Target detection method and target detection device based on clustering optimization

Abstract

The application provides a target detection method and device based on cluster optimization. The target detection method comprises: performing cluster processing on at least original point cloud data to obtain at least one cluster, determining a target ratio according to a total area of the at least one cluster and a total area corresponding to the original point cloud data, and determining the size of a target voxel based on the target ratio and a predetermined rule. The original point cloud data corresponds to an original image, and the original point cloud data is point cloud data of a target region obtained by a laser radar. The target region is a region spaced apart from an autonomous vehicle by a predetermined distance. The predetermined rule is a rule for determining the size of the target voxel. The original image corresponding to the original point cloud data is divided according to the size of the target voxel to obtain a corresponding target pseudo image, and target detection is performed based on the target pseudo image, thereby solving the problem that the division of voxels in the prior art is unreasonable.

Description

Technical Field

[0001] This application relates to the field of autonomous driving technology, and more specifically, to a target detection method based on clustering optimization, a target detection device, a computer-readable storage medium, a processor, and an autonomous vehicle. Background Technology

[0002] LiDAR is one of the most important sensors in the field of autonomous driving, and it is indispensable for high-level autonomous vehicles. Due to its high accuracy, short detection time, and ability to acquire depth information, LiDAR-based autonomous driving algorithms are constantly emerging, especially in the field of object detection.

[0003] In existing technologies, target detection algorithms based on LiDAR can be mainly divided into two directions: voxel-based and point-based. Among them, voxel-based algorithms have gradually become the mainstream due to their performance advantages. However, voxel-based algorithms also have some related drawbacks. For example, the accuracy and performance of voxel-based algorithms are highly dependent on the size of the voxels. If the voxels are large, there will be more information loss in a frame of point cloud data; if the voxels are small, the computational load of the algorithm will be large and the memory consumption will also increase accordingly.

[0004] Currently, in voxel-based algorithms, the division of voxels in a frame of point cloud data often relies on human experience. This makes the voxel division of a frame of point cloud data rather unreasonable, resulting in inaccurate target detection results.

[0005] Therefore, there is an urgent need for a more reasonable method to automatically divide voxels. Summary of the Invention

[0006] The main objective of this application is to provide a target detection method, target detection device, computer-readable storage medium, processor, and autonomous vehicle based on clustering optimization, so as to solve the problem of unreasonable voxel division in the prior art.

[0007] According to one aspect of the present invention, a target detection method based on clustering optimization is provided, comprising: performing clustering processing on at least one original point cloud data to obtain at least one cluster; determining a target ratio based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data; determining the size of a target voxel based on the target ratio and a predetermined rule; wherein the original point cloud data corresponds to one frame of original image, and the original point cloud data is point cloud data of a target region acquired by a LiDAR, the target region is a region separated from an autonomous driving vehicle by a predetermined distance, and the predetermined rule is a rule for determining the size of the target voxel; dividing the original image corresponding to the original point cloud data according to the size of the target voxel to obtain a corresponding target pseudo image; and performing target detection based on the target pseudo image.

[0008] Optionally, at least the original point cloud data is subjected to clustering processing to obtain at least one cluster, including: filtering the original point cloud data to obtain filtered original point cloud data; and clustering the filtered original point cloud data to obtain at least one cluster.

[0009] Optionally, at least the original point cloud data is clustered to obtain at least one cluster, including: clustering the original point cloud data to obtain at least one cluster; and filtering the at least one cluster to obtain at least one filtered cluster.

[0010] Optionally, determining the target ratio based on the total area of ​​at least one of the clusters and the total area corresponding to the original point cloud data includes: performing a viewpoint transformation on the target point cloud data in at least one of the clusters to obtain the total area of ​​at least one of the clusters from the BEV viewpoint; performing a viewpoint transformation on the original point cloud data to obtain the total area corresponding to the original point cloud data from the BEV viewpoint; and calculating the ratio of the total area of ​​at least one of the clusters from the BEV viewpoint to the total area corresponding to the original point cloud data from the BEV viewpoint to obtain the target ratio.

[0011] Optionally, determining the size of the target voxel based on the target ratio and a predetermined rule includes: determining the size of the target voxel corresponding to the target ratio from the predetermined rule based on the target ratio.

[0012] Optionally, target detection based on the target pseudo-image includes: not performing a predetermined operation on regions in the target pseudo-image that do not have the cluster, wherein the predetermined operation includes at least one of the following: convolution operation and pooling operation; and performing the predetermined operation on regions in the target pseudo-image that have the cluster.

[0013] According to another aspect of the present invention, a target detection device based on clustering optimization is also provided, comprising: a clustering processing unit, configured to perform clustering processing on at least one original point cloud data to obtain at least one cluster, and determine a target ratio based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data, and determine the size of a target voxel based on the target ratio and a predetermined rule, wherein the original point cloud data corresponds to one frame of original image, and the original point cloud data is point cloud data of a target region acquired by a lidar, the target region is a region separated from an autonomous driving vehicle by a predetermined distance, and the predetermined rule is a rule for determining the size of the target voxel; and a segmentation unit, configured to segment the original image corresponding to the original point cloud data according to the size of the target voxel to obtain a corresponding target pseudo image, and perform target detection based on the target pseudo image.

[0014] According to another aspect of the present invention, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored program, wherein the program executes any of the described clustering-based target detection methods.

[0015] According to another aspect of the present invention, a processor is also provided, the processor being configured to run a program, wherein the program, when running, executes any of the described clustering-based target detection methods.

[0016] According to one aspect of the present invention, an autonomous vehicle is also provided, comprising: a lidar; and a target detection device based on clustering optimization, the target detection device being used to execute any of the target detection methods based on clustering optimization described above.

[0017] In this embodiment of the invention, the target detection method based on clustering optimization firstly performs clustering processing on the acquired original point cloud data of the target region to obtain one or more clusters; if one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data; if multiple clusters are obtained, the target ratio is determined based on the total area of ​​the multiple clusters and the total area of ​​the original point cloud data; then, the size of the target voxel is determined based on the target ratio and a predetermined rule; finally, based on the determined size of the target voxel, the original image corresponding to the original point cloud is divided to obtain a target pseudo-image corresponding to the original image, and target detection is performed based on the target pseudo-image. Compared to existing technologies that rely on human experience to divide the original image into voxels, the target detection method in this application determines the target ratio based on the total area of ​​the clusters obtained from clustering processing and the total area corresponding to the original point cloud data. Then, based on the target ratio and predetermined rules, the size of the target voxels is determined. This ensures that the determined target voxel sizes are reasonable, thus solving the problem of unreasonable voxel division in existing technologies. This application divides the original image into voxels based on the size of the target voxels, obtaining a target pseudo-image corresponding to the original image. Target detection is then performed based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, and thus improves the performance of the target detection algorithm, balancing the relationship between detection accuracy and algorithm performance. Attached Figure Description

[0018] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0019] Figure 1 A flowchart of a cluster-optimized target detection method according to an embodiment of this application is shown;

[0020] Figure 2 A schematic diagram of the structure of a target detection device based on clustering optimization according to an embodiment of this application is shown. Detailed Implementation

[0021] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0022] 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.

[0023] 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 for the embodiments of this application 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.

[0024] As mentioned in the background section, the classification of voxels in the prior art is rather unreasonable. In order to solve the above problems, in a typical embodiment of this application, a target detection method, a target detection device, a computer-readable storage medium, a processor, and an autonomous vehicle based on cluster optimization are provided.

[0025] According to an embodiment of this application, a target detection method based on clustering optimization is provided.

[0026] Figure 1 This is a flowchart of a cluster-based target detection method according to an embodiment of this application. Figure 1 As shown, the target detection method includes the following steps:

[0027] Step S101: At least the original point cloud data is clustered to obtain at least one cluster. The target ratio is determined based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data. The size of the target voxel is determined based on the target ratio and a predetermined rule. The original point cloud data corresponds to one frame of original image. The original point cloud data is the point cloud data of the target area acquired by the LiDAR. The target area is the area at a predetermined distance from the autonomous vehicle. The predetermined rule is the rule used to determine the size of the target voxel.

[0028] Step S102: Based on the size of the target voxel, the original image corresponding to the original point cloud data is divided to obtain the corresponding target pseudo image, and target detection is performed based on the target pseudo image.

[0029] In the aforementioned clustering-based target detection method, firstly, the original point cloud data of the acquired target region is clustered to obtain one or more clusters. If only one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data. If multiple clusters are obtained, the target ratio is determined based on the total area of ​​the clusters and the total area of ​​the original point cloud data. Then, the target voxel size is determined based on the target ratio and a predetermined rule. Finally, based on the determined target voxel size, the original image corresponding to the original point cloud is divided to obtain a target pseudo-image, and target detection is performed based on the target pseudo-image. Compared with the prior art's voxel division of the original image based on human experience, the target detection method of this application determines the target ratio based on the total area of ​​the clusters obtained through clustering and the total area corresponding to the original point cloud data; then, based on the target ratio and a predetermined rule, the target voxel size is determined. This ensures that the determined target voxel size is more reasonable, thereby solving the problem of unreasonable voxel division in the prior art. This application divides the original image into voxels based on the size of the target voxels to obtain a target pseudo-image corresponding to the original image, and performs target detection based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, thereby improving the performance of the target detection algorithm and balancing the relationship between detection accuracy and algorithm performance.

[0030] In practical applications, when clustering the original point cloud data to obtain a single cluster, the target ratio is determined based on the area of ​​the cluster and the total area of ​​the original point cloud data. When clustering the original point cloud data to obtain multiple clusters, the target ratio is determined based on the sum of the areas of the multiple clusters (i.e., the total area) and the total area of ​​the original point cloud data.

[0031] Specifically, the aforementioned raw point cloud data is the environmental information around the autonomous vehicle acquired by the LiDAR at a certain moment. That is, the aforementioned raw point cloud data can be the point cloud data corresponding to a frame of raw image.

[0032] Specifically, this application does not limit the specific clustering method used to cluster the original point cloud data; any feasible clustering method in the prior art can be used to cluster the original point cloud data. For example, Spectral Clustering (SC), Hierarchical Clustering (HC), and Mean-shift clustering.

[0033] In one specific embodiment of this application, the size of the target voxel corresponding to each cluster can be determined based on the area of ​​each cluster and the total area corresponding to the original point cloud data. Based on the position information of the minimum bounding rectangle surrounding each cluster in the original image, the target image corresponding to each cluster is extracted from the original image. Then, based on the size of the target voxel corresponding to each cluster, the corresponding target image is divided into voxel segments to obtain each target image after voxel segmentation. Finally, target detection is performed using each target image after voxel segmentation.

[0034] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0035] To reduce the impact of isolated point cloud data in the original point cloud data on clustering processing and to ensure that the obtained clusters are relatively accurate, and further to ensure that the size of the target voxel determined based on the ratio of the total area of ​​at least one cluster to the total area of ​​the original point cloud data is relatively accurate, in one embodiment of this application, at least the original point cloud data is subjected to clustering processing to obtain at least one cluster, including: filtering the original point cloud data to obtain filtered original point cloud data; and performing clustering processing on the filtered original point cloud data to obtain at least one cluster.

[0036] In another embodiment of this application, at least one clustering process is performed on the original point cloud data to obtain at least one cluster, including: performing clustering processing on the original point cloud data to obtain at least one cluster; and performing filtering processing on the at least one cluster to obtain at least one filtered cluster. This ensures that the obtained clusters are relatively accurate, and further ensures that the size of the target voxel determined based on the ratio of the total area of ​​the at least one cluster to the total area of ​​the original point cloud data is relatively accurate and reasonable. Subsequently, the original image is divided into voxels based on the size of the target voxel to obtain a target pseudo image, which further ensures that there is more effective point cloud data in the obtained target pseudo image.

[0037] In practical applications, any feasible filtering method in the existing technology can be used to filter the original point cloud data or the clusters. This application does not limit the specific filtering methods described above. Examples include voxel filters, bilateral filtering, and spatial distribution-based denoising algorithms (Statistical Outlier Removal, SOR), etc.

[0038] To simplify the calculation of the target ratio and further ensure that the computational load of the target detection method of this application is low, in another embodiment of this application, the target ratio is determined based on the total area of ​​at least one of the above-mentioned clusters and the total area corresponding to the original point cloud data. This includes: performing a viewpoint transformation on the target point cloud data in at least one of the above-mentioned clusters to obtain the total area of ​​at least one of the above-mentioned clusters from the BEV viewpoint; performing a viewpoint transformation on the original point cloud data to obtain the total area corresponding to the original point cloud data from the BEV viewpoint; and calculating the ratio of the total area of ​​at least one of the above-mentioned clusters from the BEV viewpoint to the total area corresponding to the original point cloud data from the BEV viewpoint to obtain the target ratio.

[0039] In one specific embodiment of this application, the process of performing a perspective transformation on the target point cloud data in at least one of the above-mentioned clusters to obtain the total area of ​​at least one of the above-mentioned clusters from the BEV perspective is as follows: performing a perspective transformation on the target point cloud data in at least one cluster to obtain the clusters from the BEV perspective; and then calculating the total area of ​​at least one cluster.

[0040] In another specific embodiment of this application, the specific process of performing a viewpoint transformation on the above-mentioned original point cloud data to obtain the total area corresponding to the above-mentioned original point cloud data under the BEV viewpoint is as follows: performing a viewpoint transformation on the original point cloud data to obtain the original point cloud data under the BEV viewpoint; calculating the original point cloud data under the BEV viewpoint.

[0041] In another embodiment of this application, determining the size of a target voxel based on the target ratio and predetermined rules includes: determining the size of the target voxel corresponding to the target ratio from the predetermined rules based on the target ratio. This allows for a simpler determination of the size of the target voxel based on the target ratio and predetermined rules.

[0042] In practical applications, the aforementioned predetermined rules can also be the predefined relationship between the target ratio and the size of the target voxel. After determining the target ratio, the target voxel corresponding to the target ratio is determined based on the target ratio.

[0043] In practical applications, target pseudo-images can be input into a neural network for target detection. To further improve the computational performance of target detection, in one embodiment of this application, target detection based on the aforementioned target pseudo-image includes: not performing a predetermined operation on regions in the target pseudo-image that do not have the aforementioned clusters, wherein the predetermined operation includes at least one of the following: convolution operation and pooling operation; and performing the aforementioned predetermined operation on regions in the target pseudo-image that have the aforementioned clusters. In this embodiment, the predetermined operation is not performed on regions in the target pseudo-image that do not have clusters, i.e., regions without features; and the predetermined operation is performed on regions in the target pseudo-image that have clusters, i.e., regions with features, thereby further improving the computational performance of the detection algorithm.

[0044] In practical applications, the target detection algorithm of this application can be any feasible target detection algorithm in the prior art. This application does not limit the specific target detection algorithm, and it can be flexibly adjusted according to the actual use case.

[0045] This application also provides a target detection device based on clustering optimization. It should be noted that the target detection device based on clustering optimization provided in this application can be used to execute the target detection method based on clustering optimization provided in this application. The target detection device based on clustering optimization provided in this application will be described below.

[0046] Figure 2 This is a schematic diagram of the target detection device based on clustering optimization according to an embodiment of this application. Figure 2 As shown, the target detection device includes:

[0047] Clustering processing unit 10 is used to perform clustering processing on the original point cloud data to obtain at least one cluster, and determine a target ratio based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data, and determine the size of the target voxel based on the target ratio and a predetermined rule, wherein the original point cloud data corresponds to one frame of original image, and the original point cloud data is point cloud data of the target area acquired by the LiDAR, the target area is the area at a predetermined distance from the autonomous vehicle, and the predetermined rule is a rule used to determine the size of the target voxel.

[0048] The segmentation unit 20 is used to segment the original image corresponding to the original point cloud data according to the size of the target voxel to obtain the corresponding target pseudo image, and to perform target detection based on the target pseudo image.

[0049] In the aforementioned target detection device based on clustering optimization, the clustering processing unit performs clustering processing on the acquired original point cloud data of the target region to obtain one or more clusters; when one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data; when multiple clusters are obtained, a target ratio is determined based on the total area of ​​the multiple clusters and the total area of ​​the original point cloud data; then, based on the target ratio and a predetermined rule, the size of the target voxel is determined; the segmentation unit is used to segment the original image corresponding to the original point cloud based on the determined size of the target voxel to obtain a target pseudo-image corresponding to the original image, and to perform target detection based on the target pseudo-image. Compared to existing technologies that rely on human experience to divide the original image into voxels, the target detection device in this application determines the target ratio based on the total area of ​​the clusters obtained from clustering processing and the total area corresponding to the original point cloud data. Then, based on the target ratio and predetermined rules, the size of the target voxels is determined. This ensures that the determined target voxel sizes are reasonable, thus solving the problem of unreasonable voxel division in existing technologies. This application divides the original image into voxels based on the size of the target voxels, obtaining a target pseudo-image corresponding to the original image. Target detection is then performed based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, and thus improves the performance of the target detection algorithm, balancing the relationship between detection accuracy and algorithm performance.

[0050] In practical applications, when clustering the original point cloud data to obtain a single cluster, the target ratio is determined based on the area of ​​the cluster and the total area of ​​the original point cloud data. When clustering the original point cloud data to obtain multiple clusters, the target ratio is determined based on the sum of the areas of the multiple clusters (i.e., the total area) and the total area of ​​the original point cloud data.

[0051] Specifically, the aforementioned raw point cloud data is the environmental information around the autonomous vehicle acquired by the LiDAR at a certain moment. That is, the aforementioned raw point cloud data can be the point cloud data corresponding to a frame of raw image.

[0052] Specifically, this application does not limit the specific clustering method used to cluster the original point cloud data; any feasible clustering method in the prior art can be used to cluster the original point cloud data. For example, Spectral Clustering (SC), Hierarchical Clustering (HC), and Mean-shift clustering.

[0053] In one specific embodiment of this application, the size of the target voxel corresponding to each cluster can be determined based on the area of ​​each cluster and the total area corresponding to the original point cloud data. Based on the position information of the minimum bounding rectangle surrounding each cluster in the original image, the target image corresponding to each cluster is extracted from the original image. Then, based on the size of the target voxel corresponding to each cluster, the corresponding target image is divided into voxel segments to obtain each target image after voxel segmentation. Finally, target detection is performed using each target image after voxel segmentation.

[0054] To reduce the impact of isolated point cloud data in the original point cloud data on clustering processing and to ensure that the obtained clusters are relatively accurate, and further to ensure that the size of the target voxel determined based on the ratio of the total area of ​​at least one cluster to the total area of ​​the original point cloud data is relatively accurate, in one embodiment of this application, the clustering processing unit includes a filtering processing module and a clustering processing module. The filtering processing module is used to filter the original point cloud data to obtain filtered original point cloud data; the clustering processing module is used to perform clustering processing on the filtered original point cloud data to obtain at least one cluster.

[0055] In another embodiment of this application, the clustering processing module is further configured to perform clustering processing on the original point cloud data to obtain at least one cluster; the filtering processing module is configured to perform filtering processing on the at least one cluster to obtain at least one filtered cluster. This ensures that the obtained clusters are relatively accurate, and further ensures that the size of the target voxel determined based on the ratio of the total area of ​​at least one cluster to the total area of ​​the original point cloud data is relatively accurate and reasonable. Subsequently, the original image is divided into voxels based on the size of the target voxel to obtain a target pseudo image, further ensuring that there is more effective point cloud data in the obtained target pseudo image.

[0056] In practical applications, any feasible filtering method in the existing technology can be used to filter the original point cloud data or the clusters. This application does not limit the specific filtering methods described above. Examples include voxel filters, bilateral filtering, and spatial distribution-based denoising algorithms (Statistical Outlier Removal, SOR), etc.

[0057] To simplify the calculation of the target ratio and further ensure that the computational load of the target detection method of this application is low, in another embodiment of this application, the clustering processing unit further includes a first view conversion module, a second view conversion module, and a calculation module. The first view conversion module performs view conversion on the target point cloud data in at least one of the clusters to obtain the total area of ​​at least one cluster under the BEV view. The second view conversion module performs view conversion on the original point cloud data to obtain the total area corresponding to the original point cloud data under the BEV view. The calculation module calculates the ratio of the total area of ​​at least one cluster under the BEV view to the total area corresponding to the original point cloud data under the BEV view, thus obtaining the target ratio.

[0058] In one specific embodiment of this application, the process of performing a perspective transformation on the target point cloud data in at least one of the above-mentioned clusters to obtain the total area of ​​at least one of the above-mentioned clusters from the BEV perspective is as follows: performing a perspective transformation on the target point cloud data in at least one cluster to obtain the clusters from the BEV perspective; and then calculating the total area of ​​at least one cluster.

[0059] In another specific embodiment of this application, the specific process of performing a viewpoint transformation on the above-mentioned original point cloud data to obtain the total area corresponding to the above-mentioned original point cloud data under the BEV viewpoint is as follows: performing a viewpoint transformation on the original point cloud data to obtain the original point cloud data under the BEV viewpoint; calculating the original point cloud data under the BEV viewpoint.

[0060] In another embodiment of this application, the clustering processing unit further includes a determination module, which is used to determine the size of the target voxel corresponding to the target ratio from the predetermined rules based on the target ratio. This allows the size of the target voxel to be determined more simply based on the target ratio and the predetermined rules.

[0061] In practical applications, the aforementioned predetermined rules can also be the predefined relationship between the target ratio and the size of the target voxel. After determining the target ratio, the target voxel corresponding to the target ratio is determined based on the target ratio.

[0062] In practical applications, target pseudo-images can be input into a neural network for target detection. To further improve the computational performance of target detection, in one embodiment of this application, the partitioning unit includes a first execution module and a second execution module. The first execution module is used to not perform a predetermined operation on regions in the target pseudo-image that do not have the aforementioned clusters. The predetermined operation includes at least one of the following: convolution operation and pooling operation. The second execution module is used to perform the predetermined operation on regions in the target pseudo-image that have the aforementioned clusters. In this embodiment, the predetermined operation is not performed on regions in the target pseudo-image that do not have clusters, i.e., regions without features; the predetermined operation is performed on regions in the target pseudo-image that have clusters, i.e., regions with features. This can further improve the computational performance of the detection algorithm.

[0063] In practical applications, the target detection algorithm of this application can be any feasible target detection algorithm in the prior art. This application does not limit the specific target detection algorithm, and it can be flexibly adjusted according to the actual use case.

[0064] The aforementioned target detection device based on clustering optimization includes a processor and a memory. The clustering processing unit and partitioning unit are stored in the memory as program units, and the processor executes the program units stored in the memory to achieve the corresponding functions.

[0065] The processor contains a kernel, which retrieves the corresponding program units from memory. One or more kernels can be configured, and adjusting kernel parameters can address the problem of unreasonable voxel partitioning in existing technologies.

[0066] The memory may include non-permanent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0067] This invention provides a computer-readable storage medium storing a program that, when executed by a processor, implements the above-described target detection method based on clustering optimization.

[0068] This invention provides a processor for running a program, wherein the program executes the target detection method based on clustering optimization.

[0069] In a typical embodiment of this application, an autonomous vehicle is also provided. The autonomous vehicle includes a lidar system and a cluster-optimized target detection device, which is used to execute any of the above-described cluster-optimized target detection methods.

[0070] The aforementioned autonomous vehicle includes a LiDAR and a cluster-optimized target detection device, which executes any of the aforementioned cluster-optimized target detection methods. In the aforementioned target detection method, firstly, the acquired original point cloud data of the target region is clustered to obtain one or more clusters; if one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data; if multiple clusters are obtained, the target ratio is determined based on the total area of ​​the multiple clusters and the total area of ​​the original point cloud data; then, the target voxel size is determined based on the target ratio and a predetermined rule; finally, based on the determined target voxel size, the original image corresponding to the original point cloud is divided to obtain a target pseudo-image corresponding to the original image, and target detection is performed based on the target pseudo-image. Compared to existing technologies that rely on human experience to divide the original image into voxels, the target detection method in this application determines the target ratio based on the total area of ​​the clusters obtained from clustering processing and the total area corresponding to the original point cloud data. Then, based on the target ratio and predetermined rules, the size of the target voxels is determined. This ensures that the determined target voxel sizes are reasonable, thus solving the problem of unreasonable voxel division in existing technologies. This application divides the original image into voxels based on the size of the target voxels, obtaining a target pseudo-image corresponding to the original image. Target detection is then performed based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, and thus improves the performance of the target detection algorithm, balancing the relationship between detection accuracy and algorithm performance.

[0071] This invention provides a device including a processor, a memory, and a program stored in the memory and executable on the processor. When the processor executes the program, it performs at least the following steps:

[0072] Step S101: At least the original point cloud data is clustered to obtain at least one cluster. The target ratio is determined based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data. The size of the target voxel is determined based on the target ratio and a predetermined rule. The original point cloud data corresponds to one frame of original image. The original point cloud data is the point cloud data of the target area acquired by the LiDAR. The target area is the area at a predetermined distance from the autonomous vehicle. The predetermined rule is the rule used to determine the size of the target voxel.

[0073] Step S102: Based on the size of the target voxel, the original image corresponding to the original point cloud data is divided to obtain the corresponding target pseudo image, and target detection is performed based on the target pseudo image.

[0074] The devices mentioned in this article can be servers, PCs, tablets, mobile phones, etc.

[0075] This application also provides a computer program product, which, when executed on a data processing device, is suitable for executing an initialization program having at least the following method steps:

[0076] Step S101: At least the original point cloud data is clustered to obtain at least one cluster. The target ratio is determined based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data. The size of the target voxel is determined based on the target ratio and a predetermined rule. The original point cloud data corresponds to one frame of original image. The original point cloud data is the point cloud data of the target area acquired by the LiDAR. The target area is the area at a predetermined distance from the autonomous vehicle. The predetermined rule is the rule used to determine the size of the target voxel.

[0077] Step S102: Based on the size of the target voxel, the original image corresponding to the original point cloud data is divided to obtain the corresponding target pseudo image, and target detection is performed based on the target pseudo image.

[0078] In the above embodiments of the present invention, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0079] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units described above can be a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0080] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0081] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0082] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0083] As can be seen from the above description, the embodiments of this application achieve the following technical effects:

[0084] 1) In the target detection method based on clustering optimization of this application, firstly, the original point cloud data of the acquired target region is clustered to obtain one or more clusters; if one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data; if multiple clusters are obtained, a target ratio is determined based on the total area of ​​the clusters and the total area of ​​the original point cloud data; then, the size of the target voxel is determined based on the target ratio and a predetermined rule; finally, based on the determined size of the target voxel, the original image corresponding to the original point cloud is divided to obtain a target pseudo-image corresponding to the original image, and target detection is performed based on the target pseudo-image. Compared to existing technologies that rely on human experience to divide the original image into voxels, the target detection method in this application determines the target ratio based on the total area of ​​the clusters obtained from clustering processing and the total area corresponding to the original point cloud data. Then, based on the target ratio and predetermined rules, the size of the target voxels is determined. This ensures that the determined target voxel sizes are reasonable, thus solving the problem of unreasonable voxel division in existing technologies. This application divides the original image into voxels based on the size of the target voxels, obtaining a target pseudo-image corresponding to the original image. Target detection is then performed based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, and thus improves the performance of the target detection algorithm, balancing the relationship between detection accuracy and algorithm performance.

[0085] 2) In the target detection device based on clustering optimization of this application, the clustering processing unit is used to perform clustering processing on the acquired original point cloud data of the target region to obtain one or more clusters; when one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data; when multiple clusters are obtained, a target ratio is determined based on the total area of ​​the multiple clusters and the total area of ​​the original point cloud data; then, the size of the target voxel is determined based on the target ratio and a predetermined rule; the segmentation unit is used to segment the original image corresponding to the original point cloud based on the determined size of the target voxel to obtain a target pseudo image corresponding to the original image, and to perform target detection based on the target pseudo image. Compared to existing technologies that rely on human experience to divide the original image into voxels, the target detection device in this application determines the target ratio based on the total area of ​​the clusters obtained from clustering processing and the total area corresponding to the original point cloud data. Then, based on the target ratio and predetermined rules, the size of the target voxels is determined. This ensures that the determined target voxel sizes are reasonable, thus solving the problem of unreasonable voxel division in existing technologies. This application divides the original image into voxels based on the size of the target voxels, obtaining a target pseudo-image corresponding to the original image. Target detection is then performed based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, and thus improves the performance of the target detection algorithm, balancing the relationship between detection accuracy and algorithm performance.

[0086] 3) The autonomous vehicle of this application includes a LiDAR and a target detection device based on clustering optimization. The target detection device is used to execute any of the above-described target detection methods based on clustering optimization. In the above-described target detection method, firstly, the original point cloud data of the acquired target region is clustered to obtain one or more clusters; if one cluster is obtained, a target ratio is determined based on the total area of ​​the cluster and the total area of ​​the original point cloud data; if multiple clusters are obtained, a target ratio is determined based on the total area of ​​the multiple clusters and the total area of ​​the original point cloud data; then, the size of the target voxel is determined based on the target ratio and a predetermined rule; finally, based on the determined size of the target voxel, the original image corresponding to the original point cloud is divided to obtain a target pseudo-image corresponding to the original image, and target detection is performed based on the target pseudo-image. Compared to existing technologies that rely on human experience to divide the original image into voxels, the target detection method in this application determines the target ratio based on the total area of ​​the clusters obtained from clustering processing and the total area corresponding to the original point cloud data. Then, based on the target ratio and predetermined rules, the size of the target voxels is determined. This ensures that the determined target voxel sizes are reasonable, thus solving the problem of unreasonable voxel division in existing technologies. This application divides the original image into voxels based on the size of the target voxels, obtaining a target pseudo-image corresponding to the original image. Target detection is then performed based on the target pseudo-image. This ensures that the target pseudo-image contains more effective information, reduces the amount of point cloud data in the target pseudo-image, and thus improves the performance of the target detection algorithm, balancing the relationship between detection accuracy and algorithm performance.

[0087] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A target detection method based on clustering optimization, characterized in that, include: At least one clustering process is performed on the original point cloud data to obtain at least one cluster. A target ratio is determined based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data. The size of the target voxel is determined based on the target ratio and a predetermined rule. The original point cloud data corresponds to one frame of original image. The original point cloud data is the point cloud data of the target area acquired by the LiDAR. The target area is the area at a predetermined distance from the autonomous vehicle. The predetermined rule is a rule used to determine the size of the target voxel. The predetermined rule is a predefined relationship between the target ratio and the size of the target voxel. Based on the size of the target voxel, the original image corresponding to the original point cloud data is divided to obtain the corresponding target pseudo image, and target detection is performed based on the target pseudo image; Determining the size of a target voxel based on the target ratio and predetermined rules includes: determining the size of the target voxel corresponding to the target ratio from the predetermined rules based on the target ratio; Determining a target ratio based on the total area of ​​at least one of the clusters and the total area corresponding to the original point cloud data includes: performing a viewpoint transformation on the target point cloud data in at least one of the clusters to obtain the total area of ​​at least one of the clusters from a BEV viewpoint; performing a viewpoint transformation on the original point cloud data to obtain the total area corresponding to the original point cloud data from the BEV viewpoint; and calculating the ratio of the total area of ​​at least one of the clusters from the BEV viewpoint to the total area corresponding to the original point cloud data from the BEV viewpoint to obtain the target ratio.

2. The target detection method according to claim 1, characterized in that, At least one cluster should be obtained by performing clustering processing on the original point cloud data, including: The original point cloud data is filtered to obtain the filtered original point cloud data. The filtered original point cloud data is subjected to clustering processing to obtain at least one cluster.

3. The target detection method according to claim 1, characterized in that, At least one cluster should be obtained by performing clustering processing on the original point cloud data, including: The original point cloud data is subjected to clustering processing to obtain at least one cluster. At least one of the clusters is filtered to obtain at least one filtered cluster.

4. The target detection method according to claim 1, characterized in that, Target detection based on the target pseudo-image includes: For regions in the target pseudo-image that do not have the cluster, no predetermined operation is performed. The predetermined operation includes at least one of the following: convolution operation and pooling operation. The predetermined operation is performed on the regions in the target pseudo-image that contain the clusters.

5. A target detection device based on clustering optimization, characterized in that, include: A clustering processing unit is configured to perform clustering processing on at least the original point cloud data to obtain at least one cluster, determine a target ratio based on the total area of ​​the at least one cluster and the total area corresponding to the original point cloud data, and determine the size of the target voxel based on the target ratio and a predetermined rule. The original point cloud data corresponds to one frame of original image, and the original point cloud data is point cloud data of a target area acquired by a LiDAR. The target area is an area separated from the autonomous vehicle by a predetermined distance. The predetermined rule is a rule for determining the size of the target voxel, and the predetermined rule is a predefined relationship between the target ratio and the size of the target voxel. The segmentation unit is used to segment the original image corresponding to the original point cloud data according to the size of the target voxel to obtain the corresponding target pseudo image, and to perform target detection based on the target pseudo image; Determining the size of a target voxel based on the target ratio and predetermined rules includes: determining the size of the target voxel corresponding to the target ratio from the predetermined rules based on the target ratio; The clustering processing unit further includes a first perspective conversion module, a second perspective conversion module, and a calculation module. The first perspective conversion module performs perspective conversion on target point cloud data within at least one of the clusters to obtain the total area of ​​at least one cluster under the BEV perspective. The second perspective conversion module performs perspective conversion on the original point cloud data to obtain the total area corresponding to the original point cloud data under the BEV perspective. The calculation module calculates the ratio of the total area of ​​at least one cluster under the BEV perspective to the total area corresponding to the original point cloud data under the BEV perspective, obtaining the target ratio.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein the program executes the target detection method based on clustering optimization as described in any one of claims 1 to 4.

7. A processor, characterized in that, The processor is used to run a program, wherein the program executes the target detection method based on clustering optimization as described in any one of claims 1 to 4.

8. An autonomous vehicle, characterized in that, include: LiDAR; The target detection device based on cluster optimization is used to execute the target detection method based on cluster optimization as described in any one of claims 1 to 4.

Images