Point cloud segmentation method and apparatus, and electronic device

Abstract

The invention provides a point cloud segmentation method and device and electronic equipment, and relates to the field of smart mines, automatic driving and unmanned vehicles. The point cloud segmentation method comprises the following steps: dividing an original point cloud to obtain a plurality of partitioned point clouds; determining a surface feature corresponding to each block point cloud, wherein the surface feature comprises at least one of a plane feature, a non-plane feature and a roughness feature; determining that each block point cloud has a corresponding ground attribute or a non-ground attribute based on the surface feature corresponding to each block point cloud; and based on the ground attribute or the non-ground attribute corresponding to each partitioned point cloud, segmenting the original point cloud to obtain a ground point cloud and a non-ground point cloud. The partitioning strategy can be flexibly adjusted for different scenes, and ground point clouds and non-ground point clouds can be clearly distinguished without complex parameter adjustment.

Description

Technical Field

[0001] The present application relates to the fields of smart mines, autonomous driving, and unmanned vehicle technology, and specifically to a point cloud segmentation method and device, and electronic equipment. Background Art

[0002] In the current production process of high-precision maps of mining areas, the quality of point cloud segmentation has a direct impact on the quality of boundary updates, which in turn determines the effect of map updates.

[0003] At present, traditional point cloud segmentation methods, such as clustering and segmenting through region growing using features such as the normal vector curvature of the point cloud, or using RANSAC (Random Sample Consensus) plane fitting, are not only time-consuming, but also rely on too many parameters and cannot adapt to all scenarios, especially in lane-changing scenarios that require rapid support for the passage of driverless vehicles. This has become a pain point for map updates. Summary of the invention

[0004] In view of this, embodiments of the present application provide a point cloud segmentation method and device, and an electronic device.

[0005] In a first aspect, an embodiment of the present application provides a point cloud segmentation method, including: dividing an original point cloud to obtain a plurality of block point clouds; determining the surface features corresponding to each block point cloud, the surface features including at least one of planar features, non-planar features and roughness features; based on the surface features corresponding to each block point cloud, determining that each block point cloud has corresponding ground attributes or non-ground attributes; based on the ground attributes or non-ground attributes corresponding to each block point cloud, segmenting the original point cloud to obtain a ground point cloud and a non-ground point cloud.

[0006] In combination with the first aspect, in certain implementations of the first aspect, based on the surface features corresponding to each block point cloud, it is determined that each block point cloud has corresponding ground attributes or non-ground attributes, including: for a first block point cloud with planar features among multiple block point clouds, if the first block point cloud meets the ground judgment condition, it is determined that the first block point cloud has ground attributes, otherwise, it is determined that the first block point cloud has non-ground attributes; based on the first block point cloud with ground attributes, an ideal ground model is generated; for a second block point cloud with non-planar features and / or roughness features among multiple block point clouds, based on the ideal ground model, it is determined that the second block point cloud has corresponding ground attributes or non-ground attributes.

[0007] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: determining an initial ground seed point, and searching for a first block point cloud within the neighborhood of the initial ground seed point; if the angle between the normal vector of the first block point cloud within the neighborhood of the initial ground seed point and the normal vector of the initial ground seed point is less than a target angle threshold, and the height difference between the first block point cloud within the neighborhood of the initial ground seed point and the initial ground seed point is less than a first height threshold, then determining that the first block point cloud within the neighborhood of the initial ground seed point meets the ground judgment condition; determining the first block point cloud within the neighborhood of the initial ground seed point as the target ground seed point, and taking the target ground seed point as a reference, repeating the steps of searching, judging, and updating the target ground seed point until all first block point clouds are traversed; preferably, determining the initial ground seed point based on the height of the point cloud acquisition device above the ground.

[0008] In combination with the first aspect, in some implementations of the first aspect, generating an ideal ground model based on the first block point cloud with ground attributes includes: expanding the resolution of the first block point cloud with ground attributes to obtain a high-resolution block point cloud; aggregating the high-resolution block point clouds within the target range of the center with any high-resolution block point cloud as the center, and fitting the target ground equation using the aggregated point cloud data;

[0009] The target ground equation is determined as the ideal ground model.

[0010] In combination with the first aspect, in certain implementations of the first aspect, based on the ideal ground model, determining that the second block point cloud has corresponding ground attributes or non-ground attributes includes: calculating the height difference between the second block point cloud and the ideal ground model; if the height difference between the second block point cloud and the ideal ground model is less than a second height threshold, determining that the second block point cloud has ground attributes; if the height difference between the second block point cloud and the ideal ground model is greater than or equal to the second height threshold, determining that the second block point cloud has non-ground attributes.

[0011] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: determining the block point clouds contained in the non-ground point cloud; detecting whether the block point clouds contained in the non-ground point cloud are mobile targets; if the detection result indicates that there is a specific block point cloud that is a mobile target, determining that the specific block point cloud has a mobile attribute; if the detection result indicates that there is a specific block point cloud that is not a mobile target, determining that the specific block point cloud has a static attribute; based on the mobile attributes or static attributes corresponding to the block point clouds contained in the non-ground point cloud, segmenting the non-ground point cloud to obtain a static point cloud and / or a dynamic point cloud.

[0012] In combination with the first aspect, in certain implementations of the first aspect, detecting whether a block point cloud included in a non-ground point cloud is a moving target includes: determining a block point cloud with a moving attribute in the block point cloud included in the non-ground point cloud; clustering the block point cloud with a moving attribute, and calculating the mobility and / or size of the clustered point cloud; based on the mobility and / or size of the clustered point cloud, detecting whether the block point cloud included in the non-ground point cloud is a moving target; preferably, if the mobility of the clustered point cloud is greater than a target mobility threshold, and the size of the clustered point cloud is greater than a target size threshold, it indicates that the block point cloud included in the non-ground point cloud is a moving target; if the mobility of the clustered point cloud is less than or equal to the target size threshold, then the block point cloud included in the non-ground point cloud is a moving target. If the target mobility threshold is less than or equal to the target size threshold, and / or the size of the clustered point cloud is less than or equal to the target size threshold, it indicates that the block point cloud included in the non-ground point cloud is not a moving target; preferably, the block point cloud with mobile attributes is determined in the block point cloud included in the non-ground point cloud, including: if the surface feature corresponding to the block point cloud with mobile attributes is a non-planar feature, then when the height difference between the block point cloud and the ideal ground model is greater than a third height threshold, it is determined that the block point cloud has mobile attributes; if the surface feature corresponding to the block point cloud with mobile attributes is a roughness feature, then when the height difference between the block point cloud and the ideal ground model is greater than a fourth height threshold, it is determined that the block point cloud has mobile attributes.

[0013] In combination with the first aspect, in certain implementations of the first aspect, the surface features corresponding to each block point cloud are determined, including: using a point cloud processing model to determine the surface features corresponding to each block point cloud; or, using a principal component analysis method to determine the surface features corresponding to each block point cloud.

[0014] In the second aspect, an embodiment of the present application provides a point cloud segmentation device, including: a division module, used to divide the original point cloud to obtain multiple block point clouds; a first determination module, used to determine the surface features corresponding to each block point cloud, the surface features including at least one of plane features, non-plane features and roughness features; a second determination module, used to determine that each block point cloud has corresponding ground attributes or non-ground attributes based on the surface features corresponding to each block point cloud; a segmentation module, used to segment the original point cloud based on the ground attributes or non-ground attributes corresponding to each block point cloud, to obtain a ground point cloud and a non-ground point cloud.

[0015] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, which stores a computer program, and the computer program is used to execute the point cloud segmentation method described in the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides an electronic device, comprising: a processor; a memory for storing processor executable instructions; the processor is used to execute the point cloud segmentation method described in the first aspect.

[0017] In the present application, the original point cloud is first divided to obtain multiple block point clouds, so that each block point cloud can be processed independently, thereby reducing the overall computational complexity and improving the processing speed. Moreover, this block strategy can be flexibly adjusted for different scenarios. Next, the surface features of each block point cloud are determined, including plane features, non-plane features, and roughness features, which can more comprehensively describe the geometric characteristics of the block point cloud, thereby providing a basis for subsequent ground attribute or non-ground attribute determination without the need for complex parameter adjustments. Moreover, this determination method based on the surface features of the block point cloud is not easily affected by noise or small changes, so that it can still maintain a high accuracy when facing complex or dynamic scenes. Finally, segmenting the original point cloud based on ground attributes or non-ground attributes can clearly distinguish between ground point clouds and non-ground point clouds, thereby improving the quality and efficiency of map updates. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 The figure is a flow chart of a point cloud segmentation method provided in one embodiment of the present application.

[0020] Figure 2 The figure is a schematic diagram of a process for determining whether each segmented point cloud has corresponding ground attributes or non-ground attributes provided by an embodiment of the present application.

[0021] Figure 3 Shown is a flow chart of a point cloud segmentation method provided in another embodiment of the present application.

[0022] Figure 4 The figure is a schematic diagram of a process for detecting whether a block point cloud included in a non-ground point cloud is a moving target according to an embodiment of the present application.

[0023] Figure 5 Shown is a schematic diagram of the effect of point cloud segmentation provided by an embodiment of the present application.

[0024] Figure 6 Shown is a schematic diagram of the effect of point cloud segmentation provided by another embodiment of the present application.

[0025] Figure 7 Shown is a schematic diagram of the effect of point cloud segmentation provided by another embodiment of the present application.

[0026] Figure 8 Shown is a schematic structural diagram of a point cloud segmentation device provided in one embodiment of the present application.

[0027] Fig. 9 Shown is a schematic structural diagram of an electronic device provided by an exemplary embodiment of the present application. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] Figure 1 FIG. 2 is a flow chart of a point cloud segmentation method provided by an embodiment of the present application. Figure 1 As shown, the point cloud segmentation method includes the following steps.

[0030] Step S110, dividing the original point cloud to obtain a plurality of block point clouds.

[0031] Exemplarily, the original point cloud is a dense point cloud. Specifically, a dense point cloud is a data set consisting of a large number of points obtained by a 3D scanning device, such as a laser scanner, a structured light scanner, or a depth camera, and each point contains additional information such as its coordinates, color, intensity, or normal in space. Compared with a sparse point cloud, the points in a dense point cloud are densely distributed in space with small intervals, and can more accurately represent the shape and surface features of an object.

[0032] In some embodiments, the original point cloud can be divided using a rasterization method, an octree method, etc. Taking the rasterization method as an example, first a target point is set in three-dimensional space. The target point is a reference point in the rasterization method, which is used to convert all points of the original point cloud into a coordinate system with the target point as the coordinate origin, so as to simplify the subsequent processing of the original point cloud. Then the coordinates of all points in the original point cloud are subtracted from the coordinates of the origin. Next, the row and column height position of each point in the original point cloud in the new coordinate system is calculated according to the set resolution parameters, thereby realizing the preliminary block division of the dense point cloud. Among them, resolution is a key parameter in the rasterization process, which determines the size of each small grid, and thus affects the accuracy and efficiency of the block division. Through such block processing, the original point cloud can be effectively divided into multiple small blocks, laying the foundation for subsequent point cloud segmentation and feature extraction.

[0033] Step S120, determining the surface features corresponding to each block point cloud.

[0034] The surface features include at least one of planar features, non-planar features, and roughness features.

[0035] In some embodiments, a point cloud processing model is used to determine the surface features corresponding to each block point cloud.

[0036] Optionally, the point cloud processing model is a coarse segmentation model, which is used to perform preliminary segmentation processing on dense point clouds, and its core function is to identify the surface features of block point clouds. Specifically, in the point cloud processing model, feature extraction is performed on each block point cloud, including calculating the center point, normal vector, density and thickness of the block point cloud. Through these features, the point cloud processing model can preliminarily determine the surface features of each block point cloud. For example, if the normal vector of a block point cloud is close to vertical and the thickness and elevation difference are small, it will be identified as a plane feature; if the surface of the block point cloud changes greatly, it will be identified as a non-planar feature; if the thickness of the block point cloud exceeds a certain threshold, it will be identified as a roughness feature.

[0037] The point cloud processing model in this embodiment provides a fast and efficient way to process dense point clouds, laying the foundation for subsequent more detailed analysis. In addition, it does not rely on complex calculations, but instead uses some basic geometric and statistical methods to achieve rough segmentation of block point clouds. In this way, it provides fast and accurate preliminary segmentation results for some applications.

[0038] In other embodiments, the principal component analysis method is used to determine the surface features corresponding to each block point cloud.

[0039] Optionally, principal component analysis is used to analyze multiple block point clouds corresponding to the original point cloud to obtain the first eigenvalue, second eigenvalue, third eigenvalue and third eigenvector of the point cloud covariance, where the third eigenvector is the normal vector of the block point cloud. Next, the block point cloud is converted from the lidar coordinate system to the inertial navigation unit coordinate system, and then converted to the projection coordinate system. The coordinate difference is obtained by subtracting the coordinate of the center point of the block point cloud from the coordinate of each point in the block point cloud in the projection coordinate system. The coordinate difference is multiplied by the normal vector of the block point cloud to obtain the offset of each point relative to the center point of the block point cloud in the normal direction. The offsets corresponding to all points are added and then divided by the number of points in the block point cloud to obtain the average offset of the points in the block point cloud and the block point cloud. The average offset is used as the thickness of the block point cloud. Exemplarily, the calculation formula for the thickness of the block point cloud is: Among them, P′ represents the coordinates of the center point of the block point cloud, P liRepresents the coordinates of the points in the block point cloud in the laser radar coordinate system, Represents the coordinate transformation parameters from the laser radar coordinate system to the inertial navigation unit coordinate system, Represents the coordinate transformation parameters from the inertial navigation unit coordinate system to the projection coordinate system, represents the normal vector of the block point cloud, and m represents the number of points in the block point cloud. Optionally, if the thickness of the block point cloud is greater than a target thickness threshold, the surface feature corresponding to the block point cloud is determined to be a roughness feature.

[0040] Further, the covariance feature ratio of the block point cloud is calculated. Optionally, the covariance feature ratio = first eigenvalue / third eigenvalue. If the covariance feature ratio is greater than the target covariance feature ratio threshold, the block point cloud is determined to be a surface. Further, if the angle between any point in the block point cloud and the normal vector of the block point cloud is less than the target angle threshold, the thickness of the block point cloud is less than the target thickness threshold, and the offset of the point in the block point cloud in the normal direction relative to the center point of the block point cloud is less than the target offset threshold, then the block point cloud is determined to correspond to a plane feature, otherwise, the block point cloud is determined to correspond to a non-planar feature, such as, the block point cloud is a slope.

[0041] In this embodiment, principal component analysis is used to determine the surface features corresponding to the block point cloud, which can reduce the dependence on a large number of parameters. In addition, the block point cloud processed by principal component analysis can significantly reduce the dimension of the data, simplify the subsequent processing flow, and improve the calculation efficiency.

[0042] Step S130 : determining that each block point cloud has a corresponding ground attribute or non-ground attribute based on the surface features corresponding to each block point cloud.

[0043] In an exemplary embodiment, for a block point cloud with plane features, a series of rules are set using the statistical characteristics of this block point cloud, such as height, density, and distribution range. For example, if the height of the block point cloud is within a preset height range and the density of the block point cloud is lower than a preset density threshold, the block point cloud is considered to have ground attributes, otherwise, the block point cloud has non-ground attributes.

[0044] In another example, a deep learning model, such as a convolutional neural network or a graph neural network, is used to identify the semantic information of a block point cloud with planar features, and then determine the ground attributes or non-ground attributes of this block point cloud.

[0045] Step S140 , based on the ground attributes or non-ground attributes corresponding to each block point cloud, the original point cloud is segmented to obtain a ground point cloud and a non-ground point cloud.

[0046] Exemplarily, all the block point clouds are traversed, and each block point cloud is classified into corresponding categories according to its attribute label (including ground attributes and non-ground attributes). Specifically, the block point clouds belonging to ground attributes will be aggregated into the ground point cloud set, while the block point clouds belonging to non-ground attributes will be aggregated into the non-ground point cloud set. In this way, the original point cloud is divided into two parts: one is the point cloud data representing the ground, and the other is the point cloud data of non-ground, such as buildings, vehicles, vegetation, etc.

[0047] In the present application, the original point cloud is first divided to obtain multiple block point clouds, so that each block point cloud can be processed independently, thereby reducing the overall computational complexity and improving the processing speed. Moreover, this block strategy can be flexibly adjusted for different scenarios. Next, the surface features of each block point cloud are determined, including plane features, non-plane features, and roughness features, which can more comprehensively describe the geometric characteristics of the block point cloud, thereby providing a basis for subsequent ground attribute or non-ground attribute determination without the need for complex parameter adjustments. Moreover, this determination method based on the surface features of the block point cloud is not easily affected by noise or small changes, so that it can still maintain a high accuracy when facing complex or dynamic scenes. Finally, segmenting the original point cloud based on ground attributes or non-ground attributes can clearly distinguish between ground point clouds and non-ground point clouds, thereby improving the quality and efficiency of map updates.

[0048] Figure 2 The figure shows a flow chart of determining whether each block point cloud has corresponding ground attributes or non-ground attributes provided by an embodiment of the present application. Figure 1 Based on the embodiment shown, Figure 2 The embodiment shown is described below in detail. Figure 2 The embodiment shown and Figure 1 The differences and similarities of the illustrated embodiments are not described in detail.

[0049] like Figure 2 As shown, in this embodiment, based on the surface features corresponding to each block point cloud, determining that each block point cloud has corresponding ground attributes or non-ground attributes includes the following steps.

[0050] Step S210 , for a first block point cloud having a plane feature among the plurality of block point clouds, it is determined whether the first block point cloud satisfies a ground determination condition.

[0051] Exemplarily, if the execution result of step S210 is that the first block point cloud satisfies the ground judgment condition, then step S220 is executed: determining that the first block point cloud has ground attributes; if the execution result of step S210 is that the first block point cloud does not meet the ground judgment condition, then step S230 is executed: determining that the first block point cloud has non-ground attributes.

[0052] Specifically, the ground determination condition refers to the standard or rule used to determine whether a block point cloud belongs to the ground during point cloud processing and analysis. The ground determination condition is usually defined based on the geometric features, statistical characteristics and environmental context information of the point cloud.

[0053] In some embodiments, if the angle between the normal vector of the first block point cloud and the vertical direction is within a specific small range, it is determined that the first block point cloud meets the ground determination condition; otherwise, it is determined that the first block point cloud does not meet the ground determination condition.

[0054] In some embodiments, an initial ground seed point is determined, and a first block point cloud within the neighborhood of the initial ground seed point is searched; if the angle between the normal vector of the first block point cloud within the neighborhood of the initial ground seed point and the normal vector of the initial ground seed point is less than a target angle threshold, and the height difference between the first block point cloud within the neighborhood of the initial ground seed point and the initial ground seed point is less than a first height threshold, then it is determined that the first block point cloud within the neighborhood of the initial ground seed point meets the ground judgment condition; the first block point cloud within the neighborhood of the initial ground seed point is determined as the target ground seed point, and with the target ground seed point as a reference, the steps of searching, judging, and updating the target ground seed point are repeated until all first block point clouds are traversed.

[0055] The initial ground seed point is a point that is assumed to be on the ground based on prior knowledge or a preprocessing step. Exemplarily, the initial ground seed point is determined based on the height of the point cloud acquisition device above the ground. Specifically, point cloud acquisition devices (such as lidar, stereo cameras, etc.) are usually able to provide their own precise position and posture information when collecting data, including height above the ground. This height information is an absolute height relative to a known ground. Furthermore, based on the height of the acquisition device above the ground, points whose height is equal to or close to the height above the ground can be selected as initial ground seed points. For example, if a lidar is installed on an autonomous driving car with a height of 1 meter, then points about 1 meter away from the point cloud acquisition device can be selected as initial ground seed points.

[0056] For the first block point cloud in the neighborhood of each initial ground seed point, compare the angle between the normal vector of this block point cloud and the normal vector of the initial ground seed point. If this angle is less than the target angle threshold, it indicates that the plane where the first block point cloud and the initial ground seed point are located is relatively parallel in space, which is a common feature of ground point clouds. Further, compare the height difference between the first block point cloud and the initial ground seed point. If the height difference is less than the set first height threshold, it indicates that the block point cloud is close to the initial ground seed point in the vertical direction, which is another feature of ground point clouds.

[0057] If the first block point cloud satisfies both of the above conditions, then the first block point cloud will be judged to meet the ground judgment condition and will be determined as the target ground seed point. This process will be repeated based on this new target ground seed point, and other block point clouds in the neighborhood of this new target ground seed point will continue to be searched, and the same judgment logic will be applied. This iterative process will continue until all the first block point clouds have been traversed and judged.

[0058] This embodiment searches for the first block point cloud in the neighborhood of the initial ground seed point, and determines whether these point clouds meet the ground determination conditions based on the two conditions of the angle of the normal vector and the height difference, thereby ensuring that only those first block point clouds that match the ground characteristics are selected as ground point clouds. The advantage of this determination method is that it can use the geometric characteristics of the first block point cloud to distinguish between ground attributes and non-ground attributes, rather than relying solely on height information, which makes it more adaptable to different terrains and scenes. By re-determining the target ground seed point and repeating the search and determination, the recognition range of the ground point cloud can be gradually expanded, and the spatial continuity of the ground point cloud can be maintained through continuous processing of the local neighborhood.

[0059] Step S240: Generate an ideal ground model based on the first block point cloud with ground attributes.

[0060] An ideal ground model refers to a theoretically perfect ground point cloud model that removes noise and non-ground features. This model represents the true geometric shape of the ground without any protrusions or depressions and is usually used as a reference or benchmark.

[0061] In some embodiments, based on all first block point clouds with ground attributes, a plane equation is fitted using the least square method to best represent the geometric shape of the ground, thereby obtaining an ideal ground model. Optionally, the generation process of the ideal ground model also includes steps such as smoothing and filling holes to ensure the continuity and integrity of the ideal ground model.

[0062] In some embodiments, the resolution of the first block point cloud with ground attributes is expanded to obtain a high-resolution block point cloud; with any high-resolution block point cloud as the center, the high-resolution block point clouds within the central target range are aggregated, and the aggregated point cloud data are used to fit the target ground equation; the target ground equation is determined as an ideal ground model.

[0063] Exemplarily, the number of points in the first block point cloud with ground attributes is increased by interpolation or data resampling to obtain a high-resolution block point cloud, which can more accurately capture the geometric details and surface features of the measured object or scene. Then, taking any one of these high-resolution block point clouds as the center point, search for other high-resolution block point clouds within the target range around the center point, and aggregate these high-resolution block point clouds together. Then, use the aggregated point cloud data to fit an optimal target ground equation. This equation can be a plane equation or a more complex surface equation, depending on the actual shape of the ground. Exemplarily, the fitting process involves least squares or other optimization algorithms to ensure that the fitted target ground equation can best represent the aggregated point cloud data.

[0064] In this embodiment, by increasing the resolution of the first block point cloud with ground attributes, more detailed ground geometry information can be obtained, thereby improving the accuracy of the target ground model. Then, by locally aggregating the high-resolution block point cloud, a more complete and continuous target ground model can be constructed in the local area. At the same time, the adaptability of the target ground model to local ground changes is enhanced, so that the target ground model can better adapt to changes in different terrains.

[0065] Step S250 , for a second block point cloud having non-planar features and / or roughness features in the plurality of block point clouds, based on an ideal ground model, determining that the second block point cloud has corresponding ground attributes or non-ground attributes.

[0066] Specifically, for those second block point clouds having non-planar features and / or roughness features, the ideal ground model is used as a reference benchmark to further determine their ground attributes or non-ground attributes.

[0067] In some embodiments, the height difference between the second block point cloud and the ideal ground model is calculated; if the height difference between the second block point cloud and the ideal ground model is less than a second height threshold, it is determined that the second block point cloud has ground attributes; if the height difference between the second block point cloud and the ideal ground model is greater than or equal to the second height threshold, it is determined that the second block point cloud has non-ground attributes.

[0068] The height difference is determined by comparing the height of the points in the second block point cloud with the height of the corresponding position of the ideal ground model, reflecting the vertical offset of the points in the second block point cloud relative to the ideal ground. The second height threshold is a predefined critical value used to determine whether the second block point cloud belongs to the ground. If the height difference between a second block point cloud and the ideal ground model is less than the second height threshold, it means that the position of the second block point cloud is lower than the ground, and it may be just a small undulation or roughness of the ground. Therefore, the second ground point cloud is judged to have ground attributes. On the contrary, if the height difference is greater than or equal to the second height threshold, it means that the second block point cloud is significantly higher than the ideal ground model and is likely to be a non-ground feature, such as a building, vehicle or vegetation. In this case, the second block point cloud is judged to have non-ground attributes.

[0069] Through the method in this embodiment, the ground attribute or non-ground attribute of the second block point cloud is quickly determined, the efficiency of point cloud classification is improved, and the ideal ground model is used as a reference benchmark to enhance the adaptability to complex terrain and scenes.

[0070] In summary, Figure 2 In the illustrated embodiment, the surface features of each first segment point cloud are used to simplify the classification process. The ideal ground model generated based on the first segment point cloud that has been determined to have ground attributes provides a stable and reliable reference, which in turn serves as a benchmark for judging whether other second segment point clouds have ground attributes, making the classification of segment point clouds with non-planar features or roughness features more accurate.

[0071] Figure 3 FIG. 1 is a flow chart of a point cloud segmentation method provided by another embodiment of the present application. Figure 1 Based on the embodiment shown, Figure 3 The embodiment shown is described below in detail. Figure 3 The embodiment shown and Figure 1 The differences and similarities of the illustrated embodiments are not described in detail.

[0072] like Figure 3 As shown, in this embodiment, the point cloud segmentation method also includes the following steps.

[0073] Step S310, determining the block point clouds included in the non-ground point cloud.

[0074] In the early stage of raw point cloud processing, the entire raw point cloud is divided into multiple block point clouds, each of which records its corresponding data and feature information. In the subsequent steps, the raw point cloud is segmented to separate the ground point cloud and the non-ground point cloud, and the non-ground point cloud still retains the information of which original block they originated from. Therefore, in this embodiment, this recorded information is used to trace the block point clouds contained in the non-ground point cloud.

[0075] Step S320: Detect whether the block point cloud included in the non-ground point cloud is a moving target.

[0076] Moving targets are those point clouds whose positions change within a certain time interval. For example, in the application scenario of autonomous vehicles, moving targets can be other vehicles, pedestrians, or any objects moving on the road.

[0077] In some embodiments, a point cloud acquisition device is used to collect multiple frames of point cloud data at different time points, and the position changes of each point in the block point cloud are tracked to detect whether the block point cloud contained in the non-ground point cloud is a moving target. Specifically, the position of each block point cloud in two adjacent frames of data is determined, and then the position change of these block point clouds in time is calculated. If the position of a block point cloud changes significantly relative to a reference point or reference plane during a continuous data acquisition cycle, it indicates that the block point cloud is a moving target.

[0078] In some other embodiments, the average speed and acceleration of the block point cloud between two adjacent frames of data may also be calculated. If these parameters exceed a preset threshold, the block point cloud may be determined as a moving target.

[0079] Exemplarily, if the execution result of step S320 is that there is a specific block point cloud that is a mobile target, then step S330 is executed: determining that the specific block point cloud has a mobile attribute; if the execution result of step S330 is that there is a specific block point cloud that is not a mobile target, then step S340 is executed: determining that the specific block point cloud has a static attribute.

[0080] In some embodiments, determining whether a specific block point cloud has a mobile attribute or a static attribute is, in fact, to assign attribute labels to the block point cloud. These attribute labels are used to describe and classify the behavioral characteristics of the block point cloud in a dynamic environment.

[0081] Step S350 , segmenting the non-ground point cloud based on the mobile attributes or static attributes corresponding to the block point clouds contained in the non-ground point cloud to obtain a static point cloud and / or a dynamic point cloud.

[0082] Exemplarily, the block point clouds contained in the non-ground point cloud are traversed and classified according to their respective attribute labels. The block point clouds belonging to dynamic attributes are aggregated to form dynamic point clouds, which represent the moving parts of the environment, such as moving vehicles, walking pedestrians or flying objects. The block point clouds belonging to static attributes are aggregated into static point clouds, which represent the fixed parts of the environment, such as buildings, trees or road signs.

[0083] In this embodiment, it is detected whether the segmented point clouds are mobile targets, and their mobile or static properties are further determined. This attribute-based segmentation method achieves accurate segmentation of non-ground point clouds, reduces the amount of data that needs to be processed simultaneously, and improves the accuracy of data processing.

[0084] Figure 4 The figure shows a flow chart of detecting whether a block point cloud included in a non-ground point cloud is a moving target according to an embodiment of the present application. Figure 3 Based on the embodiment shown, Figure 4 The embodiment shown is described below in detail. Figure 4 The embodiment shown and Figure 3 The differences and similarities of the illustrated embodiments are not described in detail.

[0085] like Figure 4 As shown, in this embodiment, detecting whether the block point cloud included in the non-ground point cloud is a moving target includes:

[0086] Step S410 , determining a block point cloud having a moving attribute in the block point clouds included in the non-ground point cloud.

[0087] In some embodiments, if the surface feature corresponding to the block point cloud with a mobile attribute is a non-planar feature, then when the height difference between the block point cloud and the ideal ground model is greater than a third height threshold, the block point cloud is determined to have a mobile attribute; if the surface feature corresponding to the block point cloud with a mobile attribute is a roughness feature, then when the height difference between the block point cloud and the ideal ground model is greater than a fourth height threshold, the block point cloud is determined to have a mobile attribute.

[0088] It is understood that if the height difference between the block point cloud of the non-planar feature and the ideal ground model is greater than the third height threshold, it indicates that these block point clouds are significantly higher than the ground in the vertical direction, so such block point clouds represent moving targets, such as moving vehicles or moving objects. Exemplarily, the third height threshold is based on a statistical or empirical estimate of the typical height range of the moving target of the non-planar feature.

[0089] Correspondingly, for the block point clouds of roughness features, if their height difference with the ideal ground model is greater than another specific fourth height threshold, it indicates that these block point clouds are also significantly higher than the ground in the vertical direction, but because their surface roughness is high, this means that they are protrusions or other irregularly shaped objects on the ground, which can also be regarded as moving targets if they move. Exemplarily, the fourth height threshold is based on a statistical or empirical estimate of the typical height range of moving targets of roughness features.

[0090] This embodiment combines the geometric features (surface features) and spatial position information (height difference) of the block point cloud, and provides a simple and effective way to identify the moving target through threshold processing. In addition, by setting these two different height thresholds, the determination process of the moving target can take into account the different performances of the block point clouds with different types of surface features in the vertical direction, thereby more accurately distinguishing the real moving target.

[0091] Step S420 , clustering the block point clouds with the mobile attribute, and calculating the mobility and / or size of the clustered point clouds.

[0092] Clustering is performed on block point clouds with mobile attributes. The purpose is to combine these block point clouds with mobile attributes into larger sets according to their spatial proximity or other related features. This can help identify point cloud sets that are close to each other in space and may belong to the same moving object or structure.

[0093] The mobility ratio is a measure of the proportion of segmented point clouds with mobility attributes in a clustered point cloud. Optionally, the mobility ratio can be obtained by dividing the number of segmented point clouds with mobility attributes in a clustered point cloud by the total number of all segmented point clouds in the clustered point cloud. The mobility ratio provides a quantitative indicator that can give a general understanding of the dynamic characteristics of the clustered point cloud. For example, if a clustered point cloud consists of 10 segmented point clouds, 7 of which are identified as having mobility attributes, then the mobility ratio of the target will be 7 / 10 or 0.7. This mobility ratio can help determine whether the clustered point cloud is mainly composed of moving objects or whether it contains a large number of stationary objects.

[0094] Step S430 , based on the mobility and / or size of the clustered point cloud, detecting whether the block point cloud included in the non-ground point cloud is a moving target.

[0095] In some embodiments, if the mobility rate of the clustered point cloud is greater than the target mobility rate threshold, and the size of the clustered point cloud is greater than the target size threshold, it indicates that the block point cloud contained in the non-ground point cloud is a moving target; if the mobility rate of the clustered point cloud is less than or equal to the target mobility rate threshold, and / or the size of the clustered point cloud is less than or equal to the target size threshold, it indicates that the block point cloud contained in the non-ground point cloud is not a moving target.

[0096] The size reflects the spatial range or volume of the clustered point cloud. The size threshold setting helps to exclude those small and meaningless moving block point clouds. Specifically, if the size of the clustered point cloud exceeds the target size threshold, it indicates that the clustered point cloud has a certain scale in space and is more likely to be a complete moving target rather than a scattered and insignificant moving block point cloud.

[0097] When the clustered point cloud satisfies the conditions of mobility and size at the same time, it can be determined that the block point cloud contained in the non-ground point cloud is a moving target. This embodiment provides a more reliable means of identifying moving targets by combining the dynamic features and geometric information of the non-ground point cloud in a quantitative manner. This method can improve the adaptability and responsiveness to the environment and is particularly suitable for point cloud data processing in a dynamic environment.

[0098] Figure 5 The figure shows the effect of point cloud segmentation provided by an embodiment of the present application. Figure 5 In , only the ground attributes and non-ground attributes are judged and segmented. Specifically, Figure 5 As shown, the blue point cloud is the retaining wall, and the red point cloud is the ground, that is, the blue point cloud is the non-ground point cloud, and the red point cloud is the ground point cloud.

[0099] Figure 6 The figure shows the effect of point cloud segmentation provided by another embodiment of the present application. Figure 5 Based on the embodiment shown, Figure 6 The static properties of the point cloud are judged, specifically, Figure 6 As shown, the red point cloud is the ground point cloud, the blue point cloud is the retaining wall with static attributes, and the black point cloud is the trajectory point cloud.

[0100] Figure 7 The figure shows the effect of point cloud segmentation provided by another embodiment of the present application. Figure 6 Based on the embodiment shown, Figure 7 The dynamic properties of the point cloud are judged, specifically, Figure 7 As shown, the red point cloud is the moving target, the green point cloud is the ground point cloud, and the blue point cloud is the retaining wall with static attributes.

[0101] Combination of the above Figures 1 to 7, describes in detail the point cloud segmentation method embodiment of the present application, and the following is combined with Figure 8 , the point cloud segmentation device embodiment of the present application is described in detail. It should be understood that the description of the point cloud segmentation method embodiment corresponds to the description of the point cloud segmentation device embodiment, so the part not described in detail can refer to the previous method embodiment.

[0102] Figure 8 FIG. 1 is a schematic diagram of the structure of a point cloud segmentation device provided in one embodiment of the present application. Figure 8 As shown, the point cloud segmentation device 80 provided in the embodiment of the present application includes:

[0103] A division module 810 is used to divide the original point cloud into multiple block point clouds;

[0104] A first determination module 820 is used to determine the surface features corresponding to each block point cloud, where the surface features include at least one of a plane feature, a non-plane feature and a roughness feature;

[0105] A second determination module 830 is used to determine that each block point cloud has a corresponding ground attribute or non-ground attribute based on the surface features corresponding to each block point cloud;

[0106] The segmentation module 840 is used to segment the original point cloud based on the ground attribute or non-ground attribute corresponding to each block point cloud to obtain a ground point cloud and a non-ground point cloud.

[0107] In one embodiment of the present application, the second determination module 830 is also used to, for a first block point cloud having planar features among multiple block point clouds, if the first block point cloud meets the ground judgment condition, determine that the first block point cloud has ground attributes, otherwise, determine that the first block point cloud has non-ground attributes; based on the first block point cloud with ground attributes, generate an ideal ground model; for a second block point cloud having non-planar features and / or roughness features among multiple block point clouds, based on the ideal ground model, determine that the second block point cloud has corresponding ground attributes or non-ground attributes.

[0108] In one embodiment of the present application, the second determination module 830 is also used to determine the initial ground seed point and search for the first block point cloud within the neighborhood of the initial ground seed point; if the angle between the normal vector of the first block point cloud within the neighborhood of the initial ground seed point and the normal vector of the initial ground seed point is less than the target angle threshold, and the height difference between the first block point cloud within the neighborhood of the initial ground seed point and the initial ground seed point is less than the first height threshold, then it is determined that the first block point cloud within the neighborhood of the initial ground seed point meets the ground judgment condition; the first block point cloud within the neighborhood of the initial ground seed point is determined as the target ground seed point, and with the target ground seed point as the reference, the search, judgment and update steps of the target ground seed point are repeated until all the first block point clouds are traversed; preferably, the initial ground seed point is determined based on the height of the point cloud acquisition device above the ground.

[0109] In one embodiment of the present application, the second determination module 830 is also used to expand the resolution of the first block point cloud with ground attributes to obtain a high-resolution block point cloud; taking any high-resolution block point cloud as the center, aggregating the high-resolution block point clouds within the central target range, and using the aggregated point cloud data to fit the target ground equation; and determining the target ground equation as an ideal ground model.

[0110] In one embodiment of the present application, the second determination module 830 is also used to calculate the height difference between the second block point cloud and the ideal ground model; if the height difference between the second block point cloud and the ideal ground model is less than the second height threshold, it is determined that the second block point cloud has ground attributes; if the height difference between the second block point cloud and the ideal ground model is greater than or equal to the second height threshold, it is determined that the second block point cloud has non-ground attributes.

[0111] In one embodiment of the present application, the segmentation module 840 is also used to determine the block point clouds contained in the non-ground point cloud; detect whether the block point clouds contained in the non-ground point cloud are mobile targets; if the detection result indicates that there is a specific block point cloud that is a mobile target, determine that the specific block point cloud has a mobile attribute; if the detection result indicates that there is a specific block point cloud that is not a mobile target, determine that the specific block point cloud has a static attribute; based on the mobile attributes or static attributes corresponding to the block point clouds contained in the non-ground point cloud, segment the non-ground point cloud to obtain a static point cloud and / or a dynamic point cloud.

[0112] In one embodiment of the present application, the segmentation module 840 is further used to determine the block point clouds with mobile attributes in the block point clouds included in the non-ground point cloud; cluster the block point clouds with mobile attributes, and calculate the mobility and / or size of the clustered point clouds; based on the mobility and / or size of the clustered point clouds, detect whether the block point clouds included in the non-ground point cloud are mobile targets; preferably, if the mobility of the clustered point cloud is greater than the target mobility threshold, and the size of the clustered point cloud is greater than the target size threshold, it indicates that the block point cloud included in the non-ground point cloud is a mobile target; if the mobility of the clustered point cloud is less than or equal to the target mobility threshold, and / or the clustered point cloud is smaller than the target mobility threshold, then the block point cloud is a mobile target. If the size of the classified point cloud is less than or equal to the target size threshold, it means that the block point cloud included in the non-ground point cloud is not a moving target; preferably, the block point cloud with mobile attributes is determined in the block point cloud included in the non-ground point cloud, including: if the surface feature corresponding to the block point cloud with mobile attributes is a non-planar feature, then when the height difference between the block point cloud and the ideal ground model is greater than the third height threshold, the block point cloud is determined to have a mobile attribute; if the surface feature corresponding to the block point cloud with mobile attributes is a roughness feature, then when the height difference between the block point cloud and the ideal ground model is greater than the fourth height threshold, the block point cloud is determined to have a mobile attribute.

[0113] In one embodiment of the present application, the first determination module 810 is further used to determine the surface features corresponding to each block point cloud using a point cloud processing model; or to determine the surface features corresponding to each block point cloud using a principal component analysis method.

[0114] Below, reference Fig. 9 To describe an electronic device according to an embodiment of the present application. Fig. 9 Shown is a schematic structural diagram of an electronic device provided by an exemplary embodiment of the present application.

[0115] like Fig. 9 As shown, the electronic device 990 includes one or more processors 901 and a memory 902 .

[0116] The processor 901 may be a central processing unit (CPU) or other forms of processing units having data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device 90 to perform desired functions.

[0117] The memory 902 may include one or more computer program products, and the computer program product may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory (cache), etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 901 may run the program instructions to implement the point cloud segmentation method of each embodiment of the present application described above and / or other desired functions. Various contents such as original point clouds, block point clouds, surface features corresponding to block point clouds, ground point clouds, and non-ground point clouds may also be stored in the computer-readable storage medium.

[0118] In one example, the electronic device 90 may further include: an input device 903 and an output device 904 , and these components are interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0119] The input device 903 may include, for example, a keyboard, a mouse, etc.

[0120] The output device 904 can output various information to the outside, including original point clouds, block point clouds, surface features corresponding to the block point clouds, ground point clouds and non-ground points, etc. The output device 904 can include, for example, a display, a speaker, a printer, a communication network and a remote output device connected thereto, etc.

[0121] Of course, to simplify, Fig. 9 Only some of the components related to the present application in the electronic device 90 are shown, and components such as a bus, an input / output interface, etc. are omitted. In addition, according to specific application situations, the electronic device 90 may also include any other appropriate components.

[0122] In addition to the above-mentioned methods and devices, an embodiment of the present application may also be a computer program product, which includes computer program instructions, which, when executed by a processor, enable the processor to execute the steps of the point cloud segmentation method according to various embodiments of the present application described above in this specification.

[0123] The computer program product may be written in any combination of one or more programming languages ​​to write program codes for performing the operations of the embodiments of the present application, including object-oriented programming languages, such as Java, C++, etc., and conventional procedural programming languages, such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as an independent software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0124] In addition, an embodiment of the present application may also be a computer-readable storage medium on which computer program instructions are stored. When the computer program instructions are executed by a processor, the processor executes the steps of the point cloud segmentation method according to various embodiments of the present application described above in this specification.

[0125] The computer readable storage medium can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can include, for example, but is not limited to, a system, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination of the above. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0126] The basic principles of the present application are described above in conjunction with specific embodiments. However, it should be noted that the advantages, strengths, effects, etc. mentioned in the present application are only examples and not limitations, and it cannot be considered that these advantages, strengths, effects, etc. are required by each embodiment of the present application. In addition, the specific details disclosed above are only for the purpose of illustration and ease of understanding, not for limitation, and the above details do not limit the present application to being implemented by adopting the above specific details.

[0127] The block diagrams of the devices, apparatuses, equipment, and systems involved in this application are only illustrative examples and are not intended to require or imply that they must be connected, arranged, and configured in the manner shown in the block diagram. As will be appreciated by those skilled in the art, these devices, apparatuses, equipment, and systems can be connected, arranged, and configured in any manner. Words such as "including", "comprising", "having", etc. are open words, referring to "including but not limited to", and can be used interchangeably with them. The words "or" and "and" used here refer to the words "and / or" and can be used interchangeably with them, unless the context clearly indicates otherwise. The words "such as" used here refer to the phrase "such as but not limited to", and can be used interchangeably with them.

[0128] It should also be noted that in the apparatus, device and method of the present application, each component or each step can be decomposed and / or recombined. Such decomposition and / or recombination should be regarded as equivalent solutions of the present application.

[0129] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the present application. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the aspects shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

[0130] The above description has been given for the purpose of illustration and description. In addition, this description is not intended to limit the embodiments of the present application to the forms disclosed herein. Although multiple example aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, changes, additions and sub-combinations thereof.

Claims

1. A point cloud segmentation method, characterized in that it includes: Divide the original point cloud to obtain multiple block point clouds; Determine a surface feature corresponding to each of the block point clouds, wherein the surface feature includes at least one of a plane feature, a non-plane feature, and a roughness feature; Based on the surface features corresponding to each of the block point clouds, determining that each of the block point clouds has corresponding ground attributes or non-ground attributes; Based on the ground attributes or non-ground attributes corresponding to each of the block point clouds, the original point cloud is segmented to obtain a ground point cloud and a non-ground point cloud.

2. The point cloud segmentation method according to claim 1, characterized in that: The determining that each of the block point clouds has a corresponding ground attribute or non-ground attribute based on the surface features corresponding to each of the block point clouds comprises: For a first block point cloud having the plane feature among the multiple block point clouds, if the first block point cloud satisfies the ground determination condition, determining that the first block point cloud has the ground attribute; otherwise, determining that the first block point cloud has the non-ground attribute; Generate an ideal ground model based on the first block point cloud having the ground attributes; For a second block point cloud having the non-planar feature and / or the roughness feature in the multiple block point clouds, based on the ideal ground model, it is determined that the second block point cloud has the corresponding ground attribute or the non-ground attribute.

3. The point cloud segmentation method according to claim 2, characterized in that: Also includes: Determine an initial ground seed point, and search for a first block point cloud within the neighborhood of the initial ground seed point; If the angle between the normal vector of the first block point cloud in the neighborhood of the initial ground seed point and the normal vector of the initial ground seed point is less than the target angle threshold, and the height difference between the first block point cloud in the neighborhood of the initial ground seed point and the initial ground seed point is less than the first height threshold, it is determined that the first block point cloud in the neighborhood of the initial ground seed point meets the ground judgment condition; The first block point cloud in the neighborhood of the initial ground seed point is determined as the target ground seed point, and the steps of searching, determining and updating the target ground seed point are repeated with the target ground seed point as a reference until all the first block point clouds are traversed; Preferably, the initial ground seed point is determined based on the height of the point cloud acquisition device above the ground.

4. The point cloud segmentation method according to claim 2, characterized in that: The step of generating an ideal ground model based on the first block point cloud having the ground attributes comprises: Enlarging the resolution of the first block point cloud having the ground attributes to obtain a high-resolution block point cloud; Taking any high-resolution block point cloud as the center, aggregate the high-resolution block point clouds within the target range of the center, and use the aggregated point cloud data to fit the target ground equation; The target ground equation is determined as the ideal ground model.

5. The point cloud segmentation method according to claim 2, characterized in that: The determining, based on the ideal ground model, that the second block point cloud has the corresponding ground attribute or the non-ground attribute includes: Calculate the height difference between the second block point cloud and the ideal ground model; If the height difference between the second block point cloud and the ideal ground model is less than a second height threshold, determining that the second block point cloud has the ground attribute; If the height difference between the second block point cloud and the ideal ground model is greater than or equal to the second height threshold, it is determined that the second block point cloud has the non-ground attribute.

6. The point cloud segmentation method according to any one of claims 1 to 5, characterized in that: Also includes: Determine the block point cloud included in the non-ground point cloud; Detecting whether the block point cloud included in the non-ground point cloud is a moving target; If the detection result indicates that there is a specific block point cloud that is the moving target, then determining that the specific block point cloud has a moving attribute; If the detection result indicates that there is a specific block point cloud that is not the moving target, determining that the specific block point cloud has a static attribute; Based on the mobile attributes or static attributes corresponding to the block point clouds contained in the non-ground point cloud, the non-ground point cloud is segmented to obtain a static point cloud and / or a dynamic point cloud.

7. The point cloud segmentation method according to claim 6, characterized in that: The detecting whether the block point cloud included in the non-ground point cloud is a moving target includes: Determining a block point cloud having a moving attribute among the block point clouds included in the non-ground point cloud; Clustering the block point clouds with the movement attribute, and calculating the movement rate and / or size of the clustered point clouds; Based on the mobility and / or size of the clustered point cloud, detecting whether the block point cloud included in the non-ground point cloud is the moving target; Preferably, if the mobility of the clustered point cloud is greater than a target mobility threshold, and the size of the clustered point cloud is greater than a target size threshold, it indicates that the block point cloud included in the non-ground point cloud is the moving target; if the mobility of the clustered point cloud is less than or equal to the target mobility threshold, and / or the size of the clustered point cloud is less than or equal to the target size threshold, it indicates that the block point cloud included in the non-ground point cloud is not the moving target; Preferably, determining the block point cloud having the mobile attribute in the block point cloud included in the non-ground point cloud comprises: If the surface feature corresponding to the block point cloud having the movement attribute is the non-planar feature, then when the height difference between the block point cloud and the ideal ground model is greater than a third height threshold, it is determined that the block point cloud has the movement attribute; If the surface feature corresponding to the block point cloud with the movement attribute is the roughness feature, then when the height difference between the block point cloud and the ideal ground model is greater than a fourth height threshold, it is determined that the block point cloud has the movement attribute.

8. The point cloud segmentation method according to any one of claims 1 to 7, characterized in that: The determining of the surface features corresponding to each of the block point clouds comprises: Using a point cloud processing model, determining the surface features corresponding to each of the block point clouds; Alternatively, the principal component analysis method is used to determine the surface features corresponding to each of the block point clouds.

9. A point cloud segmentation device, characterized in that it comprises: A partitioning module is used to partition the original point cloud to obtain multiple block point clouds; A first determination module is used to determine the surface features corresponding to each of the block point clouds, where the surface features include at least one of a plane feature, a non-plane feature and a roughness feature; A second determination module is used to determine that each of the block point clouds has a corresponding ground attribute or a non-ground attribute based on the surface features corresponding to each of the block point clouds; The segmentation module is used to segment the original point cloud based on the ground attribute or non-ground attribute corresponding to each of the block point clouds to obtain a ground point cloud and a non-ground point cloud.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is used to execute the point cloud segmentation method described in any one of claims 1 to 8.

Images