A point cloud clustering method and device, electronic equipment and storage medium

Abstract

Embodiments of the present application provide a point cloud clustering method and device, electronic equipment and storage medium. The method comprises: obtaining a first point cloud set and motion attribute information of each point cloud cluster in the first point cloud set; determining a first point cloud cluster in the first point cloud set based on the motion attribute information of each point cloud cluster in the first point cloud set; the first point cloud cluster is a point cloud cluster whose motion attribute information indicates that the point cloud cluster is in a motion state; calculating a target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set; fusing other point cloud clusters with a target distance less than a preset distance threshold with the first point cloud cluster to obtain a second point cloud set; the target distance is the minimum distance between the first point cloud cluster and other point cloud clusters. In this way, the accuracy of point cloud clustering can be improved, i.e. the accuracy of target recognition is improved, thereby the accuracy of driving decision influence can be improved.

Description

Technical Field

[0001] This application relates to the field of lidar sensing, specifically to a point cloud clustering method, apparatus, electronic device, and storage medium. Background Technology

[0002] In autonomous vehicles, the environmental perception system is the information source for their driving decisions. Identifying and tracking vehicles, non-motorized vehicles, pedestrians, and other targets in traffic flow, and promptly inputting environmental obstacle information into the decision-making system, are crucial tasks of the environmental perception system and are key to the safe operation of autonomous vehicles. LiDAR can output three-dimensional spatial data, providing reliable depth information for visual sensors. In some technologies, environmental perception is achieved using LiDAR, and it is widely used in the field of autonomous driving.

[0003] When a lidar system measures the same target, it creates a point cloud. To differentiate between multiple targets detected simultaneously and estimate their states, the point cloud must first be clustered according to the different targets. Object identification and tracking often require clustering the point cloud and then distinguishing between different targets. Due to the limitations of lidar scanning, it often fails to capture the complete outline of an object. The point cloud obtained by lidar is relatively dispersed. When clustering the point cloud based on the distance between points, multiple point clouds obtained by the lidar may be identified as multiple targets, leading to inaccurate target identification and affecting the accuracy of subsequent velocity estimation, etc. Summary of the Invention

[0004] In view of this, this application provides a point cloud clustering method, apparatus, electronic device and storage medium to help solve the problem of inaccurate point cloud clustering in the prior art.

[0005] In a first aspect, embodiments of this application provide a point cloud clustering method, including:

[0006] Obtain the motion attribute information of a first point cloud set and each point cloud cluster within the first point cloud set; wherein, the first point cloud set contains at least one point cloud cluster, and the point cloud cluster is the clustering result obtained by clustering and segmenting point cloud data; the motion attribute information of the point cloud cluster is information used to identify whether the point cloud cluster is in motion or stationary state.

[0007] Based on the motion attribute information of each point cloud cluster in the first point cloud set, a first point cloud cluster is determined in the first point cloud set; the first point cloud cluster is the point cloud cluster whose motion attribute information identifies that the point cloud cluster is in motion.

[0008] Calculate the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster. Merge other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain a second point cloud cluster. The target distance is the minimum distance between the first point cloud cluster and other point cloud clusters.

[0009] In one possible implementation of the first aspect, the step of calculating the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set, and fusing other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain a second point cloud set includes:

[0010] The target point cloud cluster is identified in the second point cloud cluster of the first point cloud cluster; the second point cloud cluster is any other first point cloud cluster in the first point cloud cluster besides the one already identified as the target point cloud cluster.

[0011] Calculate the target distance between the target point cloud and each third point cloud; the third point cloud includes other second point clouds in the first point cloud set besides the target point cloud and point cloud clouds identified by motion attribute information as being in a stationary state;

[0012] The third point cloud cluster whose target distance is less than a preset distance threshold is merged with the target point cloud cluster, and the first point cloud cluster in the first point cloud set is updated.

[0013] Detect whether a second cloud cluster exists within the first cloud cluster;

[0014] If it exists, the steps are repeated to determine the target point cloud in the second point cloud in the first point cloud set until there is no second point cloud in the first point cloud set. The updated first point cloud set is then determined as the second point cloud set.

[0015] In one possible implementation of the first aspect, fusing the third point cloud cluster with a target distance less than a preset distance threshold with the target point cloud cluster to update the first point cloud cluster in the first point cloud set includes:

[0016] The third point cloud cluster whose target distance is less than a preset distance threshold is merged with the target point cloud cluster, the target point cloud cluster in the first point cloud set is updated to the merged target point cloud cluster, and the third point cloud cluster whose target distance is less than the preset distance threshold is deleted from the first point cloud set.

[0017] In one possible implementation of the first aspect, updating the target point cloud cluster in the first point cloud set to the fused target point cloud cluster includes:

[0018] The target point cloud cluster in the first point cloud set is updated to the fused target point cloud cluster, and the updated target point cloud cluster is marked as the second point cloud cluster.

[0019] One possible implementation of the first aspect also includes:

[0020] When the target distance between the target point cloud and each third point cloud is greater than a preset distance threshold, the step of determining the target point cloud in the second point cloud in the first point cloud set is repeated until the step of detecting whether there is a second point cloud in the first point cloud set is repeated until there is no second point cloud in the first point cloud set, and the updated first point cloud set is determined as the second point cloud set.

[0021] In one possible implementation of the first aspect, calculating the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set includes:

[0022] Based on the point cloud data of each point cloud in the first point cloud cluster and the point cloud data of each other point cloud cluster, determine the outer convex hull point of the first point cloud cluster and the outer convex hull point of the other point cloud cluster.

[0023] Based on the outer convex hull points of the first point cloud and the outer convex hull points of other point clouds, calculate the minimum distance between the convex hulls of the first point cloud and other point clouds.

[0024] In one possible implementation of the first aspect, obtaining the first point cloud set and the motion attribute information of each point cloud cluster in the first point cloud set includes:

[0025] Clustering and segmenting the point cloud data yields the first point cloud set containing at least one point cloud cluster.

[0026] Each point cloud cluster in the first point cloud set is tracked and processed to determine the motion attribute information of each point cloud cluster in the first point cloud set.

[0027] One possible implementation of the first aspect also includes:

[0028] The point cloud clusters contained in the second point cloud set are tracked to determine the tracking information of each point cloud cluster in the second point cloud set; the tracking information includes motion attribute information.

[0029] Secondly, embodiments of this application provide a point cloud clustering device, comprising:

[0030] The acquisition unit is used to acquire a first point cloud set and motion attribute information of each point cloud cluster in the first point cloud set; wherein, the first point cloud set contains at least one point cloud cluster, and the point cloud cluster is a clustering result obtained by clustering and segmenting point cloud data; the motion attribute information of the point cloud cluster is information used to identify whether the point cloud cluster is in motion or stationary state.

[0031] The determining unit is used to determine a first point cloud cluster in the first point cloud set based on the motion attribute information of each point cloud cluster in the first point cloud set; the first point cloud cluster is the point cloud cluster whose motion attribute information identifies that the point cloud cluster is in motion.

[0032] The processing unit is used to calculate the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster, and to merge other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain a second point cloud cluster; the target distance is the minimum distance between the first point cloud cluster and other point cloud clusters.

[0033] Thirdly, embodiments of this application provide an electronic device, including a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the electronic device is triggered to execute the method described in any of the first aspects above.

[0034] Fourthly, embodiments of this application provide a computer-readable storage medium including a stored program, wherein, when the program is executed, it controls the device where the computer-readable storage medium is located to perform the method described in any of the first aspects.

[0035] The scheme provided in this application embodiment obtains a first point cloud set and motion attribute information of each point cloud cluster within the first point cloud set. Based on the motion attribute information of each point cloud cluster in the first point cloud set, a first point cloud cluster is determined within the first point cloud set. The target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set is calculated. Point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold are fused with the first point cloud cluster to obtain a second point cloud set. In this way, after obtaining the first point cloud set by clustering and segmenting the point cloud data, the first point cloud clusters within the first point cloud set can be fused based on the target distance between them being less than a preset distance threshold to obtain the second point cloud set. That is, in this application embodiment, after obtaining the first point cloud set by clustering and segmenting the point cloud data, the first point cloud clusters in the point cloud set in motion state can be further fused again to reduce the identification of a moving object as multiple scattered point cloud clusters. This improves the accuracy of point cloud clustering, i.e., improves the accuracy of target recognition, thereby improving the accuracy of driving decision-making. Attached Figure Description

[0036] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0037] Figure 1a A schematic diagram illustrating a point cloud clustering method provided in an embodiment of this application;

[0038] Figure 1b A schematic diagram illustrating a scenario for another point cloud clustering method provided in an embodiment of this application;

[0039] Figure 2 A flowchart illustrating a point cloud clustering method provided in an embodiment of this application;

[0040] Figure 3 A flowchart illustrating another point cloud clustering method provided in this application embodiment;

[0041] Figure 4 A schematic diagram illustrating a scenario for another point cloud clustering method provided in an embodiment of this application;

[0042] Figure 5 A flowchart illustrating another point cloud clustering method provided in this application embodiment;

[0043] Figure 6 This is a schematic diagram of the structure of a point cloud clustering device provided in an embodiment of this application;

[0044] Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0045] To better understand the technical solution of this application, the embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0046] It should be understood that the described embodiments are merely some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0047] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. The singular forms “a,” “the,” and “the” used in the embodiments of this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0048] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0049] In related technologies, LiDAR measurements of the same target create point clouds. To differentiate and estimate the state of multiple targets detected simultaneously, the point cloud must first be clustered according to different targets. Object identification and tracking often require clustering the point cloud first and then distinguishing between different targets. Due to the limitations of LiDAR scanning, it often fails to capture the complete outline of an object. The point cloud obtained by LiDAR scanning a target is relatively scattered. For example, when scanning a truck, the LiDAR obtains a point cloud like... Figure 1a The point cloud shown is as follows: Figure 1a The point cloud shown is relatively scattered. When clustering the point cloud based on the distance between them, multiple point clouds obtained by the lidar will be identified as multiple targets, such as... Figure 1b As shown, inaccurate target identification affects the accuracy of subsequent velocity estimation and other parameters.

[0050] To address the aforementioned issues, this application provides a point cloud clustering method, apparatus, electronic device, and storage medium. The method involves acquiring a first point cloud set and motion attribute information of each point cloud cluster within the first point cloud set. Based on the motion attribute information of each point cloud cluster in the first point cloud set, first point cloud clusters are identified within the first point cloud set. When the first point cloud set contains at least two first point cloud clusters, the target distance between the at least two first point cloud clusters is calculated. First point cloud clusters with target distances less than a preset distance threshold are then fused to obtain a second point cloud set. In this way, after clustering and segmenting the point cloud data to obtain the first point cloud set, the first point cloud clusters within the first point cloud set can be fused based on the target distance between them being less than a preset distance threshold to obtain the second point cloud set. Specifically, in this application embodiment, after clustering and segmenting the point cloud data to obtain the first point cloud set, the first point cloud clusters in motion within the point cloud set can be further fused again to reduce the likelihood of a moving object being identified as multiple scattered point cloud clusters. This improves the accuracy of point cloud clustering, which in turn improves the accuracy of target recognition, thereby enhancing the accuracy of driving decision-making. The following is a detailed explanation.

[0051] See Figure 2 This is a flowchart illustrating a point cloud clustering method provided in an embodiment of this application. Figure 2 As shown, the method includes:

[0052] Step S201: Obtain the motion attribute information of the first point cloud cluster and each point cloud cluster in the first point cloud cluster.

[0053] The first point cloud set contains at least one point cloud cluster, which is the clustering result obtained by clustering and segmenting the point cloud data; the motion attribute information of the point cloud cluster is used to identify whether the point cloud cluster is in motion or at rest.

[0054] In this embodiment, after obtaining point cloud data through lidar scanning, the point cloud data can be segmented into ground segments. The segmented point cloud data can then be clustered, with each clustering result forming a point cloud cluster, thus creating a first point cloud set. The point cloud clusters within the first point cloud set are then tracked to determine the motion attribute information of each cluster.

[0055] The point cloud clustering device can acquire motion attribute information of a first point cloud set and each point cloud cluster within that first point cloud set. This information can be acquired from a storage device, other electronic devices, or by receiving user input of motion attribute information for the first point cloud set and each point cloud cluster within that first point cloud set. Alternatively, the point cloud clustering device can also acquire motion attribute information for the first point cloud set and each point cloud cluster within that first point cloud set through calculations on the point cloud data. This application does not impose any limitations on this.

[0056] As one possible implementation, obtaining the motion attribute information of the first point cloud set and each point cloud cluster within the first point cloud set includes:

[0057] The point cloud data is clustered and segmented to obtain a first point cloud set containing at least one point cloud cluster. Each point cloud cluster in the first point cloud set is tracked to determine its motion attribute information.

[0058] In other words, the point cloud clustering device can obtain the first point cloud set and the motion attribute information of each point cloud cluster within it by calculating the point cloud data. At this point, the point cloud clustering device acquires point cloud data scanned by LiDAR. Ground segmentation processing is performed on the point cloud data. Since the point cloud data scanned by LiDAR is three-dimensional data, the point cloud clustering device can predefine an xyz coordinate system based on the three-dimensional data. The ground is characterized by small z-coordinate values ​​and is approximately planar, with little difference in z-coordinate values ​​within the grid. Therefore, the point cloud clustering device can determine point cloud data whose z-coordinate values ​​are less than a preset height threshold as ground point cloud data based on the z-coordinate values ​​of each point cloud data, thus filtering out other point cloud data besides the ground. To reduce computational load, a grid method can also be used for calculation. In this case, a grid of preset size can be divided on the xy plane, for example, a 0.1m × 0.1m grid. The point cloud data is divided into different grids. Since the grid is smaller than the area occupied by a possible object, each grid can be considered to belong to only one object. The point cloud height of each grid is calculated based on the point cloud data contained within it. For example, the average value of the z-coordinate of the point cloud data contained in the grid is calculated, and this average value is used as the point cloud height of the grid. Grids with point cloud heights less than a preset height threshold are identified as ground point cloud grids, thus allowing the separation of non-ground point cloud grids.

[0059] After segmenting the ground point cloud data from the point cloud dataset, the point cloud clustering device can further segment the non-ground point cloud data. At this point, the device can employ a preset clustering algorithm, such as the K-means clustering algorithm, to cluster the non-ground point cloud data, resulting in the first point cloud set. When the preset clustering algorithm is K-means, the device can determine k cluster centers from the non-ground point cloud data. For each non-ground point cloud data point, the distance between that point and each cluster center is calculated. The non-ground point cloud data point is then assigned to the cluster with the nearest cluster center, resulting in k clusters. For each cluster, the mean of the point cloud data within that cluster is recalculated using its cluster center as the center, and this mean is used to update the cluster center. Repeat the above process until the cluster centers of the clusters no longer change, resulting in multiple clusters, each of which is a point cloud, thus obtaining the first point cloud set.

[0060] Of course, the point cloud clustering device can also use other clustering algorithms, such as hierarchical clustering or mean transfer clustering, etc., and this application does not limit it.

[0061] After obtaining the first point cloud set, the point cloud clustering device can track each point cloud cluster within it to determine whether each cluster is in motion or stationary, thus determining the motion attribute information of each cluster. The point cloud clustering device can employ the Hungarian algorithm for tracking.

[0062] Step S202: Based on the motion attribute information of each point cloud cluster in the first point cloud set, determine the first point cloud cluster in the first point cloud set.

[0063] Among them, the first point cloud is the point cloud whose motion attribute information indicates that the point cloud is in motion.

[0064] In this embodiment, when using point cloud data for driving decisions, the speed information of detected moving objects is required for accurate driving decisions. The first point cloud set contains not only moving point cloud clusters but also stationary ones. Since a stationary object, even if its corresponding point cloud data is identified as multiple point cloud clusters, has little impact on driving decisions due to its stationary state, further clustering is needed only for the moving point cloud clusters to more accurately identify objects and estimate their speed, etc. Based on this, after determining the motion attribute information of each point cloud cluster in the first point cloud set, the point cloud clustering device can identify the moving point cloud clusters by examining their motion attribute information; this is equivalent to identifying the first point cloud cluster in the first point cloud set.

[0065] Step S203: Calculate the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster. Merge other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain the second point cloud cluster.

[0066] The target distance is the minimum distance between the first point cloud cluster and other point cloud clusters.

[0067] In this embodiment, point cloud clusters closer to the lidar are relatively stable and can be scanned and clustered in each working cycle, thus being identified as being in motion. Point cloud clusters farther from the lidar are unstable, sometimes being scanned and sometimes not, and may be identified as being stationary. Therefore, a moving object may be identified as including multiple first point cloud clusters, or it may be identified as at least one first point cloud cluster and at least one point cloud cluster whose motion attribute information indicates it is stationary. To improve the clustering accuracy, the target distance between a first point cloud cluster and other point cloud clusters in the first point cloud cluster can be calculated. That is, for each first point cloud cluster, the target distance between that first point cloud cluster and other point cloud clusters in the first point cloud cluster is calculated. The target distance is compared with a preset distance threshold. If the target distance between the first point cloud cluster and other point cloud clusters is less than the preset distance threshold, it indicates that the first point cloud cluster is close to the other point cloud clusters and can be considered as belonging to the same object. In this case, the first point cloud cluster needs to be merged with the other point cloud clusters into a single first point cloud cluster. If the target distance between the first point cloud cluster and other point cloud clusters is greater than a preset distance threshold, it indicates that the first point cloud cluster is far from the other point cloud clusters and can be considered as belonging to different objects. In this way, other point cloud clusters whose target distance from the first point cloud cluster is less than the preset distance threshold can be merged with the first point cloud cluster to obtain the second point cloud cluster.

[0068] It should be understood that the preset distance threshold is a pre-set distance threshold used to identify whether two point clouds belong to the same object. When the target distance between two point clouds is less than the preset distance threshold, the two point clouds can be considered to belong to the same object, and in this case, the two point clouds need to be merged. When the target distance between two point clouds is greater than the preset distance threshold, the two point clouds can be considered to not belong to the same object, that is, they belong to two different objects, and in this case, there is no need to merge the two point clouds.

[0069] As one possible implementation, such as Figure 3 As shown, the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster is calculated. Other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold are then merged with the first point cloud cluster to obtain the second point cloud cluster, which includes:

[0070] Step S2031: Determine the target point cloud cluster from the second point cloud cluster of the first point cloud cluster.

[0071] Step S2032: Calculate the target distance between the target point cloud and each third point cloud.

[0072] Step S2033: Detect whether there is a target distance between the target point cloud and each third point cloud that is less than a preset distance threshold.

[0073] Step S2034: If it exists, merge the third point cloud cluster whose target distance is less than the preset distance threshold with the target point cloud cluster, and update the first point cloud cluster in the first point cloud cluster.

[0074] Step S2035: Detect whether there is a second point cloud cluster in the first point cloud cluster.

[0075] Step S2036: If it exists, then re-execute the step to determine the target point cloud in the second point cloud in the first point cloud set, until the step checks whether the second point cloud exists in the first point cloud set, until the second point cloud does not exist in the first point cloud set, and then determine the updated first point cloud set as the second point cloud set.

[0076] The second point cloud cluster is any other first point cloud cluster besides the one already identified as the target point cloud cluster. The third point cloud cluster includes any other second point cloud clusters in the first point cloud cluster besides the target point cloud cluster, as well as point cloud clusters identified by motion attribute information that are in a stationary state.

[0077] In this embodiment, the point cloud clustering device first needs to identify second point clouds within the first point cloud cluster of the first point cloud set. This means that all other first point clouds in the first point cloud set, excluding those already identified as the target point cloud, can be identified as second point clouds. At this point, during the initial fusion of the first point clouds, all first point clouds in the first point cloud set are considered second point clouds. The point cloud clustering device can then identify the target point cloud within the second point cloud cluster. For example, the point cloud clustering device can identify any one of the second point clouds as the target point cloud. After identifying the target point cloud, the point cloud clustering device can further identify third point clouds. This means that other second point clouds in the first point cloud set besides the target point cloud and point clouds with stationary motion attribute information can be identified as third point clouds. The point cloud clustering device can calculate the target distance between the target point cloud and each third point cloud. In other words, the point cloud clustering device can calculate the target distance between the target point cloud and each third point cloud. After calculating the target distance between the target point cloud cluster and each third point cloud cluster, the target distance between the target point cloud cluster and each third point cloud cluster can be compared with a preset distance threshold to determine whether there are any third point cloud clusters whose target distance from the target point cloud cluster is less than the preset distance threshold. These third point cloud clusters with target distances less than the preset distance threshold are then merged with the target point cloud cluster, forming a single first point cloud cluster. Since the merging of the target point cloud cluster and the third point cloud clusters with target distances less than the preset distance threshold necessitates updating the first point cloud cluster record. After updating the first point cloud cluster record, it is possible to further detect whether other second point cloud clusters need to be merged. At this point, the point cloud clustering device needs to detect whether there are any second point cloud clusters in the first point cloud cluster, i.e., whether there are any other first point cloud clusters besides the one already identified as the target point cloud cluster. If it exists, it is necessary to continue determining whether the second point cloud cluster needs to be merged with other point cloud clusters in the first point cloud cluster. At this point, the steps described above for identifying the target point cloud cluster in the second point cloud cluster of the first point cloud cluster can be repeated until the step of detecting whether a second point cloud cluster exists in the first point cloud cluster is detected, until it is determined that no second point cloud cluster exists in the first point cloud cluster. When it is determined that no second point cloud cluster exists in the first point cloud cluster, it means that each first point cloud cluster in the first point cloud cluster has been identified as a target point cloud cluster, i.e., each has been determined whether it can be merged with other point cloud clusters. At this time, the point cloud clustering device can determine the updated first point cloud cluster as the second point cloud cluster.

[0078] It should be understood that when the point cloud clustering device determines that the target distance between at least two third point cloud clusters and the target point cloud cluster is less than a preset distance threshold, the target point cloud cluster is fused with both of the at least two third point cloud clusters.

[0079] As one possible implementation, the third point cloud cluster with a target distance less than a preset distance threshold is merged with the target point cloud cluster, and the first point cloud cluster in the first point cloud set is updated by:

[0080] The third point cloud cluster whose target distance is less than a preset distance threshold is merged with the target point cloud cluster. The target point cloud cluster in the first point cloud cluster is updated to the merged target point cloud cluster. The third point cloud cluster whose target distance is less than the preset distance threshold is deleted from the first point cloud cluster.

[0081] That is, when the point cloud device merges a third point cloud cluster whose target distance from the target point cloud cluster is less than a preset distance threshold and the target point cloud cluster, it merges the third point cloud cluster whose target distance from the target point cloud cluster is less than the preset distance threshold and the target point cloud cluster into a single point cloud cluster. At this time, the merged point cloud cluster can be used as the new target point cloud cluster. The target point cloud cluster recorded in the first point cloud cluster can be updated to the merged point cloud cluster. Since the third point cloud cluster whose target distance from the target point cloud cluster is less than the preset distance threshold has been merged with the target point cloud cluster, the third point cloud cluster whose target distance from the target point cloud cluster is less than the preset distance threshold in the first point cloud cluster can be deleted, thus completing the update of the first point cloud cluster in the first point cloud cluster.

[0082] As one possible implementation, to more accurately merge different point cloud clusters of an object, the target point cloud cluster to be merged can be used as a new first point cloud cluster, and the above point cloud cluster merging process can be repeated. In this case, updating the target point cloud cluster in the first point cloud cluster to the merged target point cloud cluster includes:

[0083] Update the target point cloud cluster in the first point cloud cluster to the merged target point cloud cluster, and mark the updated target point cloud cluster as the second point cloud cluster.

[0084] That is, after updating the target point cloud clusters in the first point cloud cluster to the fused target point cloud clusters, in order to perform fusion processing on the fused target point cloud clusters, the updated target point cloud clusters can be marked as the second point cloud clusters, i.e., marked as not identified as target point cloud clusters. In this way, when the point cloud clustering device determines the second point cloud cluster next time, since the updated target point cloud clusters are marked as not identified as target point cloud clusters, the second point cloud clusters determined by the point cloud clustering device at this time include the updated target point cloud clusters.

[0085] As one possible implementation, when detecting whether there is a target distance less than a preset distance threshold between the target point cloud cluster and each third point cloud cluster in step 2033 above, the point cloud clustering device can determine two possible results. One result is that there is a target distance less than the preset distance threshold, in which case the point cloud clustering device needs to execute step S2034 above. The other result is that there is no target distance less than the preset distance threshold, that is, the target distance between the target point cloud cluster and each third point cloud cluster is greater than the preset distance threshold, in which case step S2031 needs to be re-executed, such as... Figure 3 As shown.

[0086] That is, when the target distance between the target point cloud and each third point cloud is greater than the preset distance threshold, the step is repeated to determine the target point cloud in the second point cloud of the first point cloud set, until there is no second point cloud in the first point cloud set, and the updated first point cloud set is determined as the second point cloud set.

[0087] That is, when the point cloud clustering device detects that the target point cloud cluster and each third point cloud cluster have target distances greater than a preset distance threshold, it indicates that the target point cloud cluster and each third point cloud cluster do not belong to the same object. At this time, it can re-detect whether the other first point cloud clusters in the first point cloud set need to be merged with other point cloud clusters. The point cloud clustering device can re-execute the above step S2031 until there are no second point cloud clusters in the first point cloud set. At this time, the point cloud clustering device will determine the updated first point cloud set as the second point cloud set.

[0088] As one possible implementation, when the point cloud clustering device detects that the target distance between the target point cloud cluster and each third point cloud cluster is greater than a preset distance threshold, it can first detect whether a second point cloud cluster exists in the first point cloud cluster. If a second point cloud cluster exists in the first point cloud cluster, the target point cloud cluster is identified from the second point cloud clusters in the first point cloud cluster, until no second point cloud cluster exists in the first point cloud cluster, at which point the updated first point cloud cluster is identified as the second point cloud cluster.

[0089] As one possible implementation, to facilitate the calculation of the minimum distance between point cloud clusters, the minimum distance of the convex hull between two point cloud clusters can be calculated. In this case, the target distance is the minimum distance of the convex hull between the point cloud clusters. The above calculation of the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set includes:

[0090] Based on the point cloud data of each point cloud in the first point cloud cluster and the point cloud data of each other point cloud cluster, determine the outer convex hull points of the first point cloud cluster and the outer convex hull points of the other point cloud clusters; based on the outer convex hull points of the first point cloud cluster and the outer convex hull points of the other point cloud clusters, calculate the minimum distance between the convex hulls of the first point cloud cluster and the other point cloud clusters.

[0091] In this embodiment of the application, for each first point cloud cluster and other point cloud clusters in the first point cloud set, the outer convex hull point of the point cloud cluster can be determined based on the point cloud data contained within the point cloud cluster. For example, if a point cloud cluster contains 13 point cloud data points, such as... Figure 4 As shown. Since the point cloud data scanned by LiDAR is three-dimensional data, the point cloud clustering device can predefine an xyz coordinate system based on the three-dimensional data. In the first point cloud cluster, a base point p0 needs to be found among the 13 point cloud data. The point cloud data with the smallest y-value can be used as the base point p0. If there are multiple point cloud data with the smallest y-value, the point cloud data with the smallest x-value can be selected as the base point p0. After determining the base point p0, the other point cloud data can be sorted counterclockwise by the polar angles formed by the base point p0. The polar angle refers to the angle between the other point cloud data and the base point p0 relative to the x-axis. The point cloud data are labeled p1-p12 in ascending order of polar angle. (Refer to...) Figure 4As shown. If multiple polar angles are the same, only the point cloud data farthest from the base point p0 is retained, and the other point cloud data are marked as non-outer convex hull points. In this example, we will use the example of 12 point cloud data with different polar angles from the base point p0. Among the 13 point cloud data, the point cloud data are added to the convex point stack in sequence according to the size of the polar angle. In this example, when adding a point, it is necessary to consider whether the preceding line segment is on the convex hull. Starting from the base point, the rotation direction of each adjacent line segment on the convex hull should be consistent and opposite to the scanning direction. If it is found that the newly added point causes the rotation direction of the new line segment to change with the previous line segment, it can be determined that the previous point must not be on the convex hull, that is, it is not an outer convex hull point. Since point p0 is the reference point, it must be an outer convex hull point, so p0 is added to the convex point stack. When determining whether point p1 is an outer convex hull point, it is necessary to use the outer convex hull point before p1 and the point cloud data after p1 to construct two adjacent line segments with p1. At this point, points p1 and p2 can be added to the convex point stack. Base points p0 and p1 form line segment p01. Points p1 and p2 form line segment p12. Line segments p01 to p12 are right-handed, opposite to the counter-clockwise scanning direction. Therefore, it can be determined that line segment p01 lies on the convex hull, and point p1 is an outer convex hull point. Point p1 is then kept in the convex point stack. Next, it is determined whether point p2 is an outer convex hull point. To determine this, the previous outer convex hull point and the next point cloud data of point p2 are used to construct two adjacent line segments around point p2. At this point, point p3 is added to the convex point stack. The previous outer convex hull point of point p2 is p1, the line segment between p1 and p2 is p12, and the next point cloud data of p2 is p3. Thus, points p2 and p3 form line segment p23. Line segments p12 to p23 rotate leftward, which is different from the rotation direction of line segment p01 and the same as the counterclockwise scanning direction. Therefore, it can be determined that line segment p12 is not on the convex hull, and thus point p2 is not an outer convex hull point. At this point, point p2 is removed from the convex hull stack. Next, it is necessary to further determine whether point p3 is an outer convex hull point. Similarly, when determining whether point p3 is an outer convex hull point, the previous outer convex hull point and the next point cloud data of point p3 are used to construct two adjacent line segments with point p3. At this point, point p4 is added to the convex point stack. The previous outer convex hull point of point p3 is p1, the line segment formed between point p1 and point p3 is p13, and the next point cloud data of point p3 is point p4. At this point, point p3 and point p4 form line segment p34. Line segments p13 to p34 rotate right-handed, in the same direction as line segment p01, but opposite to the counter-clockwise scanning direction. Therefore, we can determine that line segment p34 lies on the convex hull, and thus point p3 is a point on the outer convex hull. We then keep point p3 in the convex hull stack. Similarly, through the above process, we can determine whether points p4 to p12 are points on the outer convex hull. In this example, we can determine that points p5, p10, and p12 are points on the outer convex hull.That is, by using the above method, the convex hull of the point cloud and the outer convex hull points are obtained, where the reference points p0, p1, p3, p5, p10, and p12 are the outer convex hull points, for reference. Figure 4 As shown.

[0092] It should be noted that when a point cloud cluster contains only one point cloud data point, that point cloud data point can be identified as the outer convex hull point of the first point cloud cluster. When a point cloud cluster contains two point cloud data points, those two point cloud data points can be identified as the outer convex hull points of the point cloud cluster.

[0093] After determining the outer convex hull points of each point cloud cluster, the point cloud clustering device can calculate the minimum distance between the convex hulls of the first point cloud cluster and other point cloud clusters in the first point cloud set, based on the outer convex hull points of that point cloud cluster. For example, the minimum distance between the convex hulls of the first point cloud cluster and other point cloud clusters in the first point cloud set can be calculated using a rotating caliper algorithm, and two point cloud clusters whose minimum distance between convex hulls is less than a preset distance threshold can be merged.

[0094] In this way, after clustering and segmenting the point cloud data to obtain the first point cloud set, the first point cloud clusters within the first point cloud set can be fused if the target distance between the point cloud clusters is less than a preset distance threshold, resulting in the second point cloud set. That is, in this embodiment, after clustering and segmenting the point cloud data to obtain the first point cloud set, the first point cloud clusters in motion within the point cloud set can be further fused again to reduce the likelihood of a moving object being identified as multiple scattered point cloud clusters. This improves the accuracy of point cloud clustering, which in turn improves the accuracy of target recognition, thereby enhancing the accuracy of driving decision-making.

[0095] See Figure 5 This is a flowchart illustrating a point cloud clustering method provided in an embodiment of this application. Figure 5 As shown, the method includes:

[0096] Step S501: Obtain the motion attribute information of the first point cloud cluster and each point cloud cluster in the first point cloud cluster.

[0097] For details, please refer to step S201, which will not be repeated here.

[0098] Step S502: Based on the motion attribute information of each point cloud cluster in the first point cloud set, determine the first point cloud cluster in the first point cloud set.

[0099] For details, please refer to step S202, which will not be repeated here.

[0100] Step S503: Calculate the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster. Merge other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain the second point cloud cluster.

[0101] For details, please refer to step S203, which will not be repeated here.

[0102] Step S504: Perform tracking processing on the point cloud clusters contained in the second point cloud set to determine the tracking information of each point cloud cluster in the second point cloud set.

[0103] The tracking information includes motion attribute information.

[0104] In this embodiment, point cloud clusters in the first point cloud set are fused to obtain a second point cloud set. The second point cloud set includes the fused first point cloud clusters, first point cloud clusters not fused with other first point cloud clusters, and point cloud clusters whose motion attribute information indicates they are stationary. Since the point cloud clusters in the second point cloud set have changed, they need to be re-tracked to obtain tracking information for subsequent velocity estimation. At this time, the point cloud clustering device can perform tracking processing on the point cloud clusters in the second point cloud set to obtain tracking information after the tracking processing. Through tracking processing, the positional changes of each point cloud cluster can be obtained, thus the resulting tracking information includes motion attribute information. Through tracking processing, the historical period during which the target corresponding to the point cloud cluster was detected can also be obtained, allowing for a more accurate estimation of its future motion trend. Because the longer the target is monitored, the richer its historical motion information becomes, and the more accurate its velocity estimation. Thus, the accuracy of the velocity estimation for the point cloud clusters can be improved using the obtained tracking information.

[0105] refer to Figure 6 This is a schematic diagram of the structure of a vehicle data acquisition device provided in an embodiment of this application. Figure 6 As shown, the device includes:

[0106] The acquisition unit 601 is used to acquire the motion attribute information of the first point cloud set and each point cloud cluster in the first point cloud set.

[0107] The first point cloud set contains at least one point cloud cluster, which is the clustering result obtained by clustering and segmenting the point cloud data; the motion attribute information of the point cloud cluster is used to identify whether the point cloud cluster is in motion or at rest.

[0108] The determining unit 602 is used to determine the first point cloud cluster in the first point cloud cluster based on the motion attribute information of each point cloud cluster in the first point cloud cluster.

[0109] Among them, the first point cloud is the point cloud whose motion attribute information indicates that the point cloud is in motion.

[0110] The processing unit 603 calculates the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster, and merges other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain the second point cloud cluster.

[0111] The target distance is the minimum distance between the first point cloud cluster and other point cloud clusters.

[0112] As one possible implementation, the processing unit 603 is specifically used to determine the target point cloud cluster in the second point cloud cluster of the first point cloud set; calculate the target distance between the target point cloud cluster and each third point cloud cluster; merge the third point cloud clusters whose target distance is less than a preset distance threshold with the target point cloud cluster to update the first point cloud cluster in the first point cloud set; detect whether there is a second point cloud cluster in the first point cloud set; if there is, repeat the step of determining the target point cloud cluster in the second point cloud cluster of the first point cloud set until the step of detecting whether there is a second point cloud cluster in the first point cloud set is executed again until there is no second point cloud cluster in the first point cloud set, and then determine the updated first point cloud set as the second point cloud set.

[0113] The second point cloud cluster is any other first point cloud cluster in the first point cloud cluster besides the one already identified as the target point cloud cluster; the third point cloud cluster includes any other second point cloud clusters in the first point cloud cluster besides the target point cloud cluster, as well as point cloud clusters identified by motion attribute information as being in a stationary state.

[0114] As one possible implementation, the processing unit 603 is specifically used to merge the third point cloud cluster whose target distance is less than a preset distance threshold with the target point cloud cluster, update the target point cloud cluster in the first point cloud cluster to the merged target point cloud cluster, and delete the third point cloud cluster whose target distance is less than the preset distance threshold from the first point cloud cluster.

[0115] As one possible implementation, the processing unit 603 is specifically used to update the target point cloud cluster in the first point cloud set to the fused target point cloud cluster, and to mark the updated target point cloud cluster as the second point cloud cluster.

[0116] As one possible implementation, when the target distance between the target point cloud and each third point cloud is greater than a preset distance threshold, the processing unit 603 is specifically used to re-execute the step of determining the target point cloud in the second point cloud of the first point cloud set until there is no second point cloud in the first point cloud set, and then determine the updated first point cloud set as the second point cloud set.

[0117] As one possible implementation, the processing unit 603 is specifically used to determine the outer convex hull points of the first point cloud cluster and the outer convex hull points of other point cloud clusters based on each point cloud data in the first point cloud cluster and each point cloud data in other point cloud clusters; and to calculate the minimum distance between the convex hulls of the first point cloud cluster and other point cloud clusters based on the outer convex hull points of the first point cloud cluster and the outer convex hull points of other point cloud clusters.

[0118] As one possible implementation, the acquisition unit 601 is specifically used to cluster and segment the point cloud data to obtain a first point cloud set containing at least one point cloud cluster; and to perform tracking processing on each point cloud cluster in the first point cloud set to determine the motion attribute information of each point cloud cluster in the first point cloud set.

[0119] As one possible implementation, the processing unit 603 is also used to perform tracking processing on the point cloud clusters contained in the second point cloud set, and determine the tracking information of each point cloud cluster in the second point cloud set; the tracking information includes motion attribute information.

[0120] Corresponding to the above embodiments, this application also provides an electronic device. Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The electronic device 700 may include a processor 701, a memory 702, and a communication unit 703. These components communicate through one or more buses. Those skilled in the art will understand that the structure of the electronic device shown in the figure does not constitute a limitation on the embodiment of the present invention. It may be a bus topology or a star topology, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0121] The communication unit 703 is used to establish a communication channel, enabling the electronic device to communicate with other devices. It can receive user data sent by other devices or send user data to other devices.

[0122] The processor 701 serves as the control center of the electronic device, connecting various parts of the device via interfaces and lines. It executes software programs and / or modules stored in the memory 702, and calls data stored in the memory to perform various functions and / or process data. The processor can be composed of integrated circuits (ICs), such as a single packaged IC or multiple packaged ICs with the same or different functions connected together. For example, the processor 701 may consist only of a central processing unit (CPU). In this embodiment, the CPU may have a single processing core or include multiple processing cores.

[0123] The memory 702 is used to store the execution instructions of the processor 701. The memory 702 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0124] When the execution instructions in memory 702 are executed by processor 701, the electronic device 700 is able to perform operations. Figure 2 or Figure 5 Some or all of the steps in the illustrated embodiments.

[0125] In a specific implementation, the present invention also provides a computer storage medium, wherein the computer storage medium may store a program, which, when executed, may include some or all of the steps of the point cloud clustering method provided by the present invention in various embodiments. The storage medium may be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0126] Those skilled in the art will clearly understand that the techniques in the embodiments of the present invention can be implemented using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0127] The same or similar parts between the various embodiments in this specification can be referred to mutually. In particular, the device embodiments and terminal embodiments are basically similar to the method embodiments, so the description is relatively simple, and the relevant parts can be referred to the description in the method embodiments.

Claims

1. A point cloud clustering method, characterized in that, include: Obtain the motion attribute information of a first point cloud set and each point cloud cluster within the first point cloud set; wherein, the first point cloud set contains at least one point cloud cluster, the point cloud cluster being the clustering result obtained by clustering and segmenting point cloud data, the point cloud data being obtained by lidar scanning; the motion attribute information of the point cloud cluster is information used to identify whether the point cloud cluster is in motion or stationary state. Based on the motion attribute information of each point cloud cluster in the first point cloud set, a first point cloud cluster is determined in the first point cloud set; the first point cloud cluster is the point cloud cluster whose motion attribute information identifies that the point cloud cluster is in motion. Calculate the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster. Merge other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain a second point cloud cluster. The target distance is the minimum distance between the first point cloud cluster and other point cloud clusters. The second point cloud cluster is used for target recognition and driving decision-making.

2. The method according to claim 1, characterized in that, The step of calculating the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set, and fusing other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain the second point cloud set includes: The target point cloud cluster is identified from the second point cloud cluster of the first point cloud cluster; the second point cloud cluster is any other first point cloud cluster in the first point cloud cluster besides the one already identified as the target point cloud cluster. Calculate the target distance between the target point cloud and each third point cloud; the third point cloud includes other second point clouds in the first point cloud set besides the target point cloud and point cloud clouds identified by motion attribute information as being in a stationary state; The third point cloud cluster whose target distance is less than a preset distance threshold is merged with the target point cloud cluster, and the first point cloud cluster in the first point cloud set is updated. Detect whether a second cloud cluster exists within the first cloud cluster; If it exists, the step is repeated to determine the target point cloud in the second point cloud in the first point cloud set. The step continues to check whether the second point cloud exists in the first point cloud set until the second point cloud set does not exist. The updated first point cloud set is then determined as the second point cloud set.

3. The method according to claim 2, characterized in that, The step of fusing the third point cloud cluster whose target distance is less than a preset distance threshold with the target point cloud cluster to update the first point cloud cluster in the first point cloud set includes: The third point cloud cluster whose target distance is less than a preset distance threshold is merged with the target point cloud cluster, the target point cloud cluster in the first point cloud set is updated to the merged target point cloud cluster, and the third point cloud cluster whose target distance is less than the preset distance threshold is deleted from the first point cloud set.

4. The method according to claim 3, characterized in that, Updating the target point cloud cluster in the first point cloud set to the fused target point cloud cluster includes: The target point cloud cluster in the first point cloud set is updated to the fused target point cloud cluster, and the updated target point cloud cluster is marked as the second point cloud cluster.

5. The method according to claim 2, characterized in that, Also includes: When the target distance between the target point cloud and each third point cloud is greater than a preset distance threshold, the step is repeated to determine the target point cloud in the second point cloud in the first point cloud set until there is no second point cloud in the first point cloud set, and the updated first point cloud set is determined as the second point cloud set.

6. The method according to claim 1, characterized in that, The calculation of the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud set includes: Based on the point cloud data of each point cloud in the first point cloud cluster and the point cloud data of each other point cloud cluster, determine the outer convex hull point of the first point cloud cluster and the outer convex hull point of the other point cloud cluster. Based on the outer convex hull points of the first point cloud and the outer convex hull points of other point clouds, calculate the minimum distance between the convex hulls of the first point cloud and other point clouds.

7. The method according to claim 1, characterized in that, The process of obtaining the first point cloud set and the motion attribute information of each point cloud cluster in the first point cloud set includes: Clustering and segmenting the point cloud data yields the first point cloud set containing at least one point cloud cluster. Each point cloud cluster in the first point cloud set is tracked and processed to determine the motion attribute information of each point cloud cluster in the first point cloud set.

8. The method according to any one of claims 1-7, characterized in that, Also includes: The point cloud clusters contained in the second point cloud set are tracked to determine the tracking information of each point cloud cluster in the second point cloud set; The tracking information includes motion attribute information.

9. A point cloud clustering device, characterized in that, include The acquisition unit is used to acquire a first point cloud set and motion attribute information of each point cloud cluster in the first point cloud set; wherein, the first point cloud set contains at least one point cloud cluster, the point cloud cluster is the clustering result obtained by clustering and segmenting point cloud data, and the point cloud data is acquired by lidar scanning; the motion attribute information of the point cloud cluster is information used to identify whether the point cloud cluster is in motion or stationary state. The determining unit is used to determine a first point cloud cluster in the first point cloud set based on the motion attribute information of each point cloud cluster in the first point cloud set; the first point cloud cluster is the point cloud cluster whose motion attribute information identifies that the point cloud cluster is in motion. The processing unit is used to calculate the target distance between the first point cloud cluster and other point cloud clusters in the first point cloud cluster, and to fuse other point cloud clusters whose target distance from the first point cloud cluster is less than a preset distance threshold with the first point cloud cluster to obtain a second point cloud cluster; the target distance is the minimum distance between the first point cloud cluster and other point cloud clusters, and the second point cloud cluster is used for target recognition and driving decision.

10. An electronic device, characterized in that, The device includes a memory for storing computer program instructions and a processor for executing the program instructions, wherein when the computer program instructions are executed by the processor, the electronic device is triggered to perform the method according to any one of claims 1-8.

11. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the computer-readable storage medium is located to perform the method according to any one of claims 1-8.

Images