of multiple points, the multiple points are sorted to obtain a second sorting result, including: grouping the multiple points based on the target speed and the speeds of multiple points to obtain a first group and a second group, wherein the speeds of the points in the first group are greater than the target speed, and the speeds of the points in the second group are less than or equal to the target speed; sorting the points in the first group and the points in the second group respectively to obtain the sorting results of the first group and the second group; and obtaining the second sorting result based on the sorting results of the first group and the second group.

[0012] Furthermore, based on the velocity gradient and scene information, the region to be detected is detected to obtain the initial contour information of the target to be detected, including: determining at least one first point among multiple points whose velocity gradient is greater than a preset velocity gradient; dividing the multiple points based on the at least one first point to obtain multiple velocity segments, the multiple velocity segments corresponding to different velocity thresholds in different scenes, and different velocity thresholds used to detect different target types; and detecting the target to be detected based on the velocity of multiple points, different velocity thresholds, and scene information to obtain the initial contour information.

[0013] Further, target detection is performed on multiple points based on initial contour information to obtain the target type of the target to be detected, including: marking multiple points based on initial contour information to obtain marking results, wherein the marking results are used to mark points related to velocity in the initial contour information; determining at least one second point and at least one third point among the multiple points based on the marking results, wherein at least one second point represents a point used to describe the boundary corresponding to the target to be detected, and at least one third point represents a point used to describe the main body region corresponding to the target to be detected, the target to be detected consists of a boundary and a main body region; connecting at least one second point based on the scanning angle of the region to be detected to determine the boundary information of the target to be detected; and detecting the target to be detected based on the boundary information and at least one third point to obtain the target type.

[0014] Further, determining the distance gradient of multiple points based on the detection distance of multiple points includes: determining the target detection distance of a preset point among the multiple points, wherein the preset point is a point among the multiple points whose detection distance is within a preset distance range; sorting the multiple points based on the target detection distance of the preset point and the detection distance of the multiple points to obtain a third sorting result; and determining the distance gradient based on the third sorting result.

[0015] Furthermore, based on the target detection distance of the preset point and the detection distance of multiple points, the multiple points are sorted to obtain a third sorting result, including: grouping the multiple points based on the target detection distance and the detection distance of multiple points to obtain a third group and a fourth group, wherein the detection distance of the points in the third group is greater than the target detection distance, and the detection distance of the points in the fourth group is less than or equal to the target detection distance; sorting the points in the third group and the points in the fourth group respectively to obtain the point sorting result of the third group and the point sorting result of the fourth group; and obtaining the third sorting result based on the point sorting result of the third group and the point sorting result of the fourth group.

[0016] According to another aspect of the present invention, a target detection apparatus is also provided, comprising: a first acquisition module, configured to acquire target point cloud data of a region to be detected and scene information of the region to be detected, wherein the region to be detected contains a target to be detected; a first determination module, configured to determine gradient information of multiple points based on the target point cloud data, wherein the gradient information includes velocity gradients and / or distance gradients of multiple points; a first detection module, configured to detect the region to be detected based on the gradient information and scene information to obtain initial contour information of the target to be detected; and a second detection module, configured to perform target detection on multiple points based on the initial contour information to obtain the target type of the target to be detected.

[0017] According to another aspect of the present invention, an electronic device is also provided, comprising: a memory storing an executable program; and a processor for running the program, wherein the program executes the methods of various embodiments of the present invention during runtime.

[0018] According to another aspect of the present invention, a computer-readable storage medium is also provided, comprising: the computer-readable storage medium including a stored executable program, wherein, when the executable program is executed, it controls the device where the storage medium is located to perform the methods of various embodiments of the present invention.

[0019] According to another aspect of the present invention, a computer program product is also provided, comprising: a computer program that, when executed by a processor, implements the methods of various embodiments of the present invention.

[0020] According to another aspect of the present invention, a computer program is also provided, which, when executed by a processor, implements the methods of the various embodiments of the present invention.

[0021] In this embodiment of the invention, the target detection system acquires target point cloud data of the area to be detected and scene information of the area to be detected; determines gradient information of multiple points based on the target point cloud data; detects the area to be detected based on the gradient information and scene information to obtain the initial contour information of the target to be detected; and performs target detection on multiple points based on the initial contour information to obtain the target type of the target to be detected. By analyzing the gradient information, the target detection system can eliminate point cloud data that is irrelevant to target detection, thereby reducing the amount of data for subsequent depth processing and effectively optimizing computing resources. In addition, by analyzing the gradient information and scene information, the target detection system can more accurately identify the initial contour information of dynamic targets, improving the accuracy of target detection. This achieves the goal of improving both the speed and accuracy of target detection, thereby improving the technical effect of improving the efficiency of point cloud data processing and solving the technical problem of low efficiency in target detection in related technologies. Attached Figure Description

[0022] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings:

[0023] Figure 1 This is a flowchart of a target detection method according to an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of a target detection device according to an embodiment of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0026] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0027] According to an embodiment of the present invention, an embodiment of a target detection method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0028] Figure 1 This is a flowchart of a target detection method according to an embodiment of the present invention, such as... Figure 1 As shown, the method includes the following steps:

[0029] Step S102: Obtain the target point cloud data of the area to be detected, as well as the scene information of the area to be detected, wherein the area to be detected contains the target to be detected.

[0030] The aforementioned area to be detected can refer to the three-dimensional spatial area that the LiDAR sensor can cover, typically part of the environment surrounding an autonomous vehicle or robot. The aforementioned target point cloud data can refer to the set of reflection points on the surface of an object detected by the LiDAR within the area to be detected. The aforementioned scene information can refer to other environmental information within the area to be detected besides the point cloud data; for example, the scene information can be map data, lighting conditions, weather conditions, etc., but is not limited to these. The aforementioned target to be detected can be any object that may exist within the area to be detected and needs to be identified and tracked by the LiDAR system; for example, the target to be detected can be a vehicle, pedestrian, bicycle, animal, roadblock, etc., but is not limited to these.

[0031] In one optional embodiment, considering that the point cloud data generated by the LiDAR is the basis for constructing a 3D model of the vehicle's surrounding environment, the point cloud data may contain information such as the position, size, and shape of the target object in space. This helps the target detection system (hereinafter referred to as the detection system) understand and analyze the current scene, thereby identifying static and dynamic obstacles such as roads, buildings, vehicles, and pedestrians. Therefore, the detection system needs to acquire the target point cloud data of the area to be detected. Since the velocity information of multiple points in the target point cloud data helps the detection system distinguish between static and dynamic obstacles, predict the trajectory of dynamic obstacles, and thus make more accurate decisions and path planning, in addition to the target point cloud data, the detection system can also acquire scene information of the area to be detected, including weather conditions, lighting conditions, and terrain features. This scene information is crucial for correctly analyzing the point cloud data.

[0032] For example, the detection area can be assumed to be an area of ​​N meters in front of the autonomous vehicle, including the highway on which the vehicle is traveling, the guardrails on both sides, the distant mountain background, and multiple vehicles traveling ahead. The detection system can use an FMCW lidar to emit continuous wave laser light with a frequency that varies linearly with time into the detection area at a frequency of thousands of times per second. This laser light is reflected back to the radar receiver after encountering reflection points such as vehicles and guardrails in front. The detection system can collect these reflected signals, process them to generate point cloud data, and calculate the radial velocity of the reflection points. Simultaneously, the detection system can also use cameras to capture scene information such as the layout of the road ahead, traffic signs, road markers, and weather conditions (such as rain, snow, and fog), thus providing a data foundation for subsequent target detection.

[0033] Step S104: Determine the gradient information of multiple points based on the target point cloud data, wherein the gradient information includes the velocity gradient and / or distance gradient of multiple points.

[0034] The gradient information mentioned above can be the rate of change of data point attributes in space. Gradient information can be used to depict the boundaries, shape, and dynamic characteristics of an object. The velocity gradient mentioned above can refer to the rate of change of velocity of each point on the object's surface or within space in the point cloud data. The distance gradient mentioned above can be the rate of change of distance between each point on the object's surface and the observation point or radar sensor in the point cloud data.

[0035] In one alternative embodiment, the velocity gradient is extracted by analyzing the rate of change of velocity values ​​between adjacent points or multiple points in the point cloud. In the FMCW lidar system, due to the Doppler effect, each point cloud data point carries target velocity information. By calculating the velocity gradient, high-speed moving targets can be quickly located. The distance gradient reflects the rate of change of target distance in the point cloud data. In 3D point cloud modeling, the distance gradient helps the detection system identify object edges and discontinuities, thereby more accurately depicting the target's outline. This is crucial for distinguishing road boundaries, identifying static obstacles, or constructing more detailed environmental maps, and is the foundation for achieving high-precision positioning and environmental perception. Based on gradient information, the detection system can intelligently adjust its processing strategy, such as performing more detailed analysis on areas with large velocity gradients or prioritizing point cloud data with significant changes in distance gradients. This efficient allocation of resources reduces unnecessary computation and improves the overall efficiency of target detection.

[0036] Step S106: Detect the region to be detected based on gradient information and scene information to obtain the initial contour information of the target to be detected.

[0037] The aforementioned initial contour information can be a preliminary shape or boundary description of the target object obtained by analyzing the gradient information of point cloud data and fusing scene information.

[0038] In one alternative embodiment, considering that gradient information provides intuitive information about the motion state of the target object, changes in gradient information can serve as key features in point cloud data, assisting the detection system in identifying the boundaries and shapes of the target object. However, in complex scenes, especially under complex backgrounds or poor lighting conditions, the detection system may struggle to accurately distinguish object boundaries relying solely on gradient information. Therefore, the detection system can combine gradient information and scene information to detect the aforementioned region to obtain the initial contour of the target object.

[0039] For example, we can assume that the detection system has already calculated the aforementioned gradient information and extracted feature information such as road edges, traffic signs, and traffic light status from the scene information. This feature information can be combined with the gradient information of the point cloud data. Based on the combined information, the detection system can detect points in the point cloud data where the gradient information changes abruptly. Suppose the detection system detects a set of continuous gradient information change points that form a straight line or curve. The detection system can connect these points to form a preliminary contour, which can be a simple polygon used to represent the boundary and position of the target object.

[0040] Step S108: Target detection is performed on multiple points based on the initial contour information to obtain the target type of the target to be detected.

[0041] The aforementioned target detection refers to the identification and localization of target objects. The aforementioned target type refers to the category by which target objects are classified during the target detection process.

[0042] In an optional embodiment, considering that the initial contour information provides the preliminary shape and position of the target object, which is the basis for identifying and classifying the target, the detection system can perform target detection on the above multiple points based on the initial contour information to obtain the target type of the target to be detected. Classifying the above targets to be detected can make the detection system clear about which targets need to be detected in particular, thereby optimizing the allocation of the detection system's computing resources and improving the efficiency of target detection.

[0043] For example, assuming the detection system has already acquired the initial contour information of the target object using FMCW LiDAR, to accurately determine the target type, the detection system can input this initial contour information into a pre-trained deep learning model. This model can identify multiple different target types, such as vehicles, pedestrians, and bicycles, from the point cloud data of the FMCW LiDAR. The deep learning model can then perform calculations and analysis based on the input from the detection system, outputting the type probability corresponding to the target object back to the detection system. The detection system can then select the type with the higher probability as the target type of the target object.

[0044] In this embodiment of the invention, the target detection system acquires target point cloud data of the area to be detected and scene information of the area to be detected; determines gradient information of multiple points based on the target point cloud data; detects the area to be detected based on the gradient information and scene information to obtain the initial contour information of the target to be detected; and performs target detection on multiple points based on the initial contour information to obtain the target type of the target to be detected. By analyzing the gradient information, the target detection system can eliminate point cloud data that is irrelevant to target detection, thereby reducing the amount of data for subsequent depth processing and effectively optimizing computing resources. In addition, by analyzing the gradient information and scene information, the target detection system can more accurately identify the initial contour information of dynamic targets, improving the accuracy of target detection. This achieves the goal of improving both the speed and accuracy of target detection, thereby improving the technical effect of improving the efficiency of point cloud data processing and solving the technical problem of low efficiency in target detection in related technologies.

[0045] Further, acquiring target point cloud data of the region to be detected includes: acquiring initial point cloud data of the region to be detected; dividing the region to be detected into multiple sub-regions; and sorting the initial point cloud data based on the multiple sub-regions to obtain target point cloud data.

[0046] The initial point cloud data mentioned above can be unprocessed point cloud data directly obtained from the environment. The sub-regions mentioned above can be independent regions obtained by dividing the entire area to be detected.

[0047] In one optional embodiment, considering that the amount of point cloud data generated by LiDAR is usually very large, directly processing the entire point cloud dataset would consume a lot of computing resources and time. To improve detection efficiency, the detection system can first acquire the initial point cloud data of the area to be detected, and then divide the area to be detected into multiple sub-regions. The detection system can then perform parallel processing on these multiple sub-regions, that is, the detection system can sort the initial point cloud data based on multiple sub-regions to obtain the target point cloud data. Specifically, after completing the division of sub-regions, the detection system can independently sort the initial point cloud data within each sub-region. The sorting criteria can vary according to the application scenario and the needs of target detection. Furthermore, considering that each point cloud data in FMCW LiDAR also contains Doppler velocity information, the detection system can sort the point cloud data in parallel within each sub-region based on the Doppler velocity information using sorting algorithms such as quicksort and heapsort. This allows the detection system to focus on points with significant velocity changes in the point cloud data, thereby reducing the computational complexity of the target detection process.

[0048] For example, the detection system uses LiDAR to scan the surrounding environment to collect point cloud data in real time. From this point cloud dataset, the system can extract the 3D coordinates, reflection intensity, and Doppler velocity information of each reflection point, thus constructing initial point cloud data. Then, the system can use a nine-grid strategy to spatially divide the detection area. Specifically, the LiDAR detection area can be divided into nine sub-regions, each covering a specific angle and distance range. This nine-grid division helps the system focus on objects at different directions and distances during processing, while reducing overall computational complexity. Finally, since Doppler velocity information can quickly help the system identify moving targets, the system can sort the point cloud data within each sub-region based on the Doppler velocity, thereby obtaining the target point cloud data.

[0049] For example, in the scenario of autonomous driving, after constructing the initial point cloud data, the detection system can dynamically divide the area to be detected into fan-shaped regions. This division method can more flexibly adapt to different environments. Specifically, to further improve the efficiency of target detection, if the current environment is high-speed, the detection system can divide the area further ahead into more detailed fans to prioritize the detection of high-speed vehicles and obstacles. If the current environment is low-speed, the detection system can focus more on nearby objects, so the area closer to the vehicle can be divided into denser fans to facilitate the detection of pedestrians, bicycles, and other targets with slower movement speeds. After completing the above fan-shaped region division, the detection system can use a fast sorting algorithm to sort the point cloud data in each fan-shaped region from high to low according to Doppler speed, thereby quickly and efficiently identifying the target point cloud data.

[0050] Further, the initial point cloud data is sorted based on multiple sub-regions to obtain target point cloud data, including: allocating the initial point cloud data to multiple sub-regions according to the spatial location of the initial point cloud data to obtain an allocation result, wherein the allocation result is used to represent the sub-point cloud data allocated to multiple sub-regions; sorting the sub-point cloud data of multiple sub-regions respectively to obtain a first sorting result; and filtering the sub-point cloud data of multiple sub-regions based on the first sorting result and the allocation result to obtain target point cloud data.

[0051] The above allocation result can be the result of assigning all points in the initial point cloud data to multiple sub-regions according to their spatial location. The above first sorting result can refer to the result of sorting the sub-point cloud data within each sub-region according to a certain feature, such as distance, reflection intensity, Doppler velocity, etc., based on the allocation result.

[0052] In an optional embodiment, considering that the data volume of each sub-region is relatively small, locally sorting the point cloud data within each sub-region significantly reduces the complexity of sorting compared to sorting all point cloud data. Therefore, the detection system can first allocate the initial point cloud data, that is, the detection system can allocate the initial point cloud data to multiple sub-regions based on the spatial location of the initial point cloud data, thereby obtaining the above allocation result. After completing the above allocation process, in order to improve the sorting efficiency, the detection system can sort the sub-point cloud data of multiple sub-regions separately to obtain the above first sorting result. Based on this result, the detection system can filter out point cloud data directly related to target detection based on features such as distance, reflection intensity, and Doppler velocity. These data points constitute the target point cloud data, and subsequent processing steps can focus on the information contained in the above target point cloud data, thereby improving the accuracy and efficiency of target detection.

[0053] For example, to efficiently process massive initial point cloud data, the detection system can divide the area to be detected into multiple sub-regions based on spatial location, and assign all points in the initial point cloud data to these sub-regions based on their spatial location. To identify and prioritize objects with significant velocity changes, within each sub-region, the detection system can further sort the sub-point cloud data according to Doppler velocity, thus obtaining the first sorting result. After obtaining the first sorting result, the detection system can filter the sub-point cloud data based on the sorting and assignment results to determine the target point cloud data.

[0054] Furthermore, the gradient information of multiple points is determined based on the target point cloud data, including: determining the velocity of multiple points based on the target point cloud data, and determining the velocity gradient of multiple points based on the velocity of multiple points; determining the detection distance of multiple points based on the target point cloud data, and determining the distance gradient of multiple points based on the detection distance of multiple points; and determining the gradient information based on the velocity gradient and / or distance gradient.

[0055] In one alternative embodiment, determining the velocity gradient first involves extracting the velocity value of each point in the point cloud. This extraction process can be based on FMCW radar and achieved through the Doppler effect. Subsequently, the detection system determines the velocity gradient by calculating the velocity differences between adjacent points or points within a local region. Determining the velocity gradient helps identify high-speed moving objects or the dynamic characteristics of objects. For example, in autonomous driving scenarios, the detection system can quickly detect vehicles that suddenly accelerate or decelerate, which is crucial for immediate obstacle avoidance decisions. Furthermore, the introduction of the velocity gradient reduces unnecessary analysis of static backgrounds, allowing the detection system to focus on dynamic targets, thereby improving the efficiency and accuracy of target detection. The determination of the distance gradient is based on the distance from each point in the point cloud data to the radar sensor or observation point. By analyzing the distance differences within a local region, the detection system can identify the edge and shape features of objects. The calculation of the distance gradient not only helps in quickly locating objects but also supports the detection system in building high-precision environmental maps, providing accurate information for path planning and navigation. The determination of the aforementioned gradient information can further integrate velocity and distance gradients to form a more comprehensive analysis of target point cloud data. This integrated analysis process allows the detection system to understand objects in the environment from both dynamic and static dimensions, improving the robustness and flexibility of target detection. For example, on highways, the detection system may rely more on velocity gradients to quickly identify and track moving vehicles, while in urban environments, distance gradients may become crucial for identifying static obstacles and complex road layouts. Combining the gradient information from velocity and distance gradients enables the detection system to be more intelligent and flexible in handling different scenarios and targets, achieving effective perception of both dynamic and static environments.

[0056] Further, determining the velocity gradient of multiple points based on the velocities of multiple points includes: determining the target velocity of a target point among the multiple points, wherein the target point is a point among the multiple points whose velocity is within a preset velocity range; sorting the multiple points based on the target velocity of the target point and the velocities of the multiple points to obtain a second sorting result; and determining the velocity gradient based on the second sorting result.

[0057] The aforementioned target point can refer to a point whose speed falls within the aforementioned preset speed range. The aforementioned target speed can be the speed value possessed by the target point. The aforementioned preset speed range can be a speed threshold interval set when determining the target point, used to filter out point cloud data points whose speed falls within this range. The aforementioned second sorting result can refer to the result of sorting all target points according to their target speed after determining the target point.

[0058] In an optional embodiment, considering that the point cloud data contains a large number of points, in order to accurately calculate the velocity gradient of the multiple points, the detection system can set a preset velocity range and select the points whose velocities fall within the preset range as target points. These target points can serve as the benchmark for velocity sorting. After selecting the target points, the detection system can sort them based on the target velocities of the target points and the relative velocities of the multiple points to obtain the second sorting result. Based on the second sorting result, the detection system can calculate the velocity difference between each target point and its neighboring points, thereby determining the velocity gradient.

[0059] For example, the detection system can define the preset speed range as [-10m / s, 20m / s]. Suppose that in a single scan, the detection system generates point cloud data of 5000 points using LiDAR. After filtering within the preset speed range, 350 points have speeds between -10m / s and 20m / s; these 350 points are the target points. For these 350 target points, the detection system can record the Doppler velocity of each point as its target velocity. Then, the detection system can sort the point cloud data based on the target velocities and the velocity information of the points contained in the point cloud data, thus obtaining the second sorting result. Based on this second sorting result, the detection system can determine the velocity gradient by calculating the velocity difference between adjacent points.

[0060] It should be noted that the preset speed range, the number of points included in the point cloud data, and the amount of target point data mentioned above are only examples for demonstration purposes. Staff can set them according to their actual needs, and there are no restrictions here.

[0061] Furthermore, based on the target speed of the target point and the speeds of multiple points, the multiple points are sorted to obtain a second sorting result, including: grouping the multiple points based on the target speed and the speeds of multiple points to obtain a first group and a second group, wherein the speeds of the points in the first group are greater than the target speed, and the speeds of the points in the second group are less than or equal to the target speed; sorting the points in the first group and the points in the second group respectively to obtain the sorting results of the first group and the second group; and obtaining the second sorting result based on the sorting results of the first group and the second group.

[0062] The first group mentioned above can refer to the set of points in the point cloud data whose speed is greater than the target speed of the target point. Points in the first group may represent objects moving faster than the target point, such as overtaking vehicles or fast-moving pedestrians, but are not limited to these. The second group mentioned above can refer to the set of points in the point cloud data whose speed is less than or equal to the target speed of the target point. Points in the second group may represent objects moving slower or stationary. The sorting result of the points in the first group can be obtained by sorting the points in the first group based on their speed. The sorting result of the points in the second group can be obtained by sorting the points in the second group based on their speed.

[0063] In one optional embodiment, the points in the point cloud data are grouped according to their velocity, allowing the detection system to process these two groups of data separately, rather than processing all points in a single operation. This improves the efficiency of target detection, especially when dealing with large amounts of point cloud data, and reduces unnecessary computational resource consumption. Therefore, the detection system can pre-define two groups, classifying points in the point cloud data with velocities greater than the target velocity of the target point into the first group, and points with velocities less than or equal to the target velocity of the target point into the second group. After this classification, the detection system can simultaneously sort the points in both groups according to their velocity, obtaining the sorting results for the first and second groups. The detection system can then summarize the sorting results from both groups to obtain the second sorting result.

[0064] In another optional embodiment, considering that there can be multiple target velocities, the detection system can also group multiple points based on the multiple target velocities and the velocities of multiple points, thereby obtaining N groups, where N>2. In this embodiment, taking N=3 as an example, the detection system can group points in the point cloud data whose velocities are greater than target velocity 1 into the first group; simultaneously, the detection system can group points whose velocities are less than or equal to target velocity 1 but greater than target velocity 2 into the second group; and the detection system can group points whose velocities are less than or equal to target velocity 2 into the third group, thus completing the grouping of the point cloud data. Then, the detection system can simultaneously sort the points within each group according to their velocities to obtain the sorting results of the three groups. At this point, the detection system can integrate the sorting results of the three groups to obtain the second sorting result.

[0065] It should be noted that the number of groups and the number of target speeds shown above are for illustrative purposes only. Staff can set them according to their actual needs, and there are no restrictions here.

[0066] Furthermore, based on the velocity gradient and scene information, the region to be detected is detected to obtain the initial contour information of the target to be detected, including: determining at least one first point among multiple points whose velocity gradient is greater than a preset velocity gradient; dividing the multiple points based on the at least one first point to obtain multiple velocity segments, the multiple velocity segments corresponding to different velocity thresholds in different scenes, and different velocity thresholds used to detect different target types; and detecting the target to be detected based on the velocity of multiple points, different velocity thresholds, and scene information to obtain the initial contour information.

[0067] The aforementioned preset velocity gradient can be a pre-defined value for the rate of change of velocity, used to identify points in the point cloud data where the velocity change is significant. The aforementioned first point can be a point in the point cloud data where the velocity gradient is greater than the preset velocity gradient. The aforementioned multiple velocity segments can refer to the velocity range represented by each interval after dividing the point cloud data according to velocity characteristics; different velocity segments can be used to detect different target objects.

[0068] In an optional embodiment, considering that points with large velocity gradients help the detection system quickly locate the boundaries of different objects, thereby performing target detection more efficiently, the detection system, when detecting the area to be detected based on velocity gradients and scene information, can first find one or more points in the point cloud data whose velocity gradients exceed the aforementioned preset velocity gradients, i.e., the aforementioned first points. Based on the information of the first points, the detection system can divide the point cloud data into multiple velocity segments, where each velocity segment contains points with similar velocities. The division of the aforementioned velocity segments takes into account velocity thresholds under different scenes. These thresholds are used to distinguish different types of objects, such as pedestrians, bicycles, and cars. Based on this, the detection system can analyze the points within each segment separately, rather than processing the entire point cloud dataset, thereby effectively improving the efficiency of target detection. After obtaining multiple velocity segments, the detection system can detect potential target objects based on the velocity information of the points within each segment, the preset velocity thresholds, and the current scene information, in order to identify and extract the initial contour information of the target objects, i.e., the boundaries and approximate shapes of the target objects.

[0069] For example, assuming the preset velocity gradient is 0.8 m / s, the velocity of point A in the point cloud data is 0 m / s, the velocity of point B is 1.5 m / s, the velocity of point C is 2.5 m / s, the velocity of point D suddenly jumps to 10 m / s, the velocity of point E is 12 m / s, and the velocity of point F is 13.5 m / s. The detection system calculates the velocity gradient between adjacent points in the point cloud data. The velocity gradient between point D and point C is 7.5 m / s, which exceeds the preset velocity gradient of 0.8 m / s. Therefore, point D can be identified as the first point. The detection system can divide the point cloud data based on the first point D, obtaining multiple velocity segments as shown below: Velocity segment 1 can include points A, B, and C, with a velocity range of [0 m / s, 2.5 m / s], and velocity segment 2 can include points D, E, and F, with a velocity range of [10 m / s, 13.5 m / s]. In speed segment 1, the detection system can use a speed threshold of 3 m / s to detect potential pedestrians and bicycles. Points A, B, and C all have speeds within this threshold range, therefore they are identified as potential pedestrian or bicycle targets. The detection system can further analyze the distribution and distance information of points A, B, and C to determine their relative positions and sizes, thus obtaining the initial contour information of points A, B, and C in speed segment 1. In speed segment 2, the detection system can use a speed threshold of 5 m / s to detect normally moving vehicles. Points D, E, and F all have speeds within this threshold range, therefore they are identified as potential vehicle targets. Similarly, the detection system can analyze the distribution and distance information of points D, E, and F to determine their relative positions and sizes, thus obtaining the initial contour information of points D, E, and F in speed segment 2.

[0070] It should be noted that all the specific data such as the preset speed gradient and speed threshold mentioned above are only shown as examples. Staff can set them according to their actual needs, and there are no restrictions here.

[0071] Further, target detection is performed on multiple points based on initial contour information to obtain the target type of the target to be detected, including: marking multiple points based on initial contour information to obtain marking results, wherein the marking results are used to mark points related to velocity in the initial contour information; determining at least one second point and at least one third point among the multiple points based on the marking results, wherein at least one second point represents a point used to describe the boundary corresponding to the target to be detected, and at least one third point represents a point used to describe the main body region corresponding to the target to be detected, the target to be detected consists of a boundary and a main body region; connecting at least one second point based on the scanning angle of the region to be detected to determine the boundary information of the target to be detected; and detecting the target to be detected based on the boundary information and at least one third point to obtain the target type.

[0072] The aforementioned labeling results can refer to the labeling of points in point cloud data that are closely related to the initial contour information. The second point can be a point used to describe the boundary of the target to be detected, i.e., a point located at the edge of the target object in the point cloud data. The third point can be a point used to represent the main body region corresponding to the target to be detected, i.e., a point located inside the target object. The aforementioned boundary information can be information used to describe the boundary contour of the target to be detected.

[0073] In one optional embodiment, the detection system can identify which points match the preliminary contour information of the target by comparing the velocity, distance, and distribution of each point in the point cloud data, and then mark these identified points to obtain a marking result. This marking result helps reduce the amount of data processed subsequently, improving the overall computational efficiency and response speed of target detection. Then, based on the marking result, the detection system can determine at least one second point and at least one third point from a plurality of points. The determination of the second and third points further refines the description of the target object. Specifically, since the target to be detected consists of a boundary and a main body region, the second point can be used to describe the boundary of the target object, facilitating the subsequent determination of the target's precise contour. The third point can be used to describe the main body region of the target. By analyzing the distribution and velocity of these points, the size and shape of the target can be determined more accurately, thus providing a detailed data foundation for subsequent target object detection. Subsequently, the detection system connects the second points according to the scanning order or angle of the point cloud data to form the boundary contour of the target object. This process involves the continuity analysis of the points to ensure that the connected boundary lines accurately reflect the true edge of the target, which helps to more accurately identify and track the target in subsequent steps. Finally, the detection system can combine the boundary information mentioned above with the velocity, position, and distribution data provided in the third point, and apply deep learning, feature matching, or other algorithms to determine the type of the target object.

[0074] For example, the target could be a pedestrian. Assume the detection system has constructed initial outline information of the pedestrian and pre-trained a pedestrian recognition model. By analyzing velocity gradients, the detection system can mark points with velocities between 0 m / s and 3 m / s as points associated with potential pedestrian outline information. Then, the detection system can mark points with velocities around 3 m / s as a set of possible pedestrian points. From the marking results, the detection system can select points in the point cloud that exhibit the shape of a human body's edge as the second point, used to describe the pedestrian's outer boundary, i.e., the pedestrian's outline. Subsequently, the detection system can select points located within the main body region of the pedestrian from the marking results as the third point. The velocity of the third point can be stabilized at around 3 m / s, and it should be within the outline described by the second point, providing information about the pedestrian's specific shape and size. After selecting the second and third points, the detection system can connect the second points according to the scanning angle order of the point cloud data to form the pedestrian's outline boundary. This step, by connecting points with large velocity gradient changes, can depict the pedestrian's edge, thereby determining the pedestrian's boundary information. Finally, the system will input the features of the marked boundaries and interior points into the pedestrian recognition model above for comparison to confirm whether the currently detected target is a pedestrian.

[0075] It should be noted that the specific speed values ​​mentioned above are for illustrative purposes only, and staff can adjust them according to actual needs; no restrictions are imposed here.

[0076] Further, determining the distance gradient of multiple points based on the detection distance of multiple points includes: determining the target detection distance of a preset point among the multiple points, wherein the preset point is a point among the multiple points whose detection distance is within a preset distance range; sorting the multiple points based on the target detection distance of the preset point and the detection distance of the multiple points to obtain a third sorting result; and determining the distance gradient based on the third sorting result.

[0077] The aforementioned preset points can refer to points selected from point cloud data based on a specific detection distance range. In the point cloud processing workflow, the selection of preset points can be based on this preset distance range. The aforementioned target detection distance can refer to the straight-line distance from each point in the point cloud data, measured by a lidar or laser sensor, to the sensor. Target detection distance is a fundamental attribute of point cloud data and the basis for subsequent distance gradient calculations. The aforementioned preset distance range can be a pre-defined range of detection distances for the point cloud data. The aforementioned third sorting result can refer to the point cloud data list or structure sorted by detection distance after sorting based on the target detection distances of the preset points and the detection distances of the entire point cloud dataset. The third sorting result provides an ordered dataset for subsequent distance gradient calculations, facilitating the detection system's analysis of the rate of change of distance with spatial location, thereby identifying the target's contour features.

[0078] In one optional embodiment, the detection system can first capture the detection range of each point in the point cloud data using a lidar sensor. Next, the detection system can filter out preset points within a preset distance range. The selection of these preset points can be based on a preliminary estimate of the target distance or the specific needs of the application scenario. For example, in an urban driving environment, the detection system may focus more on targets at near to mid-range distances, so the preset distance range may be set to a relatively close range. By focusing on preset points, the detection system can more effectively process information directly related to the target, avoiding redundant analysis of distant or background data, thereby improving processing speed and resource utilization. Based on the target detection distance of the preset points and the detection distance of the entire point cloud dataset, the detection system can sort the point cloud data to form a third sorting result. This sorting process can employ algorithms such as quicksort, mergesort, or heapsort to ensure efficiency. The third sorting result reflects the distribution of distances in the point cloud data, providing a basis for calculating the distance gradient.

[0079] Furthermore, based on the target detection distance of the preset point and the detection distance of multiple points, the multiple points are sorted to obtain a third sorting result, including: grouping the multiple points based on the target detection distance and the detection distance of multiple points to obtain a third group and a fourth group, wherein the detection distance of the points in the third group is greater than the target detection distance, and the detection distance of the points in the fourth group is less than or equal to the target detection distance; sorting the points in the third group and the points in the fourth group respectively to obtain the point sorting result of the third group and the point sorting result of the fourth group; and obtaining the third sorting result based on the point sorting result of the third group and the point sorting result of the fourth group.

[0080] The third group mentioned above can be a set of points in the point cloud data whose detection distance is greater than the preset target detection distance. The points in the third group may represent the far boundary of the target object, objects behind it, or distant features in the background environment. By focusing on the points in the third group, the detection system can analyze the contour features of the target object more deeply and identify distant obstacles related to the target object or in the background. The fourth group mentioned above can be a set of points in the point cloud data whose detection distance is less than or equal to the preset target detection distance. The points in the fourth group may represent the near boundary of the target object, objects in front of it, or close-range features in the background environment.

[0081] In one optional embodiment, the detection system can divide multiple points in the point cloud data into two groups, namely a third group and a fourth group, based on the target detection distance of a preset point. The third group contains points whose detection distance is greater than the target detection distance of the preset point, while the fourth group contains points whose detection distance is less than or equal to the target detection distance of the preset point. This grouping method can quickly filter out the possible locations and ranges of the target, providing a basis for subsequent sorting and distance gradient calculation. After grouping, the detection system can sort the points in the third and fourth groups respectively based on the detection distance. Sorting based on the detection distance helps the detection system quickly identify the edges and contours of objects. In particular, sorting based on grouping can reduce the overall amount of data processed and improve the real-time performance of target detection. Finally, the detection system can merge the sorting results of the third and fourth groups to form a third sorting result. This result contains the sorting information of all points in the point cloud data according to the detection distance, providing an ordered dataset for subsequent distance gradient calculation. By merging the sorting results, the detection system can comprehensively analyze the target and its surrounding environment, improving the comprehensiveness and accuracy of target detection.

[0082] According to an embodiment of the present invention, a target detection device is provided. It should be noted that this device can be used to execute the target detection method described above. The specific implementation and application scenarios are the same as those in the above embodiment, and will not be repeated here. Figure 2 This is a schematic diagram of a target detection device according to an embodiment of the present invention, such as... Figure 2 As shown, the device includes:

[0083] The first acquisition device 202 is used to acquire target point cloud data of the area to be detected and scene information of the area to be detected, wherein the area to be detected contains the target to be detected.

[0084] The first determining device 204 is used to determine gradient information of multiple points based on target point cloud data, wherein the gradient information includes velocity gradients and / or distance gradients of multiple points.

[0085] The first detection device 206 is used to detect the area to be detected based on gradient information and scene information, and obtain the initial contour information of the target to be detected.

[0086] The second detection device 208 is used to perform target detection on multiple points based on initial contour information to obtain the target type of the target to be detected.

[0087] Furthermore, the first acquisition device is also used to: acquire initial point cloud data of the area to be detected; divide the area to be detected into multiple sub-regions; and sort the initial point cloud data based on the multiple sub-regions to obtain target point cloud data.

[0088] Furthermore, the first acquisition device is also used to: allocate the initial point cloud data to multiple sub-regions according to the spatial location of the initial point cloud data, and obtain an allocation result, wherein the allocation result is used to represent the sub-point cloud data allocated to the multiple sub-regions; sort the sub-point cloud data of the multiple sub-regions respectively, and obtain a first sorting result; and filter the sub-point cloud data of the multiple sub-regions based on the first sorting result and the allocation result to obtain the target point cloud data.

[0089] Furthermore, the first determining device is also used to: determine the velocity of multiple points based on target point cloud data, and determine the velocity gradient of multiple points based on the velocity of multiple points; determine the detection distance of multiple points based on target point cloud data, and determine the distance gradient of multiple points based on the detection distance of multiple points; and determine gradient information based on the velocity gradient and / or distance gradient.

[0090] Furthermore, the first determining device is also used to: determine the target velocity of a target point among multiple points, wherein the target point is a point among multiple points whose velocity is within a preset velocity range; sort the multiple points based on the target velocity of the target point and the velocities of the multiple points to obtain a second sorting result; and determine the velocity gradient based on the second sorting result.

[0091] Furthermore, the first determining device is also used to: group the multiple points based on the target speed and the speed of the multiple points to obtain a first group and a second group, wherein the speed of the points in the first group is greater than the target speed, and the speed of the points in the second group is less than or equal to the target speed; sort the points in the first group and the points in the second group respectively to obtain the sorting result of the first group and the sorting result of the second group; and obtain a second sorting result based on the sorting result of the first group and the sorting result of the second group.

[0092] Furthermore, the first detection device is also used to: determine at least one first point among multiple points whose velocity gradient is greater than a preset velocity gradient; divide the multiple points based on the at least one first point to obtain multiple velocity segments, the multiple velocity segments corresponding to different velocity thresholds in different scenarios, the different velocity thresholds being used to detect different target types; and detect the target to be detected based on the velocity of the multiple points, the different velocity thresholds, and the scene information to obtain initial contour information.

[0093] Furthermore, the second detection device is also used to: mark multiple points based on initial contour information to obtain marking results, wherein the marking results are used to mark points related to the initial contour information and velocity; determine at least one second point and at least one third point among the multiple points based on the marking results, wherein at least one second point represents a point used to describe the boundary corresponding to the target to be detected, and at least one third point represents a point used to describe the main body region corresponding to the target to be detected, the target to be detected being composed of a boundary and a main body region; connect at least one second point based on the scanning angle of the region to be detected to determine the boundary information of the target to be detected; and detect the target to be detected based on the boundary information and at least one third point to obtain the target type.

[0094] Furthermore, the first determining device is also used to: determine the target detection distance of a preset point among multiple points, wherein the preset point is a point among multiple points whose detection distance is within a preset distance range; sort the multiple points based on the target detection distance of the preset point and the detection distance of the multiple points to obtain a third sorting result; and determine the distance gradient based on the third sorting result.

[0095] Furthermore, the first determining device is also used to: group the multiple points based on the target detection distance and the detection distance of the multiple points to obtain a third group and a fourth group, wherein the detection distance of the points in the third group is greater than the target detection distance, and the detection distance of the points in the fourth group is less than or equal to the target detection distance; sort the points in the third group and the points in the fourth group respectively to obtain the point sorting result of the third group and the point sorting result of the fourth group; and obtain a third sorting result based on the point sorting result of the third group and the point sorting result of the fourth group.

[0096] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0097] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0098] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0099] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0100] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0101] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0102] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A target detection method, characterized in that, include: Acquire target point cloud data of the area to be detected, and scene information of the area to be detected, wherein the area to be detected contains the target to be detected; The gradient information of the plurality of points is determined based on the target point cloud data, wherein the gradient information includes the velocity gradient and / or distance gradient of the plurality of points; Based on the gradient information and the scene information, the region to be detected is detected to obtain the initial contour information of the target to be detected; Based on the initial contour information, target detection is performed on the multiple points to obtain the target type of the target to be detected.

2. The target detection method according to claim 1, characterized in that, Acquire target point cloud data for the area to be detected, including: Obtain the initial point cloud data of the area to be detected; The region to be detected is divided into multiple sub-regions; The initial point cloud data is sorted based on the multiple sub-regions to obtain the target point cloud data.

3. The target detection method according to claim 2, characterized in that, The initial point cloud data is sorted based on the multiple sub-regions to obtain the target point cloud data, including: The initial point cloud data is allocated to the plurality of sub-regions according to the spatial location of the initial point cloud data to obtain an allocation result, wherein the allocation result is used to represent the sub-point cloud data allocated to the plurality of sub-regions; The sub-point cloud data of the multiple sub-regions are sorted respectively to obtain the first sorting result; Based on the first sorting result and the allocation result, the sub-point cloud data of the multiple sub-regions are filtered to obtain the target point cloud data.

4. The target detection method according to claim 1, characterized in that, Determining the gradient information of the multiple points based on the target point cloud data includes: The velocity of the plurality of points is determined based on the target point cloud data, and the velocity gradient of the plurality of points is determined based on the velocity of the plurality of points. The detection distance of the plurality of points is determined based on the target point cloud data, and the distance gradient of the plurality of points is determined based on the detection distance of the plurality of points; The gradient information is determined based on the velocity gradient and / or the distance gradient.

5. The target detection method according to claim 4, characterized in that, Determining the velocity gradient of the multiple points based on their velocities includes: Determine the target speed of a target point among the plurality of points, wherein the target point is a point among the plurality of points whose speed is within a preset speed range; Based on the target speed of the target point and the speed of the multiple points, the multiple points are sorted to obtain a second sorting result; The velocity gradient is determined based on the second sorting result.

6. The target detection method according to claim 5, characterized in that, Based on the target velocity of the target point and the velocities of the multiple points, the multiple points are sorted to obtain a second sorting result, including: Based on the target speed and the speed of the multiple points, the multiple points are grouped into a first group and a second group, wherein the speed of the points in the first group is greater than the target speed, and the speed of the points in the second group is less than or equal to the target speed; Sort the points in the first group and the points in the second group respectively to obtain the sorting results of the points in the first group and the sorting results of the points in the second group; The second sorting result is obtained based on the point sorting results of the first group and the point sorting results of the second group.

7. The target detection method according to claim 1, characterized in that, Based on the velocity gradient and the scene information, the region to be detected is detected to obtain the initial contour information of the target, including: Determine at least one first point among the plurality of points where the velocity gradient is greater than a preset velocity gradient; Based on the at least one first point, the plurality of points are divided to obtain a plurality of speed segments. The plurality of speed segments correspond to different speed thresholds in different scenarios. The different speed thresholds are used to detect different target types. The target to be detected is obtained by detecting the target based on the velocity of the multiple points, the different velocity thresholds, and the scene information, thereby obtaining the initial contour information.

8. The target detection method according to claim 7, characterized in that, Based on the initial contour information, target detection is performed on the multiple points to obtain the target type of the target to be detected, including: The points are marked based on the initial contour information to obtain a marking result, wherein the marking result is used to mark the points that are associated with the initial contour information and velocity; Based on the marking results, at least one second point and at least one third point are determined from the plurality of points, wherein the at least one second point represents a point used to describe the boundary corresponding to the target to be detected, and the at least one third point represents a point used to describe the main body region corresponding to the target to be detected, and the target to be detected is composed of the boundary and the main body region; Based on the scanning angle of the area to be detected, the at least one second point is connected to determine the boundary information of the target to be detected; The target to be detected is obtained by detecting the target based on the boundary information and the at least one third point.

9. The target detection method according to claim 4, characterized in that, Determining the distance gradient of the multiple points based on the detection distance of the multiple points includes: Determine the target detection distance of a preset point among the plurality of points, wherein the preset point is a point among the plurality of points whose detection distance is within a preset distance range; Based on the target detection distance of the preset point and the detection distance of the multiple points, the multiple points are sorted to obtain a third sorting result; The distance gradient is determined based on the third sorting result.

10. The target detection method according to claim 9, characterized in that, Based on the target detection distance of the preset point and the detection distance of the multiple points, the multiple points are sorted to obtain a third sorting result, including: Based on the target detection distance and the detection distance of the multiple points, the multiple points are grouped to obtain a third group and a fourth group. The detection distance of the points in the third group is greater than the target detection distance, and the detection distance of the points in the fourth group is less than or equal to the target detection distance. Sort the points in the third group and the points in the fourth group respectively to obtain the sorting results of the points in the third group and the sorting results of the points in the fourth group; The third sorting result is obtained based on the point sorting results of the third group and the point sorting results of the fourth group.

Images