Target post-processing method and system based on in-frame point cloud, and product

Abstract

The invention provides an in-frame point cloud-based target post-processing method, system and product. The method comprises the steps of processing point cloud data to determine a target frame and target drivable area points in the target frame; determining a convex hull point set including all the target drivable area points in the frame according to the target drivable area points; determining longitudinal maximum and transverse maximum and minimum convex hull points in the set; the target frame is located on the left side of the vehicle, and a right side convex hull point set is constructed by convex hull points which meet a first condition in the longitudinal maximum convex hull point and the transverse minimum convex hull point in sequence in the frame and the longitudinal distance between the convex hull points and the transverse minimum convex hull point in the longitudinal direction; the target frame is located on the right side of the vehicle, and a left side convex hull point set is constructed by convex hull points which are located in the longitudinal maximum convex hull point and the transverse maximum convex hull point in sequence in the frame and have the longitudinal distance meeting a second condition with the transverse maximum convex hull point; the frame rotates leftwards on the left side of the vehicle, and a right side boundary line is corrected through a right side convex hull point; the frame is arranged on the right side of the vehicle and turns right, and the left-side boundary line is corrected through the left-side convex hull point. The invention aims to improve the accuracy of radar output results.

Description

Technical Field

[0001] The present application relates to the field of point cloud processing technology, and in particular to a target post-processing method, system and product based on in-frame point cloud. Background Art

[0002] Self-driving cars rely on sensors on the car to obtain information about the surrounding environment. The environmental perception module processes the sensor information and passes it to subsequent modules for vehicle behavior planning. Therefore, in order to ensure the safety of self-driving cars, the reliability and accuracy of sensor information are crucial. The current mainstream sensors are mainly cameras, millimeter waves, and lidars. However, due to the high detection accuracy of lidars, dust, water vapor in the air, and small particles in car exhaust can also be detected. In these scenes and in severe weather environments such as rain, snow, and fog, the performance of lidars may be reduced to a certain extent. At the same time, the output of the lidar deep learning model and the fitting of the traditional clustering target box will inevitably have certain deviations. For these two types of problems, there is a need for response strategies to minimize the decline in perception performance caused by adverse weather or environmental factors and improve the accuracy of lidar output information. Summary of the invention

[0003] In view of this, the present application provides a target post-processing method, system and product based on point cloud within a frame, aiming to improve the accuracy of radar output results.

[0004] The first aspect of the present application provides a target post-processing method based on a point cloud within a frame, the method comprising:

[0005] Perform target recognition on point cloud data through the target recognition model to obtain multiple target frames;

[0006] By analyzing and processing the point cloud data, a target drivable area point located in the target frame is determined;

[0007] According to the target drivable area points in the target frame, determine the convex hull of the target frame and the convex hull point set corresponding to the convex hull, wherein the convex hull points in the convex hull point set are all the target drivable area points in the target frame;

[0008] Determine the maximum longitudinal convex hull point, the maximum transverse convex hull point, and the minimum transverse convex hull point in the convex hull point set of the target frame;

[0009] When the target frame is located on the left side of the vehicle, the right convex hull point set of the target frame is constructed by sequentially selecting the right convex hull points of the target frame that are located in the maximum convex hull point in the longitudinal direction and the minimum convex hull point in the transverse direction and whose longitudinal distance from the minimum convex hull point in the transverse direction meets the first condition;

[0010] When the target frame is located on the right side of the vehicle, the left convex hull point set of the target frame is constructed by sequentially selecting the left convex hull points of the target frame that are located in the longitudinal maximum convex hull point and the transverse maximum convex hull point and whose longitudinal distance from the transverse maximum convex hull point satisfies the second condition;

[0011] When the target frame is located on the left side of the vehicle and turns left, the right boundary line of the target frame is corrected by the right convex hull points in the right convex hull point set of the target frame;

[0012] When the target frame is located at the right side of the vehicle and turns right, the left boundary line of the target frame is corrected by the left convex hull points in the left convex hull point set of the target frame.

[0013] Optionally, determining the target drivable area point located in the target frame by analyzing and processing the point cloud data includes:

[0014] Determine the drivable area points in the point cloud data by analyzing and processing the point cloud data;

[0015] According to the location of the target frame and the location of the drivable area point, the target drivable area point located in the target frame is determined.

[0016] Optionally, before determining the target drivable area point located in the target frame by analyzing and processing the point cloud data, the method further includes:

[0017] Determining a virtual target frame among the multiple target frames according to the attribute information of the target frame;

[0018] Filtering virtual target frames among the multiple target frames to obtain remaining target frames;

[0019] The step of analyzing and processing the point cloud data to determine the target drivable area point located in the target frame comprises:

[0020] By analyzing and processing the point cloud data, the target drivable area points located in the target frame remaining after filtering are determined.

[0021] Optionally, before determining the convex hull of the target box and the convex hull point set corresponding to the convex hull according to the target drivable area point in the target box, the method further includes:

[0022] According to the attribute information of the target frame, determine the target frame that needs to be oriented corrected from all the target frames;

[0023] The method of determining the convex hull of the target box and the set of convex hull points corresponding to the convex hull according to the target drivable area points in the target box includes: determining the convex hull of the target box that needs to be corrected in direction and the set of convex hull points corresponding to the convex hull according to the target drivable area points in the target box that needs to be corrected in direction.

[0024] Optionally, when the target frame is located on the left side of the vehicle and turns left, the right boundary line of the target frame is corrected using the right convex hull points in the right convex hull point set of the target frame, including:

[0025] When the target frame is located on the left side of the vehicle and turns left, traverse the right convex hull points in the right convex hull point set of the target frame, and connect the currently traversed right convex hull point with the lateral minimum convex hull point to form a right predicted boundary line;

[0026] When the angle deviation between the obtained right predicted boundary line and the right boundary line of the target frame satisfies the set condition, the traversal is terminated and the right boundary line of the target frame is corrected with the right predicted boundary line whose deviation satisfies the set condition;

[0027] When the target frame is located at the right side of the vehicle and turns right, the left boundary line of the target frame is corrected using the left convex hull point in the left convex hull point set of the target frame, including:

[0028] When the target frame is located on the right side of the vehicle and turns right, traverse the left convex hull points in the left convex hull point set of the target frame, and connect the currently traversed left convex hull point with the horizontal maximum convex hull point to form a left predicted boundary line;

[0029] When the angle deviation between the obtained left predicted boundary line and the left boundary line of the target frame satisfies the set condition, the traversal is terminated and the left boundary line of the target frame is corrected with the left predicted boundary line whose deviation satisfies the set condition.

[0030] Optionally, the method further includes:

[0031] Clustering the point cloud data using a clustering algorithm to obtain each cluster target frame;

[0032] Determine the proportion of point clouds of different semantic categories in the clustering target frame, determine the size of the clustering target frame, determine the reflectivity information of the point cloud in the clustering target frame, and determine the first point cloud density of the point cloud within a set range of the centroid of the clustering target frame;

[0033] The cluster target frame is evenly divided into multiple segments in the longitudinal direction, and the second point cloud density around the center point of the first segment in the moving direction of the cluster target frame is calculated;

[0034] Determine a virtual clustering target frame in each clustering target frame based on the point cloud proportion, the size, the reflectivity information, the first point cloud density, and the second point cloud density;

[0035] The virtual clustering target frames in the clustering target frames are filtered to obtain a clustering result of the clustering algorithm.

[0036] Optionally, determine the reflectivity information of the point cloud within the cluster target box, including:

[0037] According to the attribute information of the point cloud in the cluster target frame, the number of point clouds with low reflectivity in the cluster target frame is determined, as well as the maximum reflectivity and the total reflectivity value of the point cloud in the cluster target frame;

[0038] Determine the proportion of low-reflectivity point clouds in the cluster target frame according to the total number of point clouds in the cluster target frame and the number of low-reflectivity point clouds;

[0039] According to the total number of point clouds and the total reflectivity value in the cluster target frame, the average reflectivity of the point clouds in the cluster target frame is determined.

[0040] Optionally, determining a virtual cluster target frame in each cluster target frame based on the point cloud proportion, the size, the reflectivity information, the first point cloud density, and the second point cloud density includes:

[0041] When the proportion of the low reflectivity point cloud corresponding to the cluster target frame is a set value, the cluster target frame is determined to be a virtual cluster target frame;

[0042] When the proportion of low reflectivity point clouds corresponding to the cluster target frame exceeds a first reflectivity threshold and the corresponding average reflectivity does not exceed a second reflectivity threshold, determining that the cluster target frame is a virtual cluster target frame;

[0043] When the proportion of low reflectivity point clouds corresponding to the cluster target frame exceeds the third reflectivity threshold, the corresponding average reflectivity does not exceed the fourth reflectivity threshold, and the corresponding maximum reflectivity does not exceed the fifth reflectivity threshold, the cluster target frame is determined to be a virtual cluster target frame;

[0044] When the proportion of point clouds of the noise semantic category in the cluster target box exceeds the category threshold, determining a size relationship between the size and a first size threshold, determining a density relationship between the first point cloud density and a first density threshold, and determining a quantity relationship between the total number of point clouds and a total number threshold;

[0045] Determining whether the clustering target frame is a virtual clustering target frame according to the size relationship, the density relationship and the quantity relationship;

[0046] When the proportion of point clouds of the noise semantic category in the clustering target frame does not exceed the category threshold, determine whether the clustering target frame is a virtual clustering target frame based on the proportion of point clouds of the background semantic category in the clustering target frame, the size, and the total number of point clouds; and determine whether the clustering target frame is a virtual clustering target frame based on the proportion of point clouds of the general obstacle semantic category in the clustering target frame, the position of the clustering target frame, and the size.

[0047] Optionally, the method further includes:

[0048] Determine whether there is a small-sized single cluster target box within a preset range around the vehicle category cluster target box;

[0049] In the case that there is a small-sized single cluster target frame within a preset range around the vehicle category cluster target frame, the vehicle category cluster target frame is determined to be a virtual cluster target frame.

[0050] The second aspect of the present application provides a target post-processing system based on in-frame point cloud, the system comprising:

[0051] The target recognition module is used to perform target recognition on the point cloud data through the target recognition model to obtain multiple target frames;

[0052] An analysis and processing module, used to determine a target drivable area point located in a target frame by analyzing and processing the point cloud data;

[0053] A convex hull point set determination module is used to determine the convex hull of the target box and the convex hull point set corresponding to the convex hull according to the target drivable area points in the target box, wherein the convex hull points in the convex hull point set are all the target drivable area points in the target box;

[0054] An extreme convex hull point determination module is used to determine the maximum longitudinal convex hull point, the maximum transverse convex hull point and the minimum transverse convex hull point in the convex hull point set of the target frame;

[0055] A right convex hull point set determination module is used to construct a right convex hull point set of the target frame when the target frame is located on the left side of the vehicle, using the convex hull points in the target frame that are sequentially located between the maximum convex hull point in the longitudinal direction and the minimum convex hull point in the transverse direction and whose longitudinal distance from the minimum convex hull point in the transverse direction meets the first condition;

[0056] A left convex hull point set determination module is used to construct a left convex hull point set of the target frame when the target frame is located on the right side of the vehicle, using the left convex hull points of the target frame whose convex hull points are sequentially located in the longitudinal maximum convex hull point and the transverse maximum convex hull point and whose longitudinal distance from the transverse maximum convex hull point meets the second condition;

[0057] A first correction module is used to correct the right boundary line of the target frame with the right convex hull points in the right convex hull point set of the target frame when the target frame is located on the left side of the vehicle and turns left;

[0058] The second correction module is used to correct the left boundary line of the target frame with the left convex hull points in the left convex hull point set of the target frame when the target frame is located at the right side of the vehicle and turns right.

[0059] The third aspect of the present application provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and running on the processor, wherein when the computer program is executed by the processor, the steps in the target post-processing method based on the in-frame point cloud as described in the first aspect of the present application are implemented.

[0060] The fourth aspect of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in the target post-processing method based on the in-frame point cloud as described in the first aspect of the present application are implemented.

[0061] The target post-processing method based on the point cloud in the frame provided in this application has the following advantages:

[0062] The embodiment of the present application provides a target post-processing method based on in-frame point cloud. First, target recognition is performed on point cloud data through a target recognition model to obtain multiple target frames; target drivable area points located in the target frame are determined by analyzing and processing the point cloud data; the convex hull of the target frame and the convex hull point set corresponding to the convex hull are determined according to the target drivable area points in the target frame, and the convex hull points in the convex hull point set are all the target drivable area points in the target frame; the maximum longitudinal convex hull point, the maximum transverse convex hull point and the minimum transverse convex hull point in the convex hull point set of the target frame are determined; when the target frame is located on the left side of the vehicle, the convex hull points in the target frame are located in the maximum longitudinal convex hull point and the minimum transverse convex hull point in the order of the convex hull points in the target frame. , the right convex hull point set of the target frame is constructed by the right convex hull points whose longitudinal distance from the minimum convex hull point in the horizontal direction meets the first condition; when the target frame is located on the right side of the vehicle, the left convex hull point set of the target frame is constructed by the convex hull points in the target frame that are sequentially located in the longitudinal maximum convex hull point and the transverse maximum convex hull point, and whose longitudinal distance from the maximum convex hull point in the transverse direction meets the second condition; when the target frame is located on the left side of the vehicle and turns left, the right side boundary line of the target frame is corrected by the right side convex hull points in the right side convex hull point set of the target frame; when the target frame is located on the right side of the vehicle and turns right, the left side boundary line of the target frame is corrected by the left side convex hull points in the left side convex hull point set of the target frame. Therefore, the present application determines the convex hull of the target box based on the target drivable points within the target box, which can reduce the amount of calculation and improve the efficiency of convex hull calculation, while supplementing more complete information for the deep learning target box; in addition, the convex hull points in the convex hull within the target box are used to correct the boundary line of the target box, which can reduce the information deviation of part of the target box output by the model or traditional clustering fitting, and provide more accurate target information for the back end. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0064] Figure 1 A flowchart of a target post-processing method based on a point cloud within a frame is shown as an embodiment of the present application;

[0065] Figure 2 A schematic diagram of a target recognition result in a target post-processing method based on an in-frame point cloud according to an embodiment of the present application;

[0066] Figure 3A schematic diagram of a convex hull of a target box in a target post-processing method based on a point cloud within a box, shown as an embodiment of the present application;

[0067] Figure 4 A schematic diagram of optimizing a target frame in a target post-processing method based on a point cloud within a frame, shown as an embodiment of the present application;

[0068] Figure 5 Another flow chart of a method for post-processing a target based on a point cloud within a frame according to an embodiment of the present application;

[0069] Figure 6 A schematic diagram of a target post-processing system based on an in-frame point cloud is shown as an embodiment of the present application. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0071] refer to Figure 1 , Figure 1 The following is a schematic diagram of a target post-processing method based on a point cloud within a frame according to an embodiment of the present application. Figure 1 As shown, the method includes:

[0072] Step S1: Perform target recognition on point cloud data through a target recognition model to obtain multiple target boxes.

[0073] In this embodiment, the point cloud data of the current driving scene is collected by the radar mounted on the vehicle, and then the collected point cloud data of the current frame is input into the target recognition model for target recognition, so as to obtain the target frame of each target in the current driving scene. The target frame corresponds to the target, and the target refers to various objects in the current driving scene, including cars, bicycles, motorcycles, pedestrians, roadblocks, etc. The radar that collects point cloud data includes but is not limited to laser radar, millimeter wave radar, continuous wave radar, pulse radar, etc. If there are two vehicles in front of the radar, the target frames corresponding to the two vehicles will be obtained by inputting the collected point cloud data into the target recognition model for target recognition. It should be understood that this is just a simple explanation for the convenience of understanding the meaning of the target frame, and the target frame actually obtained by the target recognition model for target recognition processing may contain misidentification, such as a non-existent target may be identified as a target. Among them, the target recognition model is a radar model, which is used to perform target recognition on the point cloud data to obtain a target frame corresponding to the target, such as Figure 2As shown, Figure 2 The target frame recognition result obtained by performing target recognition on point cloud data through the target recognition model is shown as an example, wherein the green part is the target frame recognition result obtained by the target recognition model.

[0074] Step S2: determining the target drivable area point in the target frame by analyzing and processing the point cloud data.

[0075] In this embodiment, the target frame recognition results output by the target recognition model for target recognition all have a certain degree of deviation. Therefore, in order to avoid the influence of the target frame corresponding to the vehicle target with recognition deviation on the driving of the vehicle when the vehicles around the vehicle (i.e., the vehicle target) turn, the target frame in the target frame recognition results output by the target recognition model will be optimized.

[0076] Specifically, the collected point cloud data of the current frame is analyzed and processed to determine the freespace points in the point cloud data, that is, the drivable area points. The drivable area points (i.e., freespace points) refer to barrier-free flat road area points that can be provided for vehicles to travel freely within a certain height range without considering traffic regulations. There are many ways to implement the freespace points in the point cloud data, and they are all relatively mature determination methods, which will not be described in detail here. Then, based on the positional relationship between the positions of all the determined drivable area points and each target frame, each drivable area point in the target frame is determined, and each drivable area point in the target frame is determined as the target drivable area point.

[0077] Step S3: According to the target drivable area points in the target frame, determine the convex hull of the target frame and the convex hull point set corresponding to the convex hull, and the convex hull points in the convex hull point set are all the target drivable area points in the target frame.

[0078] In this embodiment, after determining and obtaining each target drivable area point located in the target frame, for each target frame, the target drivable area point located in the target frame is used as the boundary, and the convex hull of the target frame is determined by a corresponding algorithm. The convex hull will include all target drivable area points in the target frame. The set consisting of all target drivable area points in the convex hull is determined as the convex hull point set of the convex hull corresponding to the target frame, and a convex hull point in the convex hull point set corresponds to a target drivable area point in the convex hull. Among them, the algorithm for calculating and determining the convex hull of the target frame is preferably the Graham scanning algorithm. It should be understood that this is only a preferred algorithm, and the algorithm for determining the convex hull of the target frame can also be other algorithms, which are not specifically limited here.

[0079] Step S4: Determine the maximum longitudinal convex hull point, the maximum transverse convex hull point, and the minimum transverse convex hull point in the convex hull point set of the target frame.

[0080] In this embodiment, for ease of description and understanding, the coordinate system of each target frame is the x-axis along the longitudinal direction of its own target frame, and the y-axis in the horizontal direction. After obtaining the convex hull point sets corresponding to each target frame through step S3, for each target frame, determine the convex hull point with the largest coordinate value in the longitudinal direction in the convex hull set corresponding to itself (i.e., the largest convex hull point in the longitudinal direction), and determine the convex hull point with the largest coordinate value in the horizontal direction in the convex hull set corresponding to itself (i.e., the largest convex hull point in the horizontal direction), and determine the convex hull point with the smallest coordinate value in the horizontal direction in the convex hull set corresponding to itself (i.e., the smallest convex hull point in the horizontal direction). Thus, for each target frame, it is possible to determine a maximum convex hull point in the longitudinal direction, a maximum convex hull point in the horizontal direction, and a minimum convex hull point in the horizontal direction corresponding to itself. Figure 3 As shown, Figure 3 The convex hull corresponding to a target frame and the convex hull point set corresponding to the convex hull and the longitudinal maximum convex hull point, transverse maximum convex hull point and transverse minimum convex hull point in the determined convex hull point set are schematically shown. The rectangle in the figure is the target frame, the dotted line enclosed area is the convex hull corresponding to the target frame, the dots in the convex hull are the convex hull points, and all the convex hull points in the convex hull constitute the convex hull point set corresponding to the target frame. X_max in the figure is the longitudinal maximum convex hull point in the convex hull point set, Y_max is the transverse maximum convex hull point in the convex hull point set, and Y_min is the transverse minimum convex hull point in the convex hull point set.

[0081] Step S5: When the target frame is located on the left side of the vehicle, the right side convex hull point set of the target frame is constructed by sequentially selecting the convex hull points in the target frame from the maximum convex hull point in the longitudinal direction and the minimum convex hull point in the transverse direction, and the right side convex hull point whose longitudinal distance from the minimum convex hull point in the transverse direction meets the first condition.

[0082] In this embodiment, since the main purpose of this application is to avoid the influence of the target frame corresponding to the vehicle target with identification deviation on the driving of the vehicle when the vehicles around the vehicle (i.e., the vehicle target) turn, and when the target frame is located on the left side of the vehicle, the main thing that may be affected by the driving of the vehicle is the right side boundary of the target frame. Therefore, when the target frame to be optimized is located on the left side of the vehicle, the convex hull point scanning order is screened between the longitudinal maximum convex hull point and the transverse minimum convex hull point of the target frame, and the convex hull point whose longitudinal distance with the transverse minimum convex hull point in the target frame meets the first condition, and then the screened convex hull point is determined as the right side convex hull point of the target frame, and stored in the right side convex hull point set corresponding to the target frame. The determination method of the right side convex hull point set corresponding to each target frame located on the left side of the vehicle is the same, which will not be repeated here. The right side convex hull point set of the target frame is used to optimize the target frame based on the right side convex hull point set if the target frame needs to be optimized. Among them, the first condition is preferably that the longitudinal distance between the convex hull point and the horizontal minimum convex hull point of the target frame is greater than half of the longitudinal size of the target frame (that is, half of the length of the target frame in the longitudinal direction).

[0083] In this embodiment, the point cloud data is recorded and stored in the scanning order during the laser radar scanning. In this embodiment, when the target frame to be optimized is located on the left side of the vehicle, the maximum longitudinal convex hull point and the minimum transverse convex hull point of the target frame are used as the boundary, which refers to the convex hull points in the convex hull point set of the target frame whose recording and storage order is located after the maximum longitudinal convex hull point of the target frame and before the minimum transverse convex hull point. That is, when the convex hull points that meet the first condition are subsequently determined, the convex hull points in the convex hull point set of the target frame whose recording and storage order is located after the maximum longitudinal convex hull point of the target frame and before the minimum transverse convex hull point are targeted.

[0084] Step S6: When the target frame is located on the right side of the vehicle, the left convex hull point set of the target frame is constructed by sequentially placing the convex hull points in the target frame among the longitudinal maximum convex hull point and the transverse maximum convex hull point, and the left convex hull point whose longitudinal distance from the transverse maximum convex hull point meets the second condition.

[0085] In this embodiment, when the target frame is located on the right side of the vehicle, the left boundary of the target frame may be mainly affected by the driving of the vehicle. Therefore, when the target frame to be optimized is located on the right side of the vehicle, the convex hull point scanning order is screened to be between the longitudinal maximum convex hull point and the transverse maximum convex hull point of the target frame, and the convex hull point whose longitudinal distance from the transverse maximum convex hull point in the target frame meets the second condition, and then the screened convex hull point is determined as the left convex hull point of the target frame and stored in the left convex hull point set corresponding to the target frame. The determination method of the left convex hull point set corresponding to each target frame located on the right side of the vehicle is the same, which will not be repeated here. The left convex hull point set of the target frame is used to optimize the target frame based on the left convex hull point set if the target frame needs to be optimized. Among them, the second condition is preferably that the longitudinal distance between the convex hull point and the transverse maximum convex hull point of the target frame is greater than half of the longitudinal size of the target frame (that is, half of the length of the target frame along the longitudinal direction).

[0086] In this embodiment, the point cloud data is recorded and stored in the scanning order during the laser radar scanning. In this embodiment, when the target frame to be optimized is located on the right side of the vehicle, the maximum longitudinal convex hull point and the maximum transverse convex hull point of the target frame are used as the boundary, which refers to the convex hull points in the convex hull point set of the target frame that are located after the maximum longitudinal convex hull point of the target frame and before the maximum transverse convex hull point of the target frame in the recording and storage order. That is, when the convex hull points that meet the second condition are subsequently determined, the convex hull points in the convex hull point set of the target frame that are located after the maximum longitudinal convex hull point of the target frame and before the maximum transverse convex hull point of the target frame in the recording and storage order are targeted.

[0087] Step S7: When the target frame is located at the left side of the vehicle and turns left, the right boundary line of the target frame is corrected using the right convex hull points in the right convex hull point set of the target frame.

[0088] In this embodiment, the position relationship of the target frame relative to the vehicle is determined. When the position relationship between the target frame and the vehicle is that the target frame is located on the left side of the vehicle, it is further determined whether the target frame is turning left. When the target frame located on the left side of the vehicle turns left and the original right boundary line of the target frame invades the lane where the vehicle is located, the right boundary line of the target frame is corrected to the inside of the target frame with the right convex hull point in the right convex hull point set of the target frame to obtain a more accurate target frame. The corrected target frame may no longer invade the lane where the vehicle is located, so that the target frame that was originally identified with deviation will no longer affect the driving of the vehicle after correction, so that the vehicle can better perform control actions. For example, in an actual scenario, the vehicle target does not actually invade the lane where the vehicle is located (that is, it will not affect the normal driving of the vehicle), but because the target frame corresponding to the vehicle target obtained by the target recognition model through identification has a deviation, the identified target frame invades the lane where the vehicle is located, resulting in the vehicle target that originally would not affect the normal driving of the vehicle, because the recognition result is determined that the vehicle target will invade the lane where the vehicle is located, thereby affecting the normal driving of the vehicle, which in turn causes the subsequent control actions of the vehicle to have deviations (for example, the vehicle could originally drive normally along the lane where the vehicle is located and overtake the vehicle in front, but because of the recognition result, the vehicle cannot drive normally and overtake). The method of the present application is used to correct the target frame in the recognition result so that the target frame can be closer to the real target, so that the vehicle can better perform control actions (such as overtaking the vehicle target in front along the lane where the vehicle is located normally).

[0089] In this embodiment, one implementation method for determining whether the target frame on the left side of the vehicle is turning left is to determine the angle between the orientation of the target frame on the left side of the vehicle and the orientation of the vehicle. When the angle is greater than zero, it is determined that the target frame on the left side of the vehicle is turning left. Figure 3 As shown), when the angle between the orientation of the target frame on the left side of the vehicle and the orientation of the vehicle is greater than zero, it is determined that the target frame on the left side of the vehicle turns left.

[0090] Step S8: When the target frame is located at the right side of the vehicle and turns right, the left boundary line of the target frame is corrected using the left convex hull points in the left convex hull point set of the target frame.

[0091] In this embodiment, the position relationship of the target frame relative to the vehicle is determined. When the position relationship between the target frame and the vehicle is that the target frame is located on the right side of the vehicle, it is further determined whether the target frame is turning right. When the target frame located on the right side of the vehicle turns right and the original left boundary line of the target frame invades the lane where the vehicle is located, the left boundary line of the target frame is corrected to the inside of the target frame with the left convex hull point in the left convex hull point set of the target frame to obtain a more accurate target frame. Figure 4 As shown, Figure 4 The green box in the middle corresponds to a target box, and the orange box is a target box obtained after the target box is corrected.

[0092] The embodiment of the present application provides a target post-processing method based on in-frame point cloud. First, target recognition is performed on point cloud data through a target recognition model to obtain multiple target frames; target drivable area points located in the target frame are determined by analyzing and processing the point cloud data; the convex hull of the target frame and the convex hull point set corresponding to the convex hull are determined according to the target drivable area points in the target frame, and the convex hull points in the convex hull point set are all the target drivable area points in the target frame; the maximum longitudinal convex hull point, the maximum transverse convex hull point and the minimum transverse convex hull point in the convex hull point set of the target frame are determined; when the target frame is located on the left side of the vehicle, the convex hull points in the target frame are located in the maximum longitudinal convex hull point and the minimum transverse convex hull point in the order of the convex hull points in the target frame. , the right convex hull point set of the target frame is constructed by the right convex hull points whose longitudinal distance from the minimum convex hull point in the horizontal direction meets the first condition; when the target frame is located on the right side of the vehicle, the left convex hull point set of the target frame is constructed by the convex hull points in the target frame that are sequentially located in the longitudinal maximum convex hull point and the transverse maximum convex hull point, and whose longitudinal distance from the maximum convex hull point in the transverse direction meets the second condition; when the target frame is located on the left side of the vehicle and turns left, the right side boundary line of the target frame is corrected by the right side convex hull points in the right side convex hull point set of the target frame; when the target frame is located on the right side of the vehicle and turns right, the left side boundary line of the target frame is corrected by the left side convex hull points in the left side convex hull point set of the target frame. Therefore, the present application determines the convex hull of the target box based on the target drivable points within the target box, which can reduce the amount of calculation and improve the efficiency of convex hull calculation, while supplementing more complete information for the deep learning target box; in addition, the convex hull points in the convex hull within the target box are used to correct the boundary line of the target box, which can reduce the information deviation of part of the target box output by the model or traditional clustering fitting, and provide more accurate target information for the back end.

[0093] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a target post-processing method based on a point cloud within a frame. In the target post-processing method based on a point cloud within a frame, step S2 may include: determining a drivable area point in the point cloud data by analyzing and processing the point cloud data; and determining a target drivable area point located in the target frame according to the location of the target frame and the location of the drivable area point.

[0094] In this embodiment, the collected point cloud data of the current frame is analyzed and processed to determine the drivable area points in the point cloud data. Then, based on the center point and size information of each target frame, coordinate transformation is used to calculate the coordinates of the four corner points of each target frame in the xy plane. Then, each target frame obtains its maximum and minimum horizontal and maximum and minimum vertical ranges on the xy plane based on the coordinates of its four corner points. Then, based on the coordinates of the drivable area point, it is determined whether it falls into any target frame. If the coordinates of the drivable area point fall into any target frame, the drivable area point is determined as the target drivable area point in the target frame and stored for use in subsequent steps.

[0095] In combination with the above embodiments, in one implementation, the present application embodiment further provides a target post-processing method based on a point cloud within a frame. In the target post-processing method based on a point cloud within a frame, before step S2, the method further includes steps S01 to S02:

[0096] Step S01: determining a virtual target frame among the multiple target frames according to the attribute information of the target frame.

[0097] In this embodiment, as mentioned in the above step S1, there may be misidentification in the target frame obtained by the target recognition model through the target recognition process, such as a non-existent target may be recognized as a target. Therefore, in order to avoid the impact of the target frame originally obtained by misidentification on the driving of the vehicle, and the resource consumption caused by optimizing the target frame originally obtained by misidentification, the present application screens the various target frames obtained by the target recognition model through the recognition process of the point cloud data, determines the virtual target frames that are misidentified, and then filters them out. The subsequent optimization is only for the target frames remaining after filtering out the virtual target frames. The present application mainly determines whether each target frame is a virtual target based on the point cloud reflectivity information in the attribute information of each target frame, the confidence of the target frame, the positional relationship between the target frames, and the point cloud distribution in the target frame.

[0098] Specifically, in scenes with dense dust and dense water vapor near a sprinkler truck, if the proportion of low-reflectivity point clouds in the target frame reaches a certain threshold, and the average reflectivity information is lower than a certain set threshold and the maximum reflectivity information is lower than a certain set threshold, the target frame is determined to be a virtual target. In order to solve the problem of low accuracy of the target frame obtained by the target recognition model due to incomplete point clouds, the low-credibility target frame is determined as a virtual target frame based on the position of the target frame, the orientation of the target frame, and the distribution of the point cloud in the target frame.

[0099] Step S02: Filter the virtual target frames in the multiple target frames to obtain the remaining target frames.

[0100] In this embodiment, the virtual target frames in all the target frames determined in step S01 are filtered to obtain the remaining target frames.

[0101] In the present application, when the method further includes steps S01 to S02, step S2 may include: determining the target drivable area point located in the target frame remaining after filtering by analyzing and processing the point cloud data.

[0102] In this embodiment, before step S2, it also includes steps S01 to S02 to filter the virtual target frames in all the obtained target frames, and the implementation method of step S2 will be: by analyzing and processing the point cloud data, determine the drivable area points in the point cloud data, and then determine which drivable area points are located in the target frame remaining after filtering, and finally determine the drivable area points located in the target frame remaining after filtering as the target drivable area points. The subsequent steps of determining the convex hull point set, determining the right convex hull point set, determining the left convex hull point set, and correcting and optimizing the boundary line are all determined and corrected based on the target drivable area points located in the target frame remaining after filtering and the target frame remaining after filtering.

[0103] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a target post-processing method based on a point cloud within a frame. In the target post-processing method based on a point cloud within a frame, before step S3, the method further includes: determining a target frame that needs to be oriented corrected from all target frames according to the attribute information of the target frame.

[0104] In this embodiment, the target frame recognition results output by the target recognition model for target recognition have a certain degree of deviation, and this deviation is only when the vehicle target in the adjacent lane invades the lane where the vehicle is located, and generally only large vehicles will have such an invasion. Therefore, in order to improve processing efficiency and reduce the consumption of computing resources, this application only determines whether correction is required and performs subsequent correction for target frames that are relatively large and / or in a turning state.

[0105] Specifically, before step S3, if the present application is directed to a solution in which virtual target frame filtering is not performed, the target frame that needs to be corrected in orientation is determined from all target frames according to the category, size and / or orientation in the attribute information of the target frame. If the present application is directed to a solution in which virtual target frame filtering is performed, the target frame that needs to be corrected in orientation is determined from all target frames remaining after filtering according to the category, size and / or orientation in the attribute information of the target frame. An optional implementation method for determining a target frame that needs to be corrected in orientation is to determine whether the target frame is a BIG_VEHICLE category or a HUGE_VEHICLE category, determine whether the length of the target frame exceeds a certain threshold (such as 6 meters), determine whether the width of the target frame exceeds a certain threshold (such as 2 meters), and determine whether the orientation of the target frame is within a range of plus or minus 25°. As long as any one of the items is met, it is determined that the target frame belongs to a target frame that needs to be corrected in orientation.

[0106] In the present application, before step S3 of a target post-processing method based on an in-frame point cloud provided in the present application, the target frame that needs to be corrected in orientation is determined from all target frames according to the attribute information of the target frame. Step S3 may include: determining the convex hull of the target frame that needs to be corrected in orientation and the set of convex hull points corresponding to the convex hull according to the target drivable area points in the target frame that needs to be corrected in orientation.

[0107] In this embodiment, in the case where the present application determines the target frame that needs to be corrected in direction, subsequent correction operations will be directed to the target frame that needs to be corrected in direction. Therefore, in this case, in this implementation, the present application will only distinguish between the convex hull and the convex hull point set for the target frame that needs to be corrected in direction. The determination method is the same as the implementation method of the above-mentioned step S3, except that this time it is based on the target drivable area point of the target frame that needs to be corrected in direction, and the convex hull of the target frame that needs to be corrected in direction and the convex hull point set corresponding to the convex hull are determined.

[0108] In combination with the above embodiments, in one implementation, the present application embodiment further provides a target post-processing method based on a point cloud within a frame. In the target post-processing method based on a point cloud within a frame, step S7 may include:

[0109] Step S71: When the target frame is located on the left side of the vehicle and turns left, traverse the right convex hull points in the right convex hull point set of the target frame, and connect the currently traversed right convex hull point with the lateral minimum convex hull point to form a right predicted boundary line.

[0110] In this embodiment, when the target frame is located on the left side of the vehicle and the target frame turns left, if there is an error in model recognition and its right boundary line invades the lane where the vehicle is located, it will affect the driving of the vehicle (if it does not actually invade, it is because of an error in recognition that the recognition result is an invasion). In this case, the present application will traverse the right convex hull points in the right convex hull point set of the target frame, take out one convex hull point in the right convex hull point set at a time, and then connect the convex hull point with the lateral minimum convex hull point of the target frame to form a new right predicted boundary line.

[0111] Step S72: When the angle deviation between the obtained right predicted boundary line and the right boundary line of the target frame satisfies the set condition, the traversal is terminated and the right boundary line of the target frame is corrected with the right predicted boundary line whose deviation satisfies the set condition.

[0112] In this embodiment, after a new right predicted boundary line of the target frame is formed by step S71, the right predicted boundary line is compared with the right boundary line of the target frame (i.e., the right boundary point of the target frame obtained by model recognition) to determine whether the angle deviation between the two meets the set conditions. If the set conditions are met, the right boundary line of the target frame is replaced by the new right predicted boundary line, so as to achieve the purpose of correction, and the traversal is terminated at this time without further traversal. If the set conditions are not met, a new convex hull point is continuously taken out from the right convex hull point set of the target frame, and the newly taken convex hull point is also connected with the lateral minimum convex hull point of the target frame to form a new right predicted boundary line, and it is determined whether the angle deviation between the new right predicted boundary line and the right boundary line of the target frame meets the set conditions. If it is still not met, it is continued to be taken until the angle deviation between a new convex hull point and the right boundary line of the target frame can meet the set conditions. Among them, the preset condition is preferably that the angle deviation does not exceed 10°, which is to prevent the target frame from being over-corrected and causing more serious errors.

[0113] In combination with the above embodiments, in one implementation, the present application embodiment further provides a target post-processing method based on a point cloud within a frame. In the target post-processing method based on a point cloud within a frame, step S8 may include:

[0114] Step S81: When the target frame is located on the right side of the vehicle and turns right, traverse the left convex hull points in the left convex hull point set of the target frame, and connect the currently traversed left convex hull point with the horizontal maximum convex hull point to form a left predicted boundary line.

[0115] In this embodiment, when the target frame is located on the right side of the vehicle and the target frame turns right, if there is an error in model recognition and its left boundary line invades the lane where the vehicle is located, it will affect the driving of the vehicle (if it does not actually invade, it is because of an error in recognition that the recognition result is an invasion). In this case, the present application will traverse the left convex hull points in the left convex hull point set of the target frame, take out one convex hull point from the left convex hull point set at a time, and then connect the convex hull point with the lateral maximum convex hull point of the target frame to form a new left predicted boundary line.

[0116] Step S82: When the angle deviation between the obtained left predicted boundary line and the left boundary line of the target frame satisfies the set condition, the traversal is terminated and the left boundary line of the target frame is corrected with the left predicted boundary line whose deviation satisfies the set condition.

[0117] In this embodiment, after a new left predicted boundary line of the target frame is formed by step S81, the left predicted boundary line is compared with the left boundary line of the target frame (i.e., the left boundary point of the target frame obtained by model recognition) to determine whether the angle deviation between the two meets the set conditions. If the set conditions are met, the left boundary line of the target frame is replaced with the new left predicted boundary line, so as to achieve the purpose of correction, and the traversal is terminated at this time. If the set conditions are not met, a new convex hull point is continuously taken out from the left convex hull point set of the target frame, and the newly taken convex hull point is also connected with the horizontal maximum convex hull point of the target frame to form a new left predicted boundary line, and it is determined whether the angle deviation between the new left predicted boundary line and the left boundary line of the target frame meets the set conditions. If it is still not met, it is continued to be taken until the angle deviation between a new convex hull point and the left boundary line of the target frame can meet the set conditions. Among them, the preset condition is preferably that the angle deviation does not exceed 10°, which is to prevent the target frame from being over-corrected and causing more serious errors.

[0118] In combination with the above embodiments, in one implementation, the present application embodiment further provides a target post-processing method based on a point cloud within a frame. In the target post-processing method based on a point cloud within a frame, the method further includes:

[0119] Step S101: clustering the point cloud data using a clustering algorithm to obtain cluster target frames.

[0120] In this embodiment, in the implementation method of processing point cloud data to obtain a target frame, the laser radar model can only recognize the target categories involved when training the model. Therefore, in actual target frame recognition, traditional clustering algorithms are usually used to recognize the target frame. The recognition method of traditional clustering algorithms will also involve the recognition of some virtual targets. Based on this, a target post-processing algorithm based on the point cloud in the frame provided in this application is also used to filter the virtual clustering target frame of the clustered target frame obtained by traditional clustering algorithm recognition based on the attribute information of the point cloud in the frame.

[0121] Specifically, the point cloud data is first clustered using a traditional clustering algorithm to obtain each cluster target frame. Figure 2 The clustering target frame recognition result obtained by clustering point cloud data using a traditional clustering algorithm is shown as an example, wherein the red part is the clustering target frame recognition result obtained by clustering using the traditional clustering algorithm.

[0122] Step S102: Determine the proportion of point clouds of different semantic categories in the clustering target frame, determine the size of the clustering target frame, determine the reflectivity information of the point cloud in the clustering target frame, and determine the first point cloud density of the point cloud within a set range of the centroid of the clustering target frame.

[0123] In this embodiment, after the point cloud data is clustered by a traditional clustering algorithm to obtain each cluster target frame, the proportion of point clouds of different semantic categories in each frame is determined for each cluster target frame, as well as the size (i.e., length, width, and height) of each frame, the reflectivity information of the point cloud in each frame, and the point cloud density of the point cloud within a set range of the centroid of the cluster target frame are determined, which is the first point cloud density. The set range can be dynamically adjusted according to the specific size of the cluster target frame, so as to improve the filtering accuracy of the virtual cluster target frame. If the cluster target frame is relatively large, the corresponding set range will be taken as a larger range value, and if the cluster target frame is relatively small, the corresponding set range will be taken as a smaller range value.

[0124] The semantic categories include at least: invalid points (including noise and ground), background, vehicles, cyclists, general obstacles, etc. The reflectivity information of the point cloud includes at least: the number of point clouds with low reflectivity in the cluster target box, the maximum reflectivity and total reflectivity of the point cloud in the cluster target box, the proportion of low reflectivity point clouds in the cluster target box, and the average reflectivity of the point cloud in the cluster target box.

[0125] Step S103: Divide the cluster target frame into multiple segments in the vertical direction, and calculate the second point cloud density around the center point of the first segment in the moving direction of the cluster target frame.

[0126] In this embodiment, for each clustering target frame obtained by clustering the point cloud data through the traditional clustering algorithm, each clustering target frame is divided into multiple segments (preferably 4 segments) in its longitudinal direction, and the point cloud density around the center point of the first segment in the direction of movement of the clustering target frame itself (such as the center point of the first 1 / 4 segment) is calculated, and this point cloud density is called the second density.

[0127] Step S104: determining a virtual cluster target frame in each cluster target frame based on the point cloud proportion, the size, the reflectivity information, the first point cloud density, and the second point cloud density.

[0128] In this embodiment, based on the point cloud proportion, size, reflectivity information, first point cloud density and second point cloud density of each cluster target frame determined in steps S101 to S13, it is determined which of these cluster target frames are virtual cluster target frames.

[0129] Step S105: filtering the virtual clustering target frames in the clustering target frames to obtain a clustering result of the clustering algorithm.

[0130] In this embodiment, the clustering target frames determined in step S104 are filtered, and the remaining clustering target frames are the final clustering results obtained by the clustering process of the clustering algorithm.

[0131] In combination with the above embodiments, in one implementation, the embodiment of the present application also provides a target post-processing method based on in-frame point clouds. In the target post-processing method based on in-frame point clouds, the reflectivity information of the point clouds in the clustered target frame is determined, including: determining the number of point clouds with low reflectivity in the clustered target frame according to the attribute information of the point clouds in the clustered target frame, and determining the maximum reflectivity and the total reflectivity value of the point clouds in the clustered target frame; determining the proportion of low reflectivity point clouds in the clustered target frame according to the total number of point clouds in the clustered target frame and the number of point clouds with low reflectivity; determining the average reflectivity of the point clouds in the clustered target frame according to the total number of point clouds in the clustered target frame and the total reflectivity value.

[0132] In this embodiment, the collected point cloud data records the reflectivity of the point clouds, and the point clouds with reflectivity lower than the set value are determined based on the reflectivity of the point cloud data, and these point clouds are determined as low reflectivity point clouds, and then the number of low reflectivity point clouds in each cluster target frame is determined, wherein the set value is preferably 10. At the same time, the total reflectivity value and the maximum reflectivity of the point cloud in each cluster target frame are determined based on the reflectivity of the point cloud data. The number of ground reflectivity point clouds in a single cluster target frame is divided by the total number of point clouds in the single cluster target frame to obtain the proportion of low reflectivity point clouds in the single cluster target frame. Based on the same implementation method, each cluster target frame will obtain a corresponding ground reflectivity point cloud proportion. The total reflectivity value in a single cluster target frame is divided by the total number of point clouds in the single cluster target frame to obtain the average reflectivity of the point clouds in the single cluster target frame.

[0133] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a target post-processing method based on in-frame point clouds. In the target post-processing method based on in-frame point clouds, step S104 may include: when the proportion of low-reflectivity point clouds corresponding to the cluster target frame is a set value, determining that the cluster target frame is a virtual cluster target frame; when the proportion of low-reflectivity point clouds corresponding to the cluster target frame exceeds the first reflectivity threshold and the corresponding average reflectivity does not exceed the second reflectivity threshold, determining that the cluster target frame is a virtual cluster target frame; when the proportion of low-reflectivity point clouds corresponding to the cluster target frame exceeds the third reflectivity threshold and the corresponding average reflectivity does not exceed the fourth reflectivity threshold and the corresponding maximum reflectivity does not exceed the fifth reflectivity threshold, determining that the cluster target frame is a virtual cluster target frame; when the proportion of point clouds of noise semantic categories in the cluster target frame exceeds the category threshold, determining that the cluster target frame is a virtual cluster target frame. Determine the size relationship between the size and a first size threshold, determine the density relationship between the first point cloud density and the first density threshold, and determine the quantity relationship between the total number of point clouds and the total number threshold; determine whether the clustering target box is a virtual clustering target box based on the size relationship, the density relationship and the quantity relationship; when the proportion of point clouds of the noise semantic category in the clustering target box does not exceed the category threshold, determine whether the clustering target box is a virtual clustering target box based on the proportion of point clouds of the background semantic category in the clustering target box, the size, and the total number of point clouds, and determine whether the clustering target box is a virtual clustering target box based on the proportion of point clouds of the general obstacle semantic category in the clustering target box, the position of the clustering target box and the size.

[0134] In this embodiment, if the proportion of low reflectivity point clouds corresponding to a clustering target frame obtained by a traditional clustering algorithm is a set value, the clustering target frame is determined to be a virtual clustering target frame and will be filtered out, wherein the set value is preferably 1. If the proportion of low reflectivity point clouds corresponding to a clustering target frame obtained by a traditional clustering algorithm exceeds the first reflectivity threshold (preferably 0.9) and the corresponding average reflectivity does not exceed the second reflectivity threshold (preferably 3), the clustering target frame is determined to be a virtual clustering target frame and will be filtered out. If the proportion of low reflectivity point clouds corresponding to a clustering target frame obtained by a traditional clustering algorithm exceeds the third reflectivity threshold (preferably 1) and the corresponding average reflectivity does not exceed the fourth reflectivity threshold (preferably 0.5) and the corresponding maximum reflectivity does not exceed the fifth reflectivity threshold (preferably 3), the clustering target frame is determined to be a virtual clustering target frame and will be filtered out. The judgment conditions in this part are mainly aimed at the situation where water vapor, dust and other noises are mistakenly identified as target frames. The reflectivity of water vapor, dust and other noises is generally low. For water vapor virtual targets near sprinkler trucks, dust virtual targets at the rear of the vehicle, water vapor virtual targets on rainy days, vehicle exhaust virtual targets, etc., this application filters out dust, water vapor and other noise virtual targets based on the reflectivity information of the point cloud in the frame.

[0135] In this embodiment, since semantic segmentation may cause a problem of misdetecting a general obstacle as noise or misdetecting noise as a general obstacle, for such situations, the present application determines a virtual clustering target frame based on information such as the position and size of the target frame, the number of point clouds in the frame, the point cloud density around the center point and the centroid of the target frame, and the proportion of point clouds of each semantic category in the frame. Specifically, if the proportion of point clouds of the noise semantic category in a clustering target frame obtained by a traditional clustering algorithm exceeds a category threshold (preferably 0.6), the size relationship between the size of the clustering target frame and the first size threshold is determined, as well as the density relationship between the first point cloud density of the clustering target frame and the first density threshold, and the quantitative relationship between the total number of point clouds of the clustering target frame and the total number threshold is determined, and then, based on the size relationship, the density relationship, and the quantitative relationship corresponding to the clustering target frame, it is determined whether the clustering target frame is a virtual clustering target frame. Among them, the first size threshold includes the length threshold and width threshold of the target frame, and the length threshold and width threshold can be set according to the actual application scenario, and are not specifically limited here; the first density threshold can be set according to the actual application scenario, and are not specifically limited here; the total number threshold can be set according to the actual application scenario, and are not specifically limited here. If the size relationship is that the size of the cluster target frame is not within the first size threshold range, and / or the density relationship is that the first point cloud density of the cluster target frame is greater than the first density threshold, and / or the quantity relationship is that the total number of point clouds of the cluster target frame is greater than the total number threshold, then the cluster target frame is determined to be a virtual cluster target frame and will be filtered out, that is, if at least one of the three relationships is satisfied, the cluster target frame will be determined to be a virtual cluster target frame. If the proportion of point clouds of the noise semantic category in a cluster target frame obtained by the traditional clustering algorithm does not exceed the category threshold (preferably 0.6), determine whether the cluster target frame is a virtual cluster target frame based on the proportion of point clouds of the background semantic category (or noise and background semantic category) in the cluster target frame, the size of the cluster target frame, and the total number of point clouds in the cluster target frame. Also, determine whether the cluster target frame is a virtual cluster target frame based on the proportion of point clouds of the general obstacle semantic category in the cluster target frame, the position of the cluster target frame, and the size of the cluster target frame.Specifically, if the proportion of background semantic categories (or background and noise semantic categories) of the point cloud in the cluster target frame exceeds a certain threshold (preferably 0.9), and the size of the cluster target frame exceeds a certain set threshold (that is, the size of the cluster target frame is large) and the number of point clouds in the target cluster target frame is lower than a certain threshold (or the height of the cluster target frame from the ground exceeds a certain threshold), then the cluster target frame is determined to be a virtual cluster target frame and will be filtered out; if the proportion of general obstacle semantic categories of the point cloud in the cluster target frame exceeds a certain threshold (preferably 0.9), and the cluster target frame is not located in the current lane and the size of the cluster target frame exceeds a certain threshold (that is, the size is large), then the cluster target frame is determined to be a virtual cluster target frame and will be filtered out.

[0136] In combination with the above embodiments, in one implementation, the embodiment of the present application further provides a target post-processing method based on in-frame point cloud. In the target post-processing method based on in-frame point cloud, the method further includes: determining whether there is a small-sized single cluster target frame within a preset range around the vehicle category cluster target frame; if there is a small-sized single cluster target frame within a preset range around the vehicle category cluster target frame, determining that the vehicle category cluster target frame is a virtual cluster target frame.

[0137] In this embodiment, for each vehicle category clustering target frame obtained by the traditional clustering algorithm, it is determined whether there is a small-sized single clustering target frame within a preset range around the single vehicle category clustering target frame. If there is a small-sized single clustering target frame within the preset range of the single vehicle category clustering target frame, the single vehicle category clustering target frame is determined to be a virtual clustering target frame and will be filtered out. Among them, the preset range can be set according to the actual application scenario and is not specifically limited here. At the same time, the small size is set according to the actual application scenario and is not specifically limited here.

[0138] In this embodiment, if Figure 5As shown, an optional implementation method of a target post-processing method based on a point cloud within a frame provided by the present application is: processing the point cloud data by two processing methods, a target recognition model and a traditional clustering algorithm, and obtaining multiple target frames by identifying the target recognition model, and obtaining multiple target frames by clustering the traditional clustering algorithm. For the multiple target frames obtained by the target recognition model, the virtual target frames therein are filtered based on the reflectivity information of the point cloud in the target frame and the distribution information of the point cloud in the target frame. For the multiple target frames obtained by clustering with the traditional clustering algorithm, the number and proportion of point cloud data are obtained based on each semantic category in the target frame, as well as the reflectivity information of the point cloud in the target frame and the point cloud density information around the centroid and center point in the target frame, and the virtual target frames therein are filtered. After filtering, the multiple target frames obtained by the target recognition model are corrected for the boundary lines of the remaining target frames. First, the point cloud data is analyzed and processed to determine the target drivable area points in the target frames remaining after filtering. Then, based on the target drivable area points in the single target frame, the convex hull of the single target frame is determined. Finally, when the boundary lines of the single target frame need to be corrected, the boundary lines of the single target frame are corrected based on the convex hull.

[0139] Based on the same inventive concept, an embodiment of the present application provides a target post-processing system based on a point cloud within a frame, such as Figure 6 As shown, the system 600 includes:

[0140] The target recognition module 601 is used to perform target recognition on the point cloud data through a target recognition model to obtain multiple target frames;

[0141] An analysis and processing module 602 is used to determine a target drivable area point located in a target frame by analyzing and processing the point cloud data;

[0142] A convex hull point set determination module 603 is used to determine the convex hull of the target box and the convex hull point set corresponding to the convex hull according to the target drivable area points in the target box, wherein the convex hull points in the convex hull point set are all the target drivable area points in the target box;

[0143] The limit convex hull point determination module 604 is used to determine the maximum convex hull point in the longitudinal direction, the maximum convex hull point in the transverse direction, and the minimum convex hull point in the transverse direction in the convex hull point set of the target frame;

[0144] The right convex hull point set determination module 605 is used to construct the right convex hull point set of the target frame when the target frame is located on the left side of the vehicle, using the right convex hull points in the target frame whose convex hull points are sequentially located between the maximum convex hull point in the longitudinal direction and the minimum convex hull point in the transverse direction and whose longitudinal distance from the minimum convex hull point in the transverse direction meets the first condition;

[0145] The left convex hull point set determination module 606 is used to construct the left convex hull point set of the target frame when the target frame is located on the right side of the vehicle, using the left convex hull points of the target frame whose convex hull points are sequentially located in the longitudinal maximum convex hull point and the transverse maximum convex hull point and whose longitudinal distance from the transverse maximum convex hull point meets the second condition;

[0146] A first correction module 607 is used to correct the right boundary line of the target frame using the right convex hull points in the right convex hull point set of the target frame when the target frame is located on the left side of the vehicle and turns left;

[0147] The second correction module 608 is used to correct the left boundary line of the target frame with the left convex hull points in the left convex hull point set of the target frame when the target frame is located at the right side of the vehicle and turns right.

[0148] Optionally, the analysis and processing module 602 includes:

[0149] A drivable area point determination module, used to determine drivable area points in the point cloud data by analyzing and processing the point cloud data;

[0150] The target drivable area point determination module is used to determine the target drivable area point located in the target frame according to the location of the target frame and the location of the drivable area point.

[0151] Optionally, the system 600 further includes:

[0152] A virtual target frame determining module, used for determining a virtual target frame among the multiple target frames according to the attribute information of the target frame;

[0153] A filtering module, used for filtering virtual target frames in the multiple target frames to obtain remaining target frames;

[0154] The analysis and processing module 602 is used to determine the target drivable area points located in the target frame remaining after filtering by analyzing and processing the point cloud data.

[0155] Optionally, the system 600 further includes:

[0156] A target frame determination module is used to determine a target frame that needs to be oriented corrected from all target frames according to the attribute information of the target frame;

[0157] The convex hull point set determination module 603 is used to determine the convex hull of the target frame that needs to be oriented and the convex hull point set corresponding to the convex hull in the target drivable area point in the target frame that needs to be oriented.

[0158] Optionally, the first correction module 607 includes:

[0159] A traversal module, used for traversing the right convex hull points in the right convex hull point set of the target frame when the target frame is located on the left side of the vehicle and turns left, and connecting the currently traversed right convex hull point with the lateral minimum convex hull point to form a right predicted boundary line;

[0160] A first correction submodule is used to terminate the traversal and correct the right boundary line of the target frame with the right prediction boundary line whose deviation satisfies the set condition when the angle deviation between the obtained right prediction boundary line and the right boundary line of the target frame satisfies the set condition;

[0161] The second correction module 608 includes:

[0162] A traversal module, used for traversing the left convex hull points in the left convex hull point set of the target frame when the target frame is located on the right side of the vehicle and turns right, and connecting the currently traversed left convex hull point with the horizontal maximum convex hull point to form a left predicted boundary line;

[0163] The second correction submodule is used to end the traversal and correct the left boundary line of the target frame with the left predicted boundary line whose deviation satisfies the set condition when the angle deviation between the obtained left predicted boundary line and the left boundary line of the target frame satisfies the set condition.

[0164] Optionally, the system 600 further includes:

[0165] The clustering module is used to cluster the point cloud data through the clustering algorithm to obtain each cluster target frame;

[0166] A first attribute determination module is used to determine the proportion of point clouds of different semantic categories in the cluster target frame, determine the size of the cluster target frame, determine the reflectivity information of the point cloud in the cluster target frame, and determine the first point cloud density of the point cloud within a set range of the centroid of the cluster target frame;

[0167] A second attribute determination module is used to divide the cluster target frame into multiple segments in the longitudinal direction, and calculate a second point cloud density around a center point of a first segment in a direction of movement of the cluster target frame;

[0168] A virtual cluster target frame determination module, configured to determine a virtual cluster target frame in each cluster target frame based on the point cloud proportion, the size, the reflectivity information, the first point cloud density, and the second point cloud density;

[0169] The filtering module is used to filter the virtual clustering target frames in each clustering target frame to obtain the clustering result of the clustering algorithm.

[0170] Optionally, a reflectivity information determination module is used to determine the reflectivity information of the point cloud within the cluster target frame; the reflectivity information determination module includes:

[0171] A first reflectivity information determination module is used to determine the number of point clouds with low reflectivity in the cluster target frame according to the attribute information of the point cloud in the cluster target frame, and to determine the maximum reflectivity and the total reflectivity value of the point cloud in the cluster target frame;

[0172] A second reflectivity information determination module, used to determine the proportion of low reflectivity point clouds in the cluster target frame according to the total number of point clouds in the cluster target frame and the number of low reflectivity point clouds;

[0173] The third reflectivity information determination module is used to determine the average reflectivity of the point cloud in the cluster target frame according to the total number of point clouds and the total reflectivity value in the cluster target frame.

[0174] Optionally, a virtual clustering target frame determination module includes:

[0175] A first virtual cluster target frame determination module is used to determine that the cluster target frame is a virtual cluster target frame when the proportion of low reflectivity point clouds corresponding to the cluster target frame is a set value;

[0176] A second virtual clustering target frame determination module, configured to determine that the clustering target frame is a virtual clustering target frame when the proportion of the low reflectivity point cloud corresponding to the clustering target frame exceeds the first reflectivity threshold and the corresponding average reflectivity does not exceed the second reflectivity threshold;

[0177] A third virtual clustering target frame determination module is used to determine that the clustering target frame is a virtual clustering target frame when the proportion of low reflectivity point clouds corresponding to the clustering target frame exceeds a third reflectivity threshold and the corresponding average reflectivity does not exceed a fourth reflectivity threshold and the corresponding maximum reflectivity does not exceed a fifth reflectivity threshold;

[0178] a fourth virtual cluster target frame determination module, configured to determine, when the proportion of point clouds of the noise semantic category in the cluster target frame exceeds the category threshold, a size relationship between the size and a first size threshold, a density relationship between the first point cloud density and a first density threshold, and a quantity relationship between the total number of point clouds and a total number threshold;

[0179] A fifth virtual clustering target frame determination module, configured to determine whether the clustering target frame is a virtual clustering target frame according to the size relationship, the density relationship and the quantity relationship;

[0180] The sixth virtual clustering target frame determination module is used to determine whether the clustering target frame is a virtual clustering target frame according to the proportion of point clouds of background semantic categories in the clustering target frame, the size, and the total number of point clouds when the proportion of point clouds of noise semantic categories in the clustering target frame does not exceed the category threshold, and to determine whether the clustering target frame is a virtual clustering target frame according to the proportion of point clouds of general obstacle semantic categories in the clustering target frame, the position of the clustering target frame, and the size.

[0181] Optionally, the system 600 further includes:

[0182] A single cluster target frame determination module is used to determine whether there is a small-sized single cluster target frame within a preset range around the vehicle category cluster target frame;

[0183] The seventh virtual clustering target frame determining module is used to determine that the vehicle category clustering target frame is a virtual clustering target frame when there is a small-sized single clustering target frame within a preset range around the vehicle category clustering target frame.

[0184] Based on the same inventive concept, an embodiment of the present application provides an electronic device, comprising: a processor, a memory, and a computer program stored on the memory and running on the processor. When the computer program is executed by the processor, the steps in a target post-processing method based on an in-frame point cloud as described in the first aspect of the present application are implemented.

[0185] Based on the same inventive concept, an embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps in a target post-processing method based on an in-frame point cloud as described in the first aspect of the present application are implemented.

[0186] As for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0187] It should be noted that, for the method embodiments, for the sake of simplicity, they are all described as a series of action combinations, but those skilled in the art should be aware that the embodiments of the present application are not limited by the described order of actions, because according to the embodiments of the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily required by the embodiments of the present application.

[0188] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0189] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the embodiments of the present application may adopt the form of complete hardware embodiments, complete software embodiments, or embodiments in combination with software and hardware. Moreover, the embodiments of the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0190] The embodiments of the present application are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0191] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0192] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0193] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present application.

[0194] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0195] The above is a detailed introduction to the target post-processing method, system and product based on the in-frame point cloud provided by the present application. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A target post-processing algorithm based on point cloud within a frame, characterized in that: The method comprises: Perform target recognition on point cloud data through the target recognition model to obtain multiple target frames; By analyzing and processing the point cloud data, a target drivable area point located in the target frame is determined; According to the target drivable area points in the target frame, determine the convex hull of the target frame and the convex hull point set corresponding to the convex hull, wherein the convex hull points in the convex hull point set are all the target drivable area points in the target frame; Determine the maximum longitudinal convex hull point, the maximum transverse convex hull point, and the minimum transverse convex hull point in the convex hull point set of the target frame; When the target frame is located on the left side of the vehicle, the right convex hull point set of the target frame is constructed by sequentially selecting the right convex hull points of the target frame that are located in the maximum convex hull point in the longitudinal direction and the minimum convex hull point in the transverse direction and whose longitudinal distance from the minimum convex hull point in the transverse direction meets the first condition; When the target frame is located on the right side of the vehicle, the left convex hull point set of the target frame is constructed by sequentially selecting the left convex hull points of the target frame that are located in the longitudinal maximum convex hull point and the transverse maximum convex hull point and whose longitudinal distance from the transverse maximum convex hull point satisfies the second condition; When the target frame is located on the left side of the vehicle and turns left, the right boundary line of the target frame is corrected by the right convex hull points in the right convex hull point set of the target frame; When the target frame is located at the right side of the vehicle and turns right, the left boundary line of the target frame is corrected by the left convex hull points in the left convex hull point set of the target frame.

2. The target post-processing algorithm based on the point cloud in the frame according to claim 1 is characterized in that: By analyzing and processing the point cloud data, the target drivable area point located in the target frame is determined, including: Determine the drivable area points in the point cloud data by analyzing and processing the point cloud data; According to the location of the target frame and the location of the drivable area point, the target drivable area point located in the target frame is determined.

3. The target post-processing algorithm based on the point cloud in the frame according to claim 1 is characterized in that: Before determining the target drivable area point located in the target frame by analyzing and processing the point cloud data, the method further includes: Determining a virtual target frame among the multiple target frames according to the attribute information of the target frame; Filtering virtual target frames among the multiple target frames to obtain remaining target frames; The step of analyzing and processing the point cloud data to determine the target drivable area point located in the target frame comprises: By analyzing and processing the point cloud data, the target drivable area points located in the target frame remaining after filtering are determined.

4. The target post-processing algorithm based on the point cloud in the frame according to claim 1 or 3, characterized in that: Before determining the convex hull of the target frame and the convex hull point set corresponding to the convex hull according to the target drivable area point in the target frame, the method further includes: According to the attribute information of the target frame, determine the target frame that needs to be oriented corrected from all the target frames; The method of determining the convex hull of the target box and the set of convex hull points corresponding to the convex hull according to the target drivable area points in the target box includes: determining the convex hull of the target box that needs to be corrected in direction and the set of convex hull points corresponding to the convex hull according to the target drivable area points in the target box that needs to be corrected in direction.

5. The target post-processing algorithm based on the point cloud within the frame according to claim 1 is characterized in that: When the target frame is located on the left side of the vehicle and turns left, the right boundary line of the target frame is corrected using the right convex hull points in the right convex hull point set of the target frame, including: When the target frame is located on the left side of the vehicle and turns left, traverse the right convex hull points in the right convex hull point set of the target frame, and connect the currently traversed right convex hull point with the lateral minimum convex hull point to form a right predicted boundary line; When the angle deviation between the obtained right predicted boundary line and the right boundary line of the target frame satisfies the set condition, the traversal is terminated and the right boundary line of the target frame is corrected with the right predicted boundary line whose deviation satisfies the set condition; When the target frame is located at the right side of the vehicle and turns right, the left boundary line of the target frame is corrected using the left convex hull point in the left convex hull point set of the target frame, including: When the target frame is located on the right side of the vehicle and turns right, traverse the left convex hull points in the left convex hull point set of the target frame, and connect the currently traversed left convex hull point with the horizontal maximum convex hull point to form a left predicted boundary line; When the angle deviation between the obtained left predicted boundary line and the left boundary line of the target frame satisfies the set condition, the traversal is terminated and the left boundary line of the target frame is corrected with the left predicted boundary line whose deviation satisfies the set condition.

6. The target post-processing algorithm based on the point cloud within the frame according to claim 1 is characterized in that: The method further comprises: Clustering the point cloud data using a clustering algorithm to obtain each cluster target frame; Determine the proportion of point clouds of different semantic categories in the clustering target frame, determine the size of the clustering target frame, determine the reflectivity information of the point cloud in the clustering target frame, and determine the first point cloud density of the point cloud within a set range of the centroid of the clustering target frame; The cluster target frame is evenly divided into multiple segments in the longitudinal direction, and the second point cloud density around the center point of the first segment in the moving direction of the cluster target frame is calculated; Determine a virtual clustering target frame in each clustering target frame based on the point cloud proportion, the size, the reflectivity information, the first point cloud density, and the second point cloud density; The virtual clustering target frames in the clustering target frames are filtered to obtain a clustering result of the clustering algorithm.

7. The target post-processing algorithm based on the point cloud within the frame according to claim 6 is characterized in that: Determine the reflectivity information of the point cloud within the cluster target box, including: According to the attribute information of the point cloud in the cluster target frame, the number of point clouds with low reflectivity in the cluster target frame is determined, as well as the maximum reflectivity and the total reflectivity value of the point cloud in the cluster target frame; Determine the proportion of low-reflectivity point clouds in the cluster target frame according to the total number of point clouds in the cluster target frame and the number of low-reflectivity point clouds; According to the total number of point clouds and the total reflectivity value in the cluster target frame, the average reflectivity of the point clouds in the cluster target frame is determined.

8. The target post-processing algorithm based on the point cloud within the frame according to claim 7 is characterized in that: Determining virtual cluster target frames in each cluster target frame based on the point cloud proportion, the size, the reflectivity information, the first point cloud density, and the second point cloud density includes: When the proportion of the low reflectivity point cloud corresponding to the cluster target frame is a set value, the cluster target frame is determined to be a virtual cluster target frame; When the proportion of low reflectivity point clouds corresponding to the cluster target frame exceeds a first reflectivity threshold and the corresponding average reflectivity does not exceed a second reflectivity threshold, determining that the cluster target frame is a virtual cluster target frame; When the proportion of low reflectivity point clouds corresponding to the cluster target frame exceeds the third reflectivity threshold, the corresponding average reflectivity does not exceed the fourth reflectivity threshold, and the corresponding maximum reflectivity does not exceed the fifth reflectivity threshold, the cluster target frame is determined to be a virtual cluster target frame; When the proportion of point clouds of the noise semantic category in the cluster target box exceeds the category threshold, determining a size relationship between the size and a first size threshold, determining a density relationship between the first point cloud density and a first density threshold, and determining a quantity relationship between the total number of point clouds and a total number threshold; Determining whether the clustering target frame is a virtual clustering target frame according to the size relationship, the density relationship and the quantity relationship; When the proportion of point clouds of the noise semantic category in the clustering target frame does not exceed the category threshold, determine whether the clustering target frame is a virtual clustering target frame based on the proportion of point clouds of the background semantic category in the clustering target frame, the size, and the total number of point clouds; and determine whether the clustering target frame is a virtual clustering target frame based on the proportion of point clouds of the general obstacle semantic category in the clustering target frame, the position of the clustering target frame, and the size.

9. The target post-processing algorithm based on the point cloud within the frame according to claim 6, characterized in that: The method further comprises: Determine whether there is a small-sized single cluster target box within a preset range around the vehicle category cluster target box; In the case that there is a small-sized single cluster target frame within a preset range around the vehicle category cluster target frame, the vehicle category cluster target frame is determined to be a virtual cluster target frame.

10. A target post-processing system based on point cloud within a frame, characterized in that: The system comprises: The target recognition module is used to perform target recognition on the point cloud data through the target recognition model to obtain multiple target frames; An analysis and processing module, used to determine a target drivable area point located in a target frame by analyzing and processing the point cloud data; A convex hull point set determination module is used to determine the convex hull of the target box and the convex hull point set corresponding to the convex hull according to the target drivable area points in the target box, wherein the convex hull points in the convex hull point set are all the target drivable area points in the target box; An extreme convex hull point determination module is used to determine the maximum longitudinal convex hull point, the maximum transverse convex hull point and the minimum transverse convex hull point in the convex hull point set of the target frame; A right convex hull point set determination module is used to construct a right convex hull point set of the target frame when the target frame is located on the left side of the vehicle, using the convex hull points in the target frame that are sequentially located between the maximum convex hull point in the longitudinal direction and the minimum convex hull point in the transverse direction and whose longitudinal distance from the minimum convex hull point in the transverse direction meets the first condition; A left convex hull point set determination module is used to construct a left convex hull point set of the target frame when the target frame is located on the right side of the vehicle, using the left convex hull points of the target frame whose convex hull points are sequentially located in the longitudinal maximum convex hull point and the transverse maximum convex hull point and whose longitudinal distance from the transverse maximum convex hull point meets the second condition; A first correction module is used to correct the right boundary line of the target frame with the right convex hull points in the right convex hull point set of the target frame when the target frame is located on the left side of the vehicle and turns left; The second correction module is used to correct the left boundary line of the target frame with the left convex hull points in the left convex hull point set of the target frame when the target frame is located at the right side of the vehicle and turns right.

11. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and running on the processor, wherein when the computer program is executed by the processor, the steps in the target post-processing method based on the in-frame point cloud as described in claims 1 to 9 are implemented.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in the target post-processing method based on the in-frame point cloud as described in claims 1 to 9 are implemented.

Images