Point cloud screening method and device, storage medium and electronic device

Abstract

The application discloses a kind of screening method and device of point cloud, storage medium and electronic device, the method comprises: in the case where the road boundary of the road to be traveled is detected for target vehicle, according to target curvature information, the region segmentation of the current point cloud signal collection area of target vehicle is carried out, and a plurality of grid regions are obtained;Determine the position information of reference point cloud in a plurality of grid regions in point cloud signal collection area, wherein the reference point cloud is used to represent the road environment of the road to be traveled;According to the position information, the first point cloud is screened out from the point cloud falling in each grid region, wherein the first point cloud is used to represent the driving boundary of target vehicle in the corresponding sector region;According to the position relationship between the first point cloud, the target point cloud is screened out from the first point cloud corresponding to a plurality of grid regions, wherein the target point cloud is used to represent the road boundary of the road to be traveled, using the above technical solution, solve the problem such as the low accuracy of road boundary detection in the related art.

Description

Technical Field

[0001] This application relates to the field of radar, and more specifically, to a method and apparatus for screening point clouds, a storage medium, and electronic devices. Background Technology

[0002] Road edge detection is a crucial component of environmental perception in intelligent driving systems. This technology assists autonomous vehicles in achieving functions such as localization, path prediction, and path planning. Sensors used for road edge detection mainly include cameras, LiDAR, and millimeter-wave radar. Therefore, road edge detection methods are primarily divided into image-based and signal feature-based methods. Signal feature-based methods are typically applied to LiDAR or millimeter-wave radar, mainly utilizing point clouds generated from radar signal processing for road edge detection. A common approach is to cluster single-frame point clouds and then perform curve fitting based on the clustered point clouds to detect road edges. This method is effective for relatively simple road scenarios, such as those with minimal road interference and where the area to be identified is close to the vehicle. However, in complex road scenarios, the detected road edges differ significantly from the actual road edges, resulting in lower detection accuracy.

[0003] There is still no effective solution to the problem of low accuracy in road boundary detection in related technologies. Summary of the Invention

[0004] This application provides a point cloud screening method and apparatus, storage medium and electronic device to at least solve the problem of low accuracy of road boundary detection in related technologies.

[0005] According to one embodiment of the present application, a point cloud filtering method is provided, including: when detecting the road boundary of the road to be driven for a target vehicle, dividing the current point cloud signal acquisition area of ​​the target vehicle into multiple grid areas according to target curvature information, wherein the target curvature information is used to characterize the road curvature of the road to be driven, and the difference between the number of point clouds acquired in any two grid areas is less than or equal to a target threshold.

[0006] The position information of a reference point cloud in the point cloud signal acquisition area is determined in the multiple grid areas, wherein the reference point cloud is used to characterize the road environment of the road to be driven;

[0007] Based on the location information, a first point cloud is selected from the point clouds falling within each of the grid areas, wherein the first point cloud is used to characterize the driving boundary of the target vehicle within the corresponding sector area.

[0008] Target point clouds are selected from the first point clouds corresponding to multiple grid regions based on the positional relationship between the first point clouds, wherein the target point clouds are used to represent the road boundary of the road to be driven.

[0009] Optionally, the step of segmenting the current point cloud signal acquisition area of ​​the target vehicle based on the target curvature information to obtain multiple grid regions includes:

[0010] The target angle ratio corresponding to the target curvature information is determined from the curvature information and angle ratio that have a corresponding relationship, wherein the target angle ratio is used to characterize the regional angle of the segmented grid region;

[0011] The point cloud signal acquisition area is divided into regions according to the target angle ratio to obtain the multiple grid regions, wherein the grid region is a fan-shaped grid region centered on the current position of the target vehicle.

[0012] Optionally, the step of filtering the target point cloud from the first point clouds corresponding to multiple grid regions based on the positional relationship between the first point clouds includes:

[0013] The lateral position difference of the first point cloud corresponding to two adjacent grid regions is calculated sequentially.

[0014] The first point cloud is clustered based on the lateral position difference to obtain multiple point cloud clusters;

[0015] Target point cloud clusters are selected from the plurality of point cloud clusters based on cluster features, wherein the cluster features are used to characterize the spatial distribution of point clouds belonging to the corresponding point cloud clusters.

[0016] Point clouds belonging to the target point cloud cluster are identified as the target point cloud.

[0017] Optionally, the step of selecting the target point cloud cluster from the plurality of point cloud clusters based on cluster characteristics includes:

[0018] Construct the connection trajectories between multiple point clouds included in the point cloud cluster;

[0019] If the trajectory length of the connection trajectory is greater than or equal to the target length, a quadratic curve fitting is performed on the multiple point clouds to obtain candidate point cloud clusters;

[0020] If the proportion of the second point cloud in the candidate point cloud cluster is greater than or equal to the target proportion, the candidate point cloud cluster is determined as the target point cloud cluster, wherein the second point cloud is the point cloud in the candidate point cloud cluster whose residual value is less than or equal to the target residual value.

[0021] Optionally, the step of clustering the first point cloud based on the lateral position difference to obtain multiple point cloud clusters includes:

[0022] If the lateral position difference is less than or equal to the target difference, the two first point clouds are classified into the same point cloud cluster.

[0023] If the lateral position difference is greater than the target difference, the position gradient relationship between the two first point clouds is obtained; if the position gradient relationship satisfies the target gradient relationship, the two first point clouds are classified into the same point cloud cluster.

[0024] Optionally, the step of filtering the first point cloud from the point clouds falling within each of the grid areas based on the location information includes:

[0025] Select stationary point clouds from the point clouds falling within each of the grid regions;

[0026] The relative distance between the stationary point cloud and the target vehicle is calculated using the location information.

[0027] The point cloud whose relative distance is less than or equal to the target distance is defined as the first point cloud.

[0028] Optionally, after filtering the target point cloud from the first point clouds corresponding to the plurality of grid regions based on the positional relationship between the first point clouds, the method further includes:

[0029] Construct the road boundary trajectory of the road to be driven based on the positional relationships of the target point cloud;

[0030] If the length of the road boundary trajectory is greater than or equal to the target length, the third point cloud in the grid area is marked as a point cloud, wherein the third point cloud is a point cloud in the reference point cloud whose point cloud position is outside the boundary area of ​​the road to be driven.

[0031] If the length of the road boundary trajectory is less than the target length, the fourth point cloud in the grid area is marked with point cloud markers. The fourth point cloud is the point cloud in the reference area whose point cloud position is outside the area position of the road to be driven in the reference point cloud. The reference area is the area corresponding to the road boundary trajectory.

[0032] Optionally, before filtering out the first point cloud from the point clouds falling within each of the grid regions based on the location information, the method further includes:

[0033] Acquire a fifth point cloud and a sixth point cloud, wherein the fifth point cloud is used to characterize the road environment of the target vehicle's driving road at a reference time before the current time, and the sixth point cloud is the point cloud collected by the radar equipment deployed on the target vehicle at the current time.

[0034] Predict the current position of the fifth point cloud to obtain the seventh point cloud;

[0035] The sixth point cloud and the seventh point cloud are merged to obtain the reference point cloud.

[0036] According to another embodiment of the present application, a point cloud filtering device is also provided, including: a segmentation module, used to segment the current point cloud signal acquisition area of ​​the target vehicle according to target curvature information to obtain multiple grid regions when detecting the road boundary of the road to be driven for the target vehicle, wherein the target curvature information is used to characterize the road curvature of the road to be driven, and the difference between the number of point clouds acquired in any two grid regions is less than or equal to a target threshold.

[0037] The determining module is used to determine the position information of a reference point cloud in the point cloud signal acquisition area in the multiple grid areas, wherein the reference point cloud is used to characterize the road environment of the road to be driven;

[0038] The first filtering module is used to filter out a first point cloud from the point clouds falling within each of the grid areas based on the location information, wherein the first point cloud is used to characterize the driving boundary of the target vehicle within the corresponding fan-shaped area.

[0039] The second filtering module is used to filter out target point clouds from the first point clouds corresponding to multiple grid regions based on the positional relationship between the first point clouds, wherein the target point clouds are used to represent the road boundary of the road to be driven.

[0040] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, wherein a computer program is stored in the computer-readable storage medium, and the computer program is configured to execute the above-described point cloud filtering method at runtime.

[0041] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the point cloud filtering method through the computer program.

[0042] In this embodiment, when detecting the road boundary of the road to be driven by the target vehicle, the current point cloud signal acquisition area of ​​the target vehicle is segmented according to the target curvature information to obtain multiple grid regions. The target curvature information is used to characterize the road curvature of the road to be driven, and the difference in the number of point clouds acquired in any two grid regions is less than or equal to a target threshold. The position information of a reference point cloud in the point cloud signal acquisition area is determined within the multiple grid regions, where the reference point cloud is used to characterize the road environment of the road to be driven. Based on the position information, a first point cloud is selected from the point clouds falling within each grid region, where the first point cloud is used to characterize the target vehicle within the corresponding sector. The driving boundary within the area; based on the positional relationship between the first point clouds, a target point cloud is selected from the first point clouds corresponding to multiple grid regions. The target point cloud is used to represent the road boundary of the road to be driven. Specifically, when road boundary detection is required, the current point cloud signal acquisition area of ​​the target vehicle is segmented according to the road curvature, thereby ensuring that the number of point clouds acquired in each grid region is uniform, avoiding the impact of uneven point cloud distribution on road boundary recognition in complex road environments. Then, when identifying the road boundary, the first point cloud representing the driving boundary of the target vehicle is first identified based on the positional information of the reference point cloud in the grid, and then further filtered based on the positional relationship between the first point clouds to obtain the target point cloud used to represent the road boundary. This technical solution solves the problem of low accuracy in road boundary detection in related technologies, achieving a significant improvement in the accuracy of road boundary detection. Attached Figure Description

[0043] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0044] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a schematic diagram of the hardware environment for a point cloud filtering method according to an embodiment of this application;

[0046] Figure 2 This is a flowchart of a point cloud screening method according to an embodiment of this application;

[0047] Figure 3 This is an optional grid region segmentation diagram according to an embodiment of this application. Figure 1 ;

[0048] Figure 4 This is an optional road boundary detection flowchart according to an embodiment of this application;

[0049] Figure 5 This is an optional grid region segmentation diagram according to an embodiment of this application. Figure 2 ;

[0050] Figure 6 This is an optional point cloud clustering diagram according to an embodiment of this application;

[0051] Figure 7 This is an optional point cloud marker schematic diagram according to an embodiment of this application;

[0052] Figure 8 This is an optional point cloud preprocessing flowchart according to an embodiment of this application;

[0053] Figure 9 This is an optional road boundary recognition flowchart according to an embodiment of this application;

[0054] Figure 10 This is a structural block diagram of a point cloud screening device according to an embodiment of this application. Detailed Implementation

[0055] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort should fall within the scope of protection of the present application.

[0056] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0057] The methods and embodiments provided in this application can be executed on a computer terminal, device terminal, or similar computing device. Taking running on a computer terminal as an example, Figure 1 This is a schematic diagram of the hardware environment for a point cloud filtering method according to an embodiment of this application. For example... Figure 1 As shown, a computer terminal may include one or more ( Figure 1 Only one is shown in the diagram. A processor 102 (which may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.) and a memory 104 for storing data are also shown. In one exemplary embodiment, the computer terminal may further include a transmission device 106 for communication functions and an input / output device 108. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the computer terminal described above. For example, the computer terminal may also include components that are more complex than those described above. Figure 1 The more or fewer components shown, or having the same Figure 1 Equivalent functions or ratios shown Figure 1 The functions shown have more different configurations.

[0058] The memory 104 can be used to store computer programs, such as application software programs and modules, like the computer program corresponding to the message push sending method in this embodiment of the invention. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, thereby implementing the above-described method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to a computer terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0059] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider for the computer terminal. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module used for wireless communication with the Internet.

[0060] This embodiment provides a point cloud filtering method. Figure 2This is a flowchart of a point cloud filtering method according to an embodiment of this application, the process including the following steps:

[0061] Step S202: When detecting the road boundary of the road to be driven for the target vehicle, the current point cloud signal acquisition area of ​​the target vehicle is divided into multiple grid areas according to the target curvature information. The target curvature information is used to characterize the road curvature of the road to be driven. The difference between the number of point clouds acquired in any two grid areas is less than or equal to the target threshold.

[0062] Step S204: Determine the position information of the reference point cloud in the point cloud signal acquisition area in the multiple grid areas, wherein the reference point cloud is used to characterize the road environment of the road to be driven;

[0063] Step S206: Based on the location information, a first point cloud is selected from the point clouds falling within each of the grid areas, wherein the first point cloud is used to characterize the driving boundary of the target vehicle within the corresponding fan-shaped area.

[0064] Step S208: Select a target point cloud from the first point clouds corresponding to the multiple grid regions according to the positional relationship between the first point clouds, wherein the target point cloud is used to represent the road boundary of the road to be driven.

[0065] Through the above steps, when road boundary detection is required, the current point cloud signal acquisition area of ​​the target vehicle is segmented according to the road curvature. This ensures that the number of point clouds acquired in each grid area is uniform, avoiding the impact of uneven point cloud distribution on road boundary recognition in complex road environments. Furthermore, when identifying the road boundary, the first point cloud representing the target vehicle's driving boundary is first identified based on the position information of the reference point cloud in the grid. Then, it is further filtered based on the positional relationships between the first point clouds to obtain the target point cloud used to represent the road boundary. This technical solution solves the problem of low accuracy in road boundary detection in related technologies, achieving a significant improvement in the accuracy of road boundary detection.

[0066] In the technical solution provided in step S202 above, the target curvature information can be obtained by identifying the road. For example, by acquiring a road image of the road to be driven by the target vehicle and performing image recognition based on the road image, the road curvature of the road to be driven can be obtained.

[0067] Optionally, in this embodiment of the application, the target curvature information can also be predicted based on the driving information of the target vehicle. For example, the vehicle steering angle of the target vehicle is obtained, and the target curvature information of the target road is predicted based on the vehicle steering angle. The specific operation of predicting the target curvature information based on the vehicle steering angle can be, but is not limited to, obtaining the vehicle speed of the target vehicle and determining the target curvature information corresponding to the current driving speed and the current vehicle steering angle from the driving speed, vehicle steering angle and curvature information that have a corresponding relationship.

[0068] Optionally, in the embodiments of this application, the target curvature information may include, but is not limited to, the radius of curvature, arc length, etc., and this solution does not limit this.

[0069] Optionally, in this embodiment, region segmentation is used to divide the point cloud signal acquisition area into multiple grid regions, thereby ensuring that the number of point clouds falling in each grid region is relatively uniform. In this embodiment, the grid region can be divided into multiple uniform grid regions or multiple non-uniform grid regions according to the road curvature information. Figure 3 This is an optional grid region segmentation diagram according to an embodiment of this application. Figure 1 ,like Figure 3 As shown, when the road is a curved road, the distribution pattern collected from the current position of the target vehicle shows that the point cloud distribution is sparse when it is close to the vehicle and dense when it is far away from the vehicle. Therefore, the signal acquisition area can be divided into multiple grids with uneven area to ensure a uniform distribution of the number of point clouds in each grid area.

[0070] In the technical solution provided in step S204 above, the position information of the reference point cloud in multiple grid regions may be, but is not limited to, the coordinate position of the reference point cloud in the corresponding grid region, such as the coordinate position in the polar coordinate system and the coordinate position in the rectangular coordinate system. This solution does not limit this.

[0071] Optionally, in this embodiment, the reference point cloud information can be a point cloud collected at the current location of the target vehicle, or it can be a point cloud obtained by merging the point cloud collected at the current time with the point cloud collected in the past time period. For example, multiple frames of point cloud images of the target vehicle collected at reference times between the current time and the current time are obtained, the point cloud positions in the multiple frames of point cloud images are predicted to the current frame to obtain the predicted point cloud of the current frame, and the predicted point cloud of the current frame is merged with the point cloud actually collected at the current time to obtain the reference point cloud. This method can effectively avoid the impact of inaccurate point cloud recognition on the accuracy of road boundary detection.

[0072] In the technical solution provided in step S206 above, the first point cloud can be a point cloud selected based on the distance relationship between the point cloud in each grid area and the target vehicle. For example, the relative distance between each point cloud and the target vehicle can be calculated using location information, and the point cloud with the smallest relative distance to the target vehicle can be determined as the first point cloud.

[0073] Optionally, in this embodiment, the first point cloud can also be a point cloud selected based on the relative positional relationship between point clouds within a grid area. For example, curve fitting can be performed on the point clouds within the grid area based on the positional information of the point clouds, and the point clouds that meet the target curve fitting conditions can be determined as the first point cloud. Alternatively, position gradient calculation can be performed on the point clouds within the grid area based on the positional information of the point clouds (which may include, but is not limited to, vertical position gradient calculation and / or horizontal position gradient calculation), and the point clouds within the grid area can be clustered based on the position gradient results to obtain multiple point cloud clusters. Then, the point clouds in the point cloud clusters whose cluster features meet the road boundary features can be determined as the first point cloud. This solution does not limit this.

[0074] In the technical solution provided in step S208 above, the first point cloud representing the vehicle driving boundary can be identified based on the location information. However, in some complex road environments, there may be stationary objects near the road boundary that are identified as the road boundary. Therefore, the positional relationship between the first point clouds is used to filter out the target point cloud from the first point clouds corresponding to multiple grid areas, thereby ensuring the accuracy of road boundary recognition.

[0075] Optionally, in the embodiments of this application, the positional relationship between the first point clouds may include, but is not limited to, the lateral distance between the first point clouds in two adjacent grid areas, the positional gradient between the point clouds, etc., and this solution does not limit this.

[0076] Figure 4 This is an optional road boundary detection flowchart according to an embodiment of this application, such as... Figure 4 As shown, it includes at least the following steps:

[0077] Step S401: Filter the current frame to establish the stationary point cloud for Freespace, obtain the historical multi-frame stationary point cloud set, and merge the current frame stationary point cloud with the multi-frame stationary point cloud set to obtain the reference point cloud.

[0078] Step S402: Establish Freespace boundaries in multiple raster regions using reference point clouds.

[0079] Step S403: Determine the position of the guardrail based on the Freespace boundary; mark the points outside the Freespace.

[0080] As an optional embodiment, the step of segmenting the current point cloud signal acquisition area of ​​the target vehicle based on the target curvature information to obtain multiple grid regions includes:

[0081] The target angle ratio corresponding to the target curvature information is determined from the curvature information and angle ratio that have a corresponding relationship, wherein the target angle ratio is used to characterize the regional angle of the segmented grid region;

[0082] The point cloud signal acquisition area is divided into regions according to the target angle ratio to obtain the multiple grid regions, wherein the grid region is a fan-shaped grid region centered on the current position of the target vehicle.

[0083] Optionally, in this embodiment, the area angles of different grid regions within multiple grid areas can be set to different angles. The larger the area angle, the larger the area covered by the current grid. Therefore, in road environments with different curvature information, the area of ​​each grid region can be adjusted by setting the grid angle, thereby controlling the number of point clouds falling in each region. For example, if the road to be driven is a straight road with a curvature of 0, then for roads with this curvature information, the point cloud signal acquisition area is divided into N sector grids. Since the road edge is basically vertical, the point cloud density on both sides of the road becomes denser as the distance from the target vehicle increases. Therefore, to ensure that the number of point clouds in each grid region is relatively uniform and to avoid a large number of point clouds being divided into the same grid region due to the higher point cloud density at distant points, thus filtering out the point clouds corresponding to the road boundary, a larger area angle is required. Figure 5 This is an optional grid region segmentation diagram according to an embodiment of this application. Figure 2 ,like Figure 5 As shown, when the curvature of the road is 0, when setting the grid angle, consider setting the grid to a pattern that is sparse on both sides and dense in the middle. The angle of the middle k grids is θ1, and the angle of the two sides (Nk) / 2 grids is θ2.

[0084] As an optional embodiment, the step of filtering the target point cloud from the first point clouds corresponding to the plurality of grid regions based on the positional relationship between the first point clouds includes:

[0085] The lateral position difference of the first point cloud corresponding to two adjacent grid regions is calculated sequentially.

[0086] The first point cloud is clustered based on the lateral position difference to obtain multiple point cloud clusters;

[0087] Target point cloud clusters are selected from the plurality of point cloud clusters based on cluster features, wherein the cluster features are used to characterize the spatial distribution of point clouds belonging to the corresponding point cloud clusters.

[0088] Point clouds belonging to the target point cloud cluster are identified as the target point cloud.

[0089] Optionally, in this embodiment, the method of clustering the first point cloud based on the lateral position difference may include, but is not limited to, grouping adjacent point clouds into the same point cloud cluster when the lateral position difference is less than or equal to the target position difference, and grouping adjacent point clouds into different point cloud clusters when the lateral position difference is greater than the target position difference. Figure 6 This is an optional point cloud clustering diagram according to an embodiment of this application, such as... Figure 6 As shown in the figure, the point cloud signal acquisition area of ​​the target vehicle includes 9 first point clouds. These first point clouds represent the current driving boundary of the target vehicle. However, in actual applications, there may be stationary obstacles at the road edge, such as the reference vehicle in the figure. The point cloud corresponding to this vehicle is identified as the first point cloud. Then, it is necessary to filter out the target point cloud corresponding to the road boundary from the first point clouds. Based on the lateral position difference between two adjacent first point clouds, the 9 point clouds in the figure can be clustered into three clusters: the cluster consisting of the five point clouds close to the target vehicle, the cluster consisting of the two point clouds in front and behind the reference vehicle, and the cluster consisting of the two point clouds at the farthest position of the target vehicle. Then, the target cluster is filtered out from the three point cloud clusters based on the cluster characteristics.

[0090] Optionally, in the embodiments of this application, the cluster features may include, but are not limited to, the number of point clouds in the cluster, the position gradient of the point clouds in the cluster, the trajectory length of the connection trajectory of the point clouds in the cluster, etc., and this solution does not limit them.

[0091] As an optional embodiment, the step of filtering the target point cloud cluster from the plurality of point cloud clusters based on cluster characteristics includes:

[0092] Construct the connection trajectories between multiple point clouds included in the point cloud cluster;

[0093] If the trajectory length of the connection trajectory is greater than or equal to the target length, a quadratic curve fitting is performed on the multiple point clouds to obtain candidate point cloud clusters;

[0094] If the proportion of the second point cloud in the candidate point cloud cluster is greater than or equal to the target proportion, the candidate point cloud cluster is determined as the target point cloud cluster, wherein the second point cloud is the point cloud in the candidate point cloud cluster whose residual value is less than or equal to the target residual value.

[0095] Optionally, in this embodiment of the application, the target length may be set according to the length of a reference object at the edge of the road, such as setting the length of the fence at the edge of the road as the target length.

[0096] Optionally, in this embodiment of the application, for the point clouds in the candidate point cloud clusters obtained by quadratic curve fitting, the residual of each point cloud is calculated. If the number of points with residual values ​​less than the target residual value is above the target proportion of the total number of points, it is identified as a target point cloud cluster composed of the target point cloud.

[0097] As an optional embodiment, the step of clustering the first point cloud based on the lateral position difference to obtain multiple point cloud clusters includes:

[0098] If the lateral position difference is less than or equal to the target difference, the two first point clouds are classified into the same point cloud cluster.

[0099] If the lateral position difference is greater than the target difference, the position gradient relationship between the two first point clouds is obtained; if the position gradient relationship satisfies the target gradient relationship, the two first point clouds are classified into the same point cloud cluster.

[0100] Optionally, in the embodiments of this application, the gradient relationship can be, but is not limited to, obtained by calculating the position gradient of the reference point cloud. For example, in step 1, the stationary point clouds in the reference point cloud are first filtered to obtain a set of stationary point clouds, Pall_sortY. In step 2, the difference between the lateral position of the current point and the lateral position of the next point in the set of stationary point clouds, Pall_sortY, is calculated sequentially to obtain a set of lateral position gradients, Grad_Y. In step 3, the set of lateral position gradients of stationary point clouds, Grad_Y, is searched sequentially. If the lateral position gradient of the nth stationary point cloud is less than the threshold δY, the nth stationary point cloud in the Pall set is taken as the starting point cloud of the road edge detection point cloud set Py1. The gradient set is traversed sequentially, and the gradient value of each point is accumulated and recorded as Gard_all. Then, the lateral position gradient set Pall is searched again, and it is determined that the current gradient is less than δY and the gradient sum is less than δY_all. Step 4: Repeat Step 3 until the lateral position gradient set Grad_Y has been searched, resulting in m road edge detection point cloud sets {Py1, Py2, Py3…Pym}, where Pym represents the m-th road edge detection point cloud set. Step 5: Sequentially search the m road edge detection point cloud sets {Py1, Py2, Py3…Pym}, and count the number of stationary point clouds Kpoints in the same road edge detection point cloud set Pm. If the number of stationary point clouds Kpoints is greater than the minimum road edge detection point cloud count threshold σ, then the point cloud set Pm is removed from the list. mThe stationary point clouds are arranged in ascending order of their ordinate positions to obtain a sorted set of stationary point clouds, Grad_Ym_X. The difference between the lateral position of the current point and the ordinate position of the next point in the Grad_Ym_X set is calculated as the longitudinal gradient, resulting in the set of longitudinal gradients, Grad_Xm. Then, step 6 is executed. If the number of stationary point clouds, Kpoints, is less than the threshold σ, step 5 is returned. This process continues until all M road edge detection point cloud sets {Py1, Py2, Py3…Pym} have been searched. In step 6, the longitudinal gradient set Grad_Xm of the stationary point clouds is searched sequentially. If the longitudinal gradient of the i-th stationary point cloud is less than the threshold δX, the i-th stationary point cloud in Grad_Ym_X is used as the starting point cloud of the road edge detection point cloud set Px1. The gradient set is then iterated through, and the gradient value of each point is accumulated and denoted as Gard_all. The search continues in the longitudinal gradient set Grad_Xm, determining that the current gradient is less than δX and the gradient sum is less than δX_all. If the number of stationary point clouds in the stationary point cloud set Px1 is greater than the minimum threshold σ for the number of road edge point clouds, then the point clouds in the stationary point cloud set Px1 are marked as being used for road edge detection, and step 5 is repeated until the longitudinal position gradient set Grad_Xm is searched. Step 7: After searching through M road edge detection point cloud sets {Py1, Py2, Py3…Pym}, the stationary point clouds marked as being used for road edge detection are added to the stationary point cloud set {Px1, Px2, Px3…Pxm} for use in detecting the current road boundary, where Pxm represents the stationary point cloud used for road edge detection, and m represents the number of stationary point clouds used for road edge detection. Furthermore, if two first point clouds are located in the same set Pxm, then the position gradient relationship between the two first point clouds is determined to satisfy the target gradient relationship.

[0101] As an optional embodiment, the step of filtering the first point cloud from the point clouds falling within each of the grid regions based on the location information includes:

[0102] Select stationary point clouds from the point clouds falling within each of the grid regions;

[0103] The relative distance between the stationary point cloud and the target vehicle is calculated using the location information.

[0104] The point cloud whose relative distance is less than or equal to the target distance is defined as the first point cloud.

[0105] Optionally, in the embodiments of this application, the first point cloud can be a single point cloud or a set of point clouds composed of multiple point clouds. For example, when multiple point clouds are at the same relative distance to the target vehicle and the relative distance is less than or equal to the target distance, the multiple point clouds are simultaneously determined as the first point cloud.

[0106] As an optional embodiment, after filtering the target point cloud from the first point clouds corresponding to the plurality of grid regions based on the positional relationship between the first point clouds, the method further includes:

[0107] Construct the road boundary trajectory of the road to be driven based on the positional relationships of the target point cloud;

[0108] If the length of the road boundary trajectory is greater than or equal to the target length, the third point cloud in the grid area is marked as a point cloud, wherein the third point cloud is a point cloud in the reference point cloud whose point cloud position is outside the boundary area of ​​the road to be driven.

[0109] If the length of the road boundary trajectory is less than the target length, the fourth point cloud in the grid area is marked with point cloud markers. The fourth point cloud is the point cloud in the reference area whose point cloud position is outside the boundary area of ​​the road to be driven. The reference area is the area corresponding to the road boundary trajectory.

[0110] Figure 7 This is an optional point cloud marker schematic diagram according to an embodiment of this application, such as... Figure 7 As shown, if the length of the left road boundary is less than the target length, it is considered that the current road is entering an intersection. Only the point cloud with a longitudinal distance outside the road boundary that is less than the length of the road boundary is marked (the point cloud in the shaded area in the figure). The point cloud outside the shaded area is not marked. When the length of the road boundary is greater than or equal to the target length (i.e., the right road boundary in the figure), it is considered to be a straight road segment. Points outside the free space can be marked. The tracking can process the point cloud outside the right road boundary based on the marking information, which can effectively remove ghosts outside the road boundary.

[0111] As an optional embodiment, before filtering out the first point cloud from the point clouds falling within each of the grid regions based on the location information, the method further includes:

[0112] Acquire a fifth point cloud and a sixth point cloud, wherein the fifth point cloud is used to characterize the road environment of the target vehicle's driving road at a reference time before the current time, and the sixth point cloud is the point cloud collected by the radar equipment deployed on the target vehicle at the current time.

[0113] Predict the current position of the fifth point cloud to obtain the seventh point cloud;

[0114] The sixth point cloud and the seventh point cloud are merged to obtain the reference point cloud.

[0115] Optionally, in this embodiment of the application, when merging the sixth point cloud and the seventh point cloud, if the sixth point cloud and the seventh point cloud exist simultaneously in a certain area, the relative distance between the sixth point cloud and the seventh point cloud can be calculated. If the relative distance is less than a preset relative distance, the sixth point cloud and the seventh point cloud are retained at the same time. If the relative distance is greater than or equal to the preset relative distance, the seventh point cloud is deleted.

[0116] Through the above steps, a multi-frame point cloud merging method is used to merge the point cloud at the reference time with the point cloud acquired at the current time, thereby avoiding the impact of inaccurate single-frame point clouds on road boundary recognition. In this embodiment, point cloud preprocessing can be performed on multiple frames of point clouds to obtain a reference point cloud. The main purpose of point cloud preprocessing is to maintain a stable point cloud list formed by multiple frames of stationary point clouds, extract a set of stationary point clouds for road edge detection, and remove interfering stationary point clouds. Figure 8 This is an optional point cloud preprocessing flowchart according to an embodiment of this application, such as... Figure 8 As shown, it includes at least the following steps:

[0117] S801: Input the point cloud set acquired by the vehicle-mounted millimeter-wave radar at the current moment, and combine it with the radar vehicle's motion information to determine and record stationary point cloud data, obtaining a stationary point cloud set Pcurrent. Calculate the x and y coordinates of each point cloud in the stationary point cloud set Pcurrent in the radar vehicle's coordinate system. Predict the point track list Ptrack (the point cloud distribution at reference times before the current moment) to the current frame, and add the current frame's point track set Pcurrent to the point track list set Ptrack, obtaining a multi-frame point track list Pall for the current frame and the previous N frames.

[0118] S802, sort the point clouds in the static point cloud set Pall according to the x-coordinate from smallest to largest to obtain the sorted static point cloud set Pall_sortY.

[0119] S803, calculate the gradient by sequentially calculating the difference between the lateral position of the current point and the lateral position of the next point in the static point cloud set Pall_sortY, and obtain the lateral position gradient set Grad_Y.

[0120] S804: Sequentially search the lateral position gradient set Grad_Y of the stationary point cloud. If the lateral position gradient of the nth stationary point cloud is less than the threshold δY, then take the nth stationary point cloud in the Pall set as the starting point cloud of the road edge detection point cloud set Py1. Continue to traverse the gradient set and accumulate the gradient value of each point, denoted as Gard_all. Then continue to search the lateral position gradient set Pall, and determine whether the current gradient is less than δY and the gradient sum is less than δY_all. Repeat this process until the lateral position gradient set Grad_Y is completely searched, resulting in m road edge detection point cloud sets {Py1, Py2, Py3…Pym}, where Pym represents the mth road edge detection point cloud set.

[0121] S805, sequentially search m road edge detection point cloud sets {Py1, Py2, Py3…Pym}, and count the number of stationary point clouds Kpoints in the same road edge detection point cloud set Pm. If the number of stationary point clouds Kpoints is greater than the minimum road edge detection point cloud count threshold σ, then the point cloud set P... m The stationary point clouds are arranged in ascending order of their ordinate positions to obtain a sorted set of stationary point clouds, Grad_Ym_X. The difference between the lateral position of the current point and the ordinate position of the next point in the set of stationary point clouds, Grad_Ym_X, is calculated as the longitudinal gradient to obtain the set of longitudinal position gradients of the point clouds, Grad_Xm. Then, step 806 is executed. If the number of stationary point clouds, Kpoints, is less than the threshold σ, the process returns to step S805 until all M road edge detection point cloud sets {Py1, Py2, Py3…Pym} have been searched.

[0122] S806: Sequentially search the set of vertical position gradients of the stationary point cloud, Grad_Xm. If the vertical position gradient of the i-th stationary point cloud is less than the threshold δX, then take the i-th stationary point cloud in Grad_Xm_X as the starting point cloud of the road edge detection point cloud set Px1. Continue to traverse the gradient set and accumulate the gradient value of each point, denoted as Gard_all. Then continue to search the set of vertical position gradients, Grad_Xm, and determine whether the current gradient is less than δX and the gradient sum is less than δX_all. If the number of stationary point clouds in the set of stationary point clouds Px1 is greater than the minimum threshold σ for the number of road edge point clouds, then mark the point clouds in the set of stationary point clouds Px1 as points used for road edge detection, and repeat step S805 until the set of vertical position gradients, Grad_Xm, is completely searched.

[0123] S807, after searching through the M road edge detection point cloud sets {Py1,Py2,Py3…Pym}, the stationary point clouds marked as road edge detection point clouds are added to the stationary point cloud set {Px1,Px2,Px3…Pxm} for use in detecting the current road boundary, where Pxm represents the stationary point cloud used for road edge detection, and m represents the number of stationary point clouds used for road edge detection.

[0124] In this application, the point cloud signal acquisition area is segmented based on road curvature information to obtain multiple fan-shaped grid regions. The actual edge positions of each fan-shaped grid are established using the position information of the stationary point cloud set {Px1, Px2, Px3…Pxm}, thus obtaining a road edge position model. in and These represent the radial distance and angle of the stationary point cloud, respectively. Road boundaries are then identified based on the point cloud location information. Figure 9 This is an optional road boundary recognition flowchart according to an embodiment of this application, such as... Figure 9 As shown, it includes at least the following steps:

[0125] S901: Using the vehicle's motion information, the road edge position set FS_last from the previous moment is predicted to the current moment, resulting in the predicted road edge position set FS_current. The stationary point clouds in the current stationary point cloud set Pall are matched to corresponding sector grids based on angle information. The occupied sector grids in the road edge position set FS_current are searched sequentially. Then, the stationary point cloud closest to the edge position in each sector grid is identified as the corresponding road edge associated point cloud, and the information of the associated point clouds for each occupied sector grid is recorded.

[0126] S902, sequentially search the occupied sector grids in the road edge location set FS_current, and determine the associated point cloud status of each sector grid. If a sector grid is not associated with a stationary point cloud, use the predicted edge position as the estimated position of the occupied sector grid; if a sector grid is associated with a stationary point cloud, use the associated stationary point cloud to filter and update the edge position of the associated sector grid, and recalculate the sequence number of the sector grid where the corresponding road edge position is located based on the filtered road edge angle, to obtain the filtered road edge location set FS_update.

[0127] S903: Iterate through the occupied sector grids in the road edge location set FS_update, and determine the status of the sector grid's association with the point cloud. If an occupied sector grid is not associated with a point cloud for four consecutive frames, then set the corresponding sector grid edge position to the initial state.

[0128] S904: Iterate through the unoccupied sector grids in the road edge location set FS_update, find the stationary point cloud in each unoccupied sector grid that belongs to the point cloud set {Px1,Px2,Px3…Pxm} and has the shortest radial distance from the radar, and use the stationary point cloud to initialize the edge position of the unoccupied sub-sector grid.

[0129] S905, Determine the position of the edge of adjacent occupied sector grid cells. and If the spacing is less than the minimum spacing threshold, then the unoccupied sector grids between the two adjacent occupied sector grids are smoothed and filtered to obtain the edge position of the filtered sector grid FS_filter_i, and the grid is marked as an occupied grid. After filtering, the final set of road edge positions FS_filter for the current frame is obtained.

[0130] S906: Select all valid Freespace sector grid boundaries, calculate the lateral position difference between adjacent boundaries, and then cluster them based on the lateral position difference. Valid Freespace boundaries that meet the lateral position difference threshold are grouped into the same cluster. If two adjacent occupied grids do not meet the lateral threshold condition, determine whether the two grids are in the same small set of the road edge detection set {Py1,Py2,Py3…Pym}. For example, if two grids belong to Py3, they also need to be grouped into the same cluster. Continue to determine the number of valid boundaries in each cluster. If the number meets the minimum threshold for the number of guardrail boundaries, calculate the distance between the nearest and farthest points in this cluster. If this distance meets the minimum fence determination length, then the valid Freespace boundaries in the current cluster are considered as guardrail boundaries. Continue to traverse until the last sector grid, and the selection of all cluster fences is completed. If a car obstructs the fence, causing it to be interrupted, the system checks whether the distance between adjacent fence boundaries is less than the maximum distance threshold. If it is, filtering is performed, the middle fence is marked as occupied, and the fence marker is set to 1. A quadratic curve is fitted to the (x,y) points of the anglebin where all fence markers are set to 1, and the residual of each point is calculated. If the number of points with residuals less than a threshold is above a certain threshold of the total number of points, the fence is considered to be completely checked. Otherwise, the fence marker is set to 0.

[0131] S907, based on the sector grid model The left and right grids are traversed from smallest to largest (1 to N / 2) and from largest to smallest (N to N / 2) respectively. The lengths of the guardrails on both sides are counted. If the length is less than a threshold, the marking range of the guardrail is set to the guardrail length. If it is greater than the threshold, the marking range is not restricted. The radar FOV boundary position is calculated using the current Freespace fence boundary information. Then, all points in the current frame are traversed in turn, the Freespace boundary number corresponding to each point is calculated, and it is determined whether there are guardrail markings on adjacent Freespace boundaries. If there are guardrails, it is determined whether the point is in the clockwise direction of the line connecting the left and right adjacent boundaries. If the fence length is less than a certain threshold, it is considered that the current point has entered an intersection, and only points whose longitudinal distance outside the fence is less than the fence length are marked. When the fence length is greater than the threshold, it is considered that it is a straight road segment, and points outside the Freespace can be marked. Tracking can process points outside the fence based on the marking information, which can effectively remove ghosts outside the fence.

[0132] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software and necessary general-purpose hardware platforms, and of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, or network device, etc.) to execute the methods of the various embodiments of this application.

[0133] Figure 10 This is a structural block diagram of a point cloud screening device according to an embodiment of this application; as shown... Figure 10 As shown, it includes: a segmentation module, used to segment the current point cloud signal acquisition area of ​​the target vehicle according to the target curvature information when detecting the road boundary of the road to be driven for the target vehicle, to obtain multiple grid regions, wherein the target curvature information is used to characterize the road curvature of the road to be driven, and the difference between the number of point clouds acquired in any two grid regions is less than or equal to a target threshold.

[0134] The determining module is used to determine the position information of a reference point cloud in the point cloud signal acquisition area in the multiple grid areas, wherein the reference point cloud is used to characterize the road environment of the road to be driven;

[0135] The first filtering module is used to filter out a first point cloud from the point clouds falling within each of the grid areas based on the location information, wherein the first point cloud is used to characterize the driving boundary of the target vehicle within the corresponding fan-shaped area.

[0136] The second filtering module is used to filter out target point clouds from the first point clouds corresponding to multiple grid regions based on the positional relationship between the first point clouds, wherein the target point clouds are used to represent the road boundary of the road to be driven.

[0137] The above embodiments, when requiring road boundary detection, segment the current point cloud signal acquisition area of ​​the target vehicle based on road curvature. This ensures that the number of point clouds acquired within each grid area is uniform, avoiding the impact of uneven point cloud distribution on road boundary recognition in complex road environments. Furthermore, when identifying the road boundary, the first point cloud representing the target vehicle's driving boundary is first identified based on the position information of the reference point cloud within the grid. Then, it is further filtered based on the positional relationships between the first point clouds to obtain the target point cloud used to represent the road boundary. This technical solution solves the problem of low accuracy in road boundary detection in related technologies, achieving a significant improvement in the accuracy of road boundary detection.

[0138] Optionally, the segmentation module includes:

[0139] The first determining unit is used to determine the target angle ratio corresponding to the target curvature information from the curvature information and angle ratio that have a corresponding relationship, wherein the target angle ratio is used to characterize the regional angle situation of the segmented grid region;

[0140] The segmentation unit is used to segment the point cloud signal acquisition area according to the target angle ratio to obtain the plurality of grid areas, wherein the grid area is a fan-shaped grid area centered on the current position of the target vehicle.

[0141] Optionally, the second filtering module includes:

[0142] The first calculation unit is used to sequentially calculate the lateral position difference of the first point cloud corresponding to two adjacent grid regions;

[0143] A clustering unit is used to cluster the first point cloud based on the lateral position difference to obtain multiple point cloud clusters;

[0144] The first filtering unit is used to filter out a target point cloud cluster from the plurality of point cloud clusters based on cluster features, wherein the cluster features are used to characterize the spatial distribution of point clouds belonging to the corresponding point cloud cluster.

[0145] The second determining unit is used to determine the point cloud belonging to the target point cloud cluster as the target point cloud.

[0146] Optionally, the first filtering unit is used for:

[0147] Construct the connection trajectories between multiple point clouds included in the point cloud cluster;

[0148] If the trajectory length of the connection trajectory is greater than or equal to the target length, a quadratic curve fitting is performed on the multiple point clouds to obtain candidate point cloud clusters;

[0149] If the proportion of the second point cloud in the candidate point cloud cluster is greater than or equal to the target proportion, the candidate point cloud cluster is determined as the target point cloud cluster, wherein the second point cloud is the point cloud in the candidate point cloud cluster whose residual value is less than or equal to the target residual value.

[0150] Optionally, the clustering unit is used for:

[0151] If the lateral position difference is less than or equal to the target difference, the two first point clouds are classified into the same point cloud cluster.

[0152] If the lateral position difference is greater than the target difference, the position gradient relationship between the two first point clouds is obtained; if the position gradient relationship satisfies the target gradient relationship, the two first point clouds are classified into the same point cloud cluster.

[0153] Optionally, the first filtering module includes:

[0154] The second filtering unit is used to filter out stationary point clouds from the point clouds falling within each of the grid areas;

[0155] The second calculation unit is used to calculate the relative distance between the stationary point cloud and the target vehicle using the location information;

[0156] A determining unit is used to determine the point cloud whose relative distance is less than or equal to the target distance as the first point cloud.

[0157] Optionally, the device further includes:

[0158] The construction module is used to construct the road boundary trajectory of the road to be driven based on the positional relationship of the target point cloud after the target point cloud is selected from the first point clouds corresponding to multiple grid regions according to the positional relationship between the first point clouds.

[0159] The first marking module is used to mark the third point cloud in the grid area when the length of the road boundary trajectory is greater than or equal to the target length, wherein the third point cloud is the point cloud in the reference point cloud whose point cloud position is outside the boundary area of ​​the road to be driven.

[0160] The second marking module is used to mark the fourth point cloud in the grid area when the length of the road boundary trajectory is less than the target length. The fourth point cloud is the point cloud in the reference area whose point cloud position is outside the area position of the road to be driven in the reference point cloud. The reference area is the area corresponding to the road boundary trajectory.

[0161] Optionally, the device further includes:

[0162] The acquisition module is used to acquire a fifth point cloud and a sixth point cloud before the first point cloud is selected from the point clouds falling in each of the grid areas according to the location information. The fifth point cloud is used to characterize the road environment of the target vehicle's driving road at a reference time before the current time, and the sixth point cloud is the point cloud collected by the radar equipment deployed on the target vehicle at the current time.

[0163] The prediction module is used to predict the position of the fifth point cloud at the current moment, so as to obtain the seventh point cloud;

[0164] The merging module is used to merge the sixth point cloud and the seventh point cloud to obtain the reference point cloud.

[0165] Embodiments of this application also provide a storage medium including a stored program, wherein the program executes any of the point cloud filtering methods described above when it is run.

[0166] Optionally, in this embodiment, the storage medium can be configured to store program code for performing the following steps: when detecting the road boundary of the road to be driven for the target vehicle, segmenting the current point cloud signal acquisition area of ​​the target vehicle according to the target curvature information to obtain multiple grid regions, wherein the target curvature information is used to characterize the road curvature of the road to be driven, and the difference in the number of point clouds acquired in any two grid regions is less than or equal to a target threshold; determining the position information of a reference point cloud in the point cloud signal acquisition area in the multiple grid regions, wherein the reference point cloud is used to characterize the road environment of the road to be driven; filtering out a first point cloud from the point clouds falling in each grid region according to the position information, wherein the first point cloud is used to characterize the driving boundary of the target vehicle in the corresponding fan-shaped area; filtering out a target point cloud from the first point clouds corresponding to the multiple grid regions according to the positional relationship between the first point clouds, wherein the target point cloud is used to characterize the road boundary of the road to be driven.

[0167] Embodiments of this application also provide an electronic device including a memory and a processor, the memory storing a computer program, the processor being configured to run the computer program to perform the steps in any of the above-described point cloud screening method embodiments.

[0168] Optionally, the electronic device may further include a transmission device and an input / output device, wherein the transmission device is connected to the processor and the input / output device is connected to the processor.

[0169] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program: when detecting the road boundary of the road to be driven for the target vehicle, segmenting the current point cloud signal acquisition area of ​​the target vehicle according to the target curvature information to obtain multiple grid regions, wherein the target curvature information is used to characterize the road curvature of the road to be driven, and the difference in the number of point clouds acquired in any two grid regions is less than or equal to a target threshold; determining the position information of a reference point cloud in the point cloud signal acquisition area in the multiple grid regions, wherein the reference point cloud is used to characterize the road environment of the road to be driven; filtering out a first point cloud from the point clouds falling in each grid region according to the position information, wherein the first point cloud is used to characterize the driving boundary of the target vehicle in the corresponding fan-shaped area; filtering out a target point cloud from the first point clouds corresponding to the multiple grid regions according to the positional relationship between the first point clouds, wherein the target point cloud is used to characterize the road boundary of the road to be driven.

[0170] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0171] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated here.

[0172] Obviously, those skilled in the art should understand that the modules or steps of this application described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented here, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, this application is not limited to any particular combination of hardware and software.

[0173] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for filtering point clouds, characterized in that, include: When detecting the road boundary of the road to be driven for the target vehicle, the current point cloud signal acquisition area of ​​the target vehicle is divided into multiple grid areas according to the target curvature information. The target curvature information is used to characterize the road curvature of the road to be driven. The difference between the number of point clouds acquired in any two grid areas is less than or equal to the target threshold. The position information of a reference point cloud in the point cloud signal acquisition area is determined in the multiple grid areas, wherein the reference point cloud is used to characterize the road environment of the road to be driven; Based on the location information, a first point cloud is selected from the point clouds falling within each of the grid areas, wherein the first point cloud is used to characterize the driving boundary of the target vehicle within the corresponding sector area. Based on the positional relationship between the first point clouds, a target point cloud is selected from the first point clouds corresponding to multiple grid regions, wherein the target point cloud is used to characterize the road boundary of the road to be driven. The step of segmenting the current point cloud signal acquisition area of ​​the target vehicle according to the target curvature information to obtain multiple grid regions includes: determining the target angle ratio corresponding to the target curvature information from the corresponding curvature information and angle ratio, wherein the target angle ratio is used to characterize the regional angle situation of the segmented grid region; segmenting the point cloud signal acquisition area according to the target angle ratio to obtain the multiple grid regions, wherein the grid region is a fan-shaped grid region centered on the current position of the target vehicle.

2. The method according to claim 1, characterized in that, The step of filtering the target point cloud from the first point clouds corresponding to multiple grid regions based on the positional relationship between the first point clouds includes: The lateral position difference of the first point cloud corresponding to two adjacent grid regions is calculated sequentially. The first point cloud is clustered based on the lateral position difference to obtain multiple point cloud clusters; Target point cloud clusters are selected from the plurality of point cloud clusters based on cluster features, wherein the cluster features are used to characterize the spatial distribution of point clouds belonging to the corresponding point cloud clusters. Point clouds belonging to the target point cloud cluster are identified as the target point cloud.

3. The method according to claim 2, characterized in that, The step of selecting the target point cloud cluster from the plurality of point cloud clusters based on cluster characteristics includes: Construct the connection trajectories between multiple point clouds included in the point cloud cluster; If the trajectory length of the connection trajectory is greater than or equal to the target length, a quadratic curve fitting is performed on the multiple point clouds to obtain candidate point cloud clusters; If the proportion of the second point cloud in the candidate point cloud cluster is greater than or equal to the target proportion, the candidate point cloud cluster is determined as the target point cloud cluster, wherein the second point cloud is the point cloud in the candidate point cloud cluster whose residual value is less than or equal to the target residual value.

4. The method according to claim 2, characterized in that, The first point cloud is clustered based on the lateral position difference to obtain multiple point cloud clusters, including: If the lateral position difference is less than or equal to the target difference, the two first point clouds are classified into the same point cloud cluster. If the lateral position difference is greater than the target difference, the position gradient relationship between the two first point clouds is obtained; if the position gradient relationship satisfies the target gradient relationship, the two first point clouds are classified into the same point cloud cluster.

5. The method according to claim 1, characterized in that, The step of filtering out the first point cloud from the point clouds falling within each of the grid areas based on the location information includes: Select stationary point clouds from the point clouds falling within each of the grid regions; The relative distance between the stationary point cloud and the target vehicle is calculated using the location information; The point cloud whose relative distance is less than or equal to the target distance is defined as the first point cloud.

6. The method according to claim 1, characterized in that, After selecting the target point cloud from the first point clouds corresponding to the multiple grid regions based on the positional relationship between the first point clouds, the method further includes: Construct the road boundary trajectory of the road to be driven based on the positional relationship of the target point cloud; If the length of the road boundary trajectory is greater than or equal to the target length, the third point cloud in the grid area is marked as a point cloud, wherein the third point cloud is a point cloud in the reference point cloud whose point cloud position is outside the boundary area of ​​the road to be driven. If the length of the road boundary trajectory is less than the target length, the fourth point cloud in the grid area is marked with point cloud markers. The fourth point cloud is the point cloud in the reference area whose point cloud position is outside the area position of the road to be driven in the reference point cloud. The reference area is the area corresponding to the road boundary trajectory.

7. The method according to claim 1, characterized in that, Before filtering out the first point cloud from the point clouds falling within each of the grid regions based on the location information, the method further includes: Acquire a fifth point cloud and a sixth point cloud, wherein the fifth point cloud is used to characterize the road environment of the target vehicle's driving road at a reference time before the current time, and the sixth point cloud is the point cloud collected by the radar equipment deployed on the target vehicle at the current time. Predict the current position of the fifth point cloud to obtain the seventh point cloud; The sixth point cloud and the seventh point cloud are merged to obtain the reference point cloud.

8. A point cloud screening device, characterized in that, include: The segmentation module is used to segment the current point cloud signal acquisition area of ​​the target vehicle according to the target curvature information when detecting the road boundary of the road to be driven by the target vehicle, and obtain multiple grid regions. The target curvature information is used to characterize the road curvature of the road to be driven, and the difference between the number of point clouds acquired in any two grid regions is less than or equal to the target threshold. The determining module is used to determine the position information of a reference point cloud in the point cloud signal acquisition area in the multiple grid areas, wherein the reference point cloud is used to characterize the road environment of the road to be driven; The first filtering module is used to filter out a first point cloud from the point clouds falling within each of the grid areas based on the location information, wherein the first point cloud is used to characterize the driving boundary of the target vehicle within the corresponding fan-shaped area. The second filtering module is used to filter out target point clouds from the first point clouds corresponding to multiple grid regions based on the positional relationship between the first point clouds, wherein the target point cloud is used to characterize the road boundary of the road to be driven. The segmentation module includes: a first determining unit, configured to determine a target angle ratio corresponding to the target curvature information from curvature information and angle ratios with corresponding relationships, wherein the target angle ratio is used to characterize the regional angle situation of the segmented grid region; and a segmentation unit, configured to segment the point cloud signal acquisition area according to the target angle ratio to obtain the plurality of grid regions, wherein the grid region is a fan-shaped grid region centered on the current position of the target vehicle.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program, when executed, performs the method of any one of claims 1 to 7.

10. An electronic device comprising a memory and a processor, characterized in that, The memory stores a computer program, and the processor is configured to execute the method of any one of claims 1 to 7 through the computer program.

Images