Laser radar point cloud clustering method, device, electronic equipment and vehicle

Abstract

This invention relates to the field of radar imaging technology and provides a method, apparatus, electronic device, and vehicle for point cloud clustering using lidar. The method includes: acquiring point cloud data within a preset range of a vehicle detected by lidar; projecting the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data; selecting the projection points closest to the vehicle at each yaw angle based on the distance between the projection points and the vehicle, obtaining multiple contour points; clustering the multiple contour points, and grouping the point cloud data in the grid containing the same type of contour points into the same point cloud cluster, thus obtaining the point cloud clustering result. This invention enables a low-latency point cloud clustering algorithm, meeting the detection requirements of autonomous driving perception functions in vehicles.

Description

Technical Field

[0001] The present invention belongs to the field of radar imaging technology, and in particular relates to a laser radar point cloud clustering method, device, electronic equipment and vehicle. Background Art

[0002] On-board laser radar is an important sensing device in autonomous driving. It can obtain the three-dimensional structure of external objects by scanning the surface of objects with laser beams and detecting their distance, shape, position and other information. Point cloud clustering algorithm is an important point cloud analysis technology that can group and identify the point cloud data detected by on-board laser radar, thereby extracting different targets and features in the data.

[0003] In the related technologies, point cloud clustering algorithms can be mainly divided into clustering algorithms based on spatial neighborhood, clustering algorithms based on feature similarity, and clustering algorithms based on model fitting. The above algorithms have good clustering effects on simple objects in simple scenes, but as part of the autonomous driving perception algorithm, the clustering algorithm also needs to have the characteristics of low latency and high precision, and can quickly and accurately identify target obstacles, respond, and avoid collisions. The above algorithms are difficult to meet the detection requirements of the vehicle's autonomous driving perception function. Summary of the invention

[0004] In view of this, an embodiment of the present invention provides a lidar point cloud clustering method, device, electronic device and vehicle to implement a low-latency point cloud clustering algorithm to meet the detection requirements of the vehicle's autonomous driving perception function.

[0005] A first aspect of an embodiment of the present invention provides a laser radar point cloud clustering method, comprising:

[0006] Obtain point cloud data within a preset range of the vehicle detected by the laser radar;

[0007] Project the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data;

[0008] According to the distance between the projection point and the vehicle, the projection point closest to the vehicle at each yaw angle of the vehicle is selected to obtain multiple contour points;

[0009] Clustering is performed on multiple contour points, and the point cloud data in the grid where the contour points of the same type are located are divided into the same point cloud cluster to obtain the point cloud clustering result.

[0010] In combination with the first aspect, in a possible implementation manner of the first aspect, clustering the multiple contour points includes:

[0011] According to the coordinate information of each contour point, the distance between adjacent contour points and the difference of the curvature of adjacent contour points are calculated;

[0012] Multiple contour points are clustered according to the distances between adjacent contour points and the differences in the curvatures of adjacent contour points.

[0013] In combination with the first aspect, in a possible implementation manner of the first aspect, calculating the difference in curvature of adjacent contour points according to the coordinate information of each contour point includes:

[0014] For each contour point, determine whether there is an adjacent contour point whose distance thereto is less than a preset first distance threshold;

[0015] If it exists, the curvature of the contour point is calculated according to the coordinate information of the contour point and the coordinate information of the adjacent contour points whose distance thereto is less than a preset first distance threshold;

[0016] If it does not exist, the curvature of the contour point is set to infinity;

[0017] According to the curvature of each contour point, the difference of the curvatures of adjacent contour points is calculated.

[0018] In combination with the first aspect, in a possible implementation manner of the first aspect, clustering a plurality of contour points according to a difference between distances between adjacent contour points and curvatures of adjacent contour points includes:

[0019] If the distance between adjacent contour points is less than a preset second distance threshold, and the difference in curvature of adjacent contour points is less than a preset first curvature difference threshold or greater than a preset second curvature difference threshold, the adjacent contour points are classified into the same category;

[0020] The first curvature difference threshold is smaller than the second curvature difference threshold.

[0021] In combination with the first aspect, in a possible implementation manner of the first aspect, after dividing the point cloud data in the grid where the same type of contour points obtained by clustering are located into the same point cloud cluster, the method further includes:

[0022] Each point cloud cluster is clustered again using the neighbor propagation clustering algorithm;

[0023] Based on the results of the secondary clustering, the point cloud clustering results are updated.

[0024] In combination with the first aspect, in a possible implementation manner of the first aspect, after acquiring point cloud data within a preset range of the vehicle detected by the laser radar, the method further includes:

[0025] Each grid is layered in height to obtain multiple levels;

[0026] The ratio of the number of point cloud data in each level to the number of point cloud data in the grid where the level is located is calculated, and the point cloud data in the level with a ratio less than a preset ratio threshold is filtered out as noise.

[0027] In combination with the first aspect, in a possible implementation manner of the first aspect, after acquiring point cloud data within a preset range of the vehicle detected by the laser radar, the method further includes at least one of the following:

[0028] Filter out point cloud data with intensity values ​​less than a preset first intensity threshold and point cloud data with intensity values ​​greater than a preset second intensity threshold as noise points, wherein the first intensity threshold is less than the second intensity threshold;

[0029] Calculate the number of point cloud data in each grid, and filter out the point cloud data in the grid whose number of point cloud data is less than a preset first number threshold as noise points;

[0030] Calculate the number of different laser radar beams in each grid, and filter out the point cloud data in the grid where the number of different laser radar beams is less than a preset second number threshold as noise points;

[0031] The number of valid grids corresponding to each grid is calculated, and the point cloud data in the grids whose corresponding valid grid number is less than a preset third number threshold is filtered out as noise points; the number of valid grids corresponding to each grid is the number of grids with point cloud data within a preset range around the center of each grid.

[0032] A second aspect of an embodiment of the present invention provides a laser radar point cloud clustering device, comprising:

[0033] An acquisition module is used to acquire point cloud data within a preset range of the vehicle detected by the laser radar;

[0034] A projection module, used to project the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data;

[0035] A screening module is used to screen out the projection points closest to the vehicle at each yaw angle of the vehicle according to the distance between the projection points and the vehicle, and obtain multiple contour points;

[0036] The clustering module is used to cluster multiple contour points, divide the point cloud data in the grid where the same type of contour points are located into the same point cloud cluster, and obtain the point cloud clustering result.

[0037] A third aspect of an embodiment of the present invention provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the method in the first aspect or any one of the implementations of the first aspect are implemented.

[0038] A fourth aspect of an embodiment of the present invention provides a vehicle, which includes the electronic device according to the third aspect.

[0039] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0040] The embodiment of the present invention takes into account that the main purpose of perception in autonomous driving is to identify targets and prevent collisions, and the focus should be on the target that may collide first. Therefore, the point cloud data is projected onto a two-dimensional grid, and the nearest projection point at each yaw angle of the vehicle is screened out according to the distance between the projection point and the vehicle to obtain multiple contour points. The contour points represent the shape of objects near the vehicle. By clustering the contour points, the point cloud data in the grid where the contour points are located is clustered, which reduces the amount of clustering processing data and greatly speeds up the clustering speed, meeting the low-latency perception strategy requirements in vehicle autonomous driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0042] Figure 1 It is a schematic diagram of the implementation process of the laser radar point cloud clustering method provided by an embodiment of the present invention;

[0043] Figure 2 is a schematic diagram of contour point selection provided by an embodiment of the present invention;

[0044] Figure 3 is a schematic diagram of contour point clustering provided by an embodiment of the present invention;

[0045] Figure 4 is a normal distribution diagram of the point cloud data intensity provided by an embodiment of the present invention;

[0046] Figure 5 is an overall flow chart of the laser radar point cloud clustering method provided by an embodiment of the present invention;

[0047] Figure 6 is a structural schematic diagram of a laser radar point cloud clustering device provided in an embodiment of the present invention;

[0048] Figure 7 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0049] In the following description, specific details such as specific system structures, technologies, etc. are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present invention. However, it should be clear to those skilled in the art that the present invention may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to prevent unnecessary details from obstructing the description of the present invention.

[0050] In order to illustrate the technical solution of the present invention, a specific embodiment is provided below for illustration.

[0051] Point cloud clustering algorithm plays an important role in connecting the previous and the next steps in the point cloud processing process. By clustering the obstacle point cloud obtained by ground segmentation, the discrete target point cloud is transformed into an operational target point cloud cluster, which can pave the way for subsequent outer bounding box fitting.

[0052] At present, traditional point cloud clustering algorithms can be mainly divided into:

[0053] (1) Clustering algorithm based on spatial neighborhood: Based on spatial distribution characteristics, similar discrete point clouds are grouped into one category. According to the point cloud division method, it can be divided into point-based clustering and grid projection-based clustering. The time consumption of point-based clustering is proportional to the number of point clouds, and the calculation speed is related to the amount of point cloud data. It is not suitable for large-scale point clouds. In addition, according to the characteristics of lidar that the objects are dense near and sparse far, it is easy to over-segment large objects and distant objects. Clustering based on grid projection has a significant improvement in time consumption, but it will sacrifice a certain degree of accuracy and is prone to under-segmentation.

[0054] (2) Clustering algorithm based on feature similarity: Calculate certain features of the point cloud, such as normal vector, curvature, density, etc., and divide the point cloud according to the feature value as the judgment standard of similarity consistency. The main difficulty of this algorithm is to find a feature with high computational efficiency, high similarity of the same type and high recognition of different types. Due to the diversity of target objects, this feature value is difficult to achieve. That is, the complexity and universality of the feature will affect the calculation speed and clustering effect accordingly.

[0055] (3) Based on model fitting algorithm: Use consistent geometric models, such as simple geometric shapes such as straight lines, planes, circles, etc. as constraints, and select points that satisfy the model as similar point clouds. This method is only effective for simple objects of specific shapes, such as relatively flat ground and roadside signs. It is not effective when used alone and is not applicable to most targets that may appear in autonomous driving scenarios.

[0056] To this end, an embodiment of the present invention provides a laser radar point cloud clustering method, which can be integrated into the vehicle's automatic driving control system as a program, or written into a separate processor, which is used in the automatic driving control system to realize the clustering of point cloud data. The laser radar point cloud clustering method of the present invention is mainly used in mine automatic driving scenarios to achieve high-response and high-precision automatic driving in mine scenarios.

[0057] See also Figure 1 As shown, the method includes:

[0058] Step S101, obtaining point cloud data within a preset range of the vehicle detected by the laser radar.

[0059] The laser radar can be installed at the four corners of the vehicle to detect 360° point cloud data around the vehicle. The point cloud data in this embodiment refers to the non-ground point cloud after the ground segmentation module.

[0060] Generally, point cloud data is in the form of (x, y, z, intensity), where x, y, z are the three-dimensional coordinates of the point, and intensity is the intensity value of the point. This embodiment redefines the point cloud data by calculating the yaw angle and beam distribution properties (ring value) of the point cloud, thereby constructing a depth map. This not only better fits the laser radar point cloud scanning structure, but also this distribution method contains the spatial distribution information of the point cloud, which can quickly locate and query adjacent points in space, and better perform point cloud clustering and segmentation.

[0061] The yaw angle Yaw is the angle between the projection of the point cloud on the horizontal plane and the vehicle direction, which can be calculated according to the following formula:

[0062] M PI is the ratio of pi.

[0063] The beam distribution attribute is the scanning line of the laser radar. There are different models of laser radar, such as 4-line, 8-line, 16-line, 32-line, and 64-line.

[0064] In addition, the distance value of the projection of each point can be calculated

[0065] The attributes of each point after conversion include: x, y, z, ring, yaw, intensity, range.

[0066] Step S102, projecting the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data.

[0067] Point cloud two-dimensional gridding is the process of mapping three-dimensional point cloud data in space to a horizontal two-dimensional grid. Each point in space corresponds to a projection point on the two-dimensional grid.

[0068] Step S103 , based on the distance between the projection point and the vehicle, the projection point at each yaw angle of the vehicle that is closest to the vehicle is screened out to obtain a plurality of contour points.

[0069] This embodiment considers that the main purpose of perception in autonomous driving is to identify targets and prevent collisions. The focus should be on the target that may collide first. Therefore, the point closest to the vehicle at the same yaw angle is selected as the focus object, which can further reduce the amount of data and improve operating efficiency. Specifically, within the yaw angle range of 0-360°, an angle threshold can be set, for example, in units of 1 degree, and the projection point closest to the vehicle within each yaw angle of 1 degree is detected as a contour point. Finally, there are 360 ​​contour points around the vehicle, forming the contour of the objects around the vehicle. Figure 2 It is an explanation of selecting the nearest point within the yaw angle.

[0070] Step S104 , clustering the multiple contour points, dividing the point cloud data in the grid where the same type of contour points are located obtained by clustering into the same point cloud cluster, and obtaining a point cloud clustering result.

[0071] By clustering the contour points, the point cloud data of the grid where the contour points of the same category are located are also divided into one category, which greatly reduces the number of cluster points and improves the clustering rate.

[0072] It can be seen that the embodiment of the present invention considers that the main purpose of perception in autonomous driving is to identify targets and prevent collisions, and the focus should be on the target that may collide first. Therefore, the point cloud data is projected onto a two-dimensional grid, and the nearest projection point at each yaw angle of the vehicle is screened out according to the distance between the projection point and the vehicle to obtain multiple contour points. The contour point represents the shape of the object near the vehicle. By clustering the contour points, the point cloud data in the grid where the contour points are located is clustered, which reduces the amount of clustering processing data and greatly speeds up the clustering speed, meeting the low-latency perception strategy requirements in vehicle autonomous driving.

[0073] In one embodiment, step S104 clusters the multiple contour points, including:

[0074] According to the coordinate information of each contour point, the distance between adjacent contour points and the difference of the curvature of adjacent contour points are calculated;

[0075] Multiple contour points are clustered according to the distances between adjacent contour points and the differences in the curvatures of adjacent contour points.

[0076] In an embodiment, for each contour point, it is determined whether there is an adjacent contour point whose distance to the contour point is less than a preset first distance threshold. If there is, the curvature of the contour point is calculated according to the coordinate information of the contour point and the coordinate information of the adjacent contour point whose distance to the contour point is less than the preset first distance threshold; if there is no such point, the curvature of the contour point is set to infinity; and the difference of the curvatures of the adjacent contour points is calculated according to the curvatures of the respective contour points. It is understandable that if the distance between the contour point and the adjacent contour point is close, they may belong to the same object, and the curvature can also be calculated. If the distance between the contour point and the adjacent contour point is very far, the contour point may be a noise point, and the curvature is directly set to infinity.

[0077] After calculating the difference in curvature of adjacent contour points, clustering is performed in the following way:

[0078] If the distance between adjacent contour points is less than a preset second distance threshold, and the difference in curvature of adjacent contour points is less than a preset first curvature difference threshold or greater than a preset second curvature difference threshold, the adjacent contour points are classified into the same category; wherein the first curvature difference threshold is less than the second curvature difference threshold.

[0079] In this embodiment, two adjacent contour points that are close to each other and have basically the same curvature can be considered to be the same object, because the same target object is mostly scanned on the same surface, which tends to be consistent in direction. On the contrary, the distribution of noise points is messy and the distribution of noise points is messy. At the same time, considering the mining environment, if the direction changes dramatically, for example, the difference in curvature of adjacent contour points is close to 90°, it can be considered that they are two mutually perpendicular surfaces of the same object.

[0080] According to the above principle, a depth-first search is performed based on the neighborhood to perform fast clustering. For example, see Figure 3 As shown in the figure: points a and b are far apart and are classified into different classes; points b and c are close and have similar curvatures, so they are classified into the same class; points c and d have a large difference in curvature and are classified into different classes; points d, e, and f are classified into the same class. After the contour points are clustered, the point cloud data in the grid where the contour points of the same class are located are divided into the same point cloud cluster, and the length, width, and height of each point cloud cluster are calculated. According to the corresponding maximum and minimum thresholds of length, width, and height, the point cloud cluster that meets the conditions is selected as the result.

[0081] Through the solution of this embodiment, the points that need to be clustered are only contour points, and there is no need to cluster each point cloud, which greatly reduces the processing volume, speeds up the clustering speed, and can achieve accurate clustering.

[0082] In some embodiments, after dividing the point cloud data in the grid where the contour points of the same type are located into the same point cloud cluster, the following method may also be included:

[0083] Each point cloud cluster is clustered again using the neighbor propagation clustering algorithm;

[0084] Based on the results of the secondary clustering, the point cloud clustering results are updated.

[0085] Considering that large-area objects may be under-segmented, the AP clustering algorithm is used to perform secondary clustering on each point cloud cluster, which can effectively solve the under-segmentation problem of objects whose area is larger than the area threshold. When the point cloud cluster is not under-segmented, the point cloud cluster remains as one category after the AP clustering algorithm performs secondary clustering. When the point cloud cluster is under-segmented, the point cloud cluster will be divided into two or more categories after the AP clustering algorithm performs secondary clustering. Although the use of the AP clustering algorithm for secondary clustering sacrifices some processing speed, it can greatly improve the clustering accuracy, and the clustering speed is still faster than the traditional clustering algorithm overall.

[0086] In some embodiments, in order to identify target obstacles with high precision, after acquiring the point cloud data around the vehicle detected by the laser radar, the point cloud data can also be filtered for noise. Noise often refers to white noise and invalid laser points caused by dust, water, fog, etc. Noise is often different from the target point in terms of spatial distribution and reflectivity, and can be filtered out according to this rule.

[0087] As a possible implementation method, noise filtering of point cloud data may include:

[0088] Each grid is layered in height to obtain multiple levels;

[0089] The ratio of the number of point cloud data in each level to the number of point cloud data in the grid where the level is located is calculated, and the point cloud data in the level with a ratio less than a preset ratio threshold is filtered out as noise.

[0090] Exemplarily, for each grid, it can be layered at a height of 0.2m (i.e., in the z-axis direction), and the number of point cloud data in each level and the number of point cloud data in the grid where the level is located (i.e., the sum of the number of point cloud data in each level of the grid) are counted. The ratio of the two is calculated. If the ratio is less than the preset ratio threshold, it means that the point cloud data in the grid where the level is located may be mis-segmented by the ground segmentation module or the ground points under the tree canopy. In this case, the point cloud data of this level will be filtered out to improve the accuracy of identifying the target object.

[0091] As a possible implementation method, noise filtering of point cloud data may also include:

[0092] Point cloud data with intensity values ​​less than a preset first intensity threshold and point cloud data with intensity values ​​greater than a preset second intensity threshold are filtered out as noise points.

[0093] The intensity values ​​of the point cloud are normally distributed as a whole, such as Figure 4 As shown in the figure, the intensity values ​​of the dust point cloud are distributed within a certain intensity threshold. The intensity threshold is selected according to the actual situation, and dust points can be filtered out to a certain extent. Figure 4 In the above, points with intensity values ​​less than μ-2σ and greater than μ+2σ are considered as noise points, μ is the mathematical expectation of the normal distribution, σ is the variance of the normal distribution. Generally, the dust intensity threshold is selected between 20 and 30, and points less than the dust intensity threshold are considered as noise points.

[0094] As a possible implementation method, noise filtering of point cloud data may also include:

[0095] The number of point cloud data in each grid is calculated, and the point cloud data in the grid whose number of point cloud data is less than a preset first number threshold is filtered out as noise points.

[0096] For real targets, more point clouds can generally be scanned. When the number of point cloud data in the grid is very small, it can be considered as noise and filtered out.

[0097] As a possible implementation method, noise filtering of point cloud data may also include:

[0098] The number of different laser radar beams in each grid is calculated, and the point cloud data in the grid where the number of different laser radar beams is less than a preset second number threshold is filtered out as noise points.

[0099] For real targets, each beam of a general radar can scan them, so the number of different laser radar beams in a grid should be large. When a point in a grid is detected by a few beams, that is, the number of different laser radar beams in the grid is less than a preset second number threshold, it can be considered a noise point.

[0100] As a possible implementation method, noise filtering of point cloud data may also include:

[0101] The number of valid grids corresponding to each grid is calculated, and the point cloud data in the grids whose corresponding valid grid number is less than a preset third number threshold is filtered out as noise points; the number of valid grids corresponding to each grid is the number of grids with point cloud data within a preset range around the center of each grid.

[0102] Under normal working conditions, the point cloud of object detection is continuous, that is, for an object, its point cloud is distributed in multiple adjacent point clouds. If it is isolated, it may be a noise point.

[0103] This embodiment can improve the target recognition accuracy by integrating multiple noise filtering solutions.

[0104] In combination with the above embodiments, the overall process of the laser radar point cloud clustering method of the present application can be as follows: Figure 5 As shown, it includes: depth map construction → noise filtering → contour point calculation → target clustering.

[0105] This application uses fast clustering of integrated grid projection and precise clustering of AP clustering, and adds corresponding noise filtering strategies for the unique scenes in mining areas to achieve faster and more accurate clustering segmentation of the target, which is in line with the low-latency and high-precision perception strategy in autonomous driving.

[0106] It should be understood that the order of execution of the steps in the above embodiment does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present invention.

[0107] An embodiment of the present invention further provides a laser radar point cloud clustering device, see Figure 6 As shown, the device 60 includes:

[0108] An acquisition module 61 is used to acquire point cloud data within a preset range of the vehicle detected by the laser radar;

[0109] A projection module 62, used to project the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data;

[0110] A screening module 63 is used to screen out the projection points closest to the vehicle at each yaw angle of the vehicle according to the distance between the projection points and the vehicle, and obtain a plurality of contour points;

[0111] The clustering module 64 is used to cluster the multiple contour points, and divide the point cloud data in the grid where the same type of contour points are located into the same point cloud cluster to obtain a point cloud clustering result.

[0112] As a possible implementation manner, the clustering module 64 is specifically used for:

[0113] According to the coordinate information of each contour point, the distance between adjacent contour points and the difference of the curvature of adjacent contour points are calculated;

[0114] Multiple contour points are clustered according to the distances between adjacent contour points and the differences in the curvatures of adjacent contour points.

[0115] As a possible implementation manner, the clustering module 64 is specifically used for:

[0116] For each contour point, determine whether there is an adjacent contour point whose distance thereto is less than a preset first distance threshold;

[0117] If it exists, the curvature of the contour point is calculated according to the coordinate information of the contour point and the coordinate information of the adjacent contour points whose distance thereto is less than a preset first distance threshold;

[0118] If it does not exist, the curvature of the contour point is set to infinity;

[0119] According to the curvature of each contour point, the difference of the curvatures of adjacent contour points is calculated.

[0120] As a possible implementation manner, the clustering module 64 is specifically used for:

[0121] If the distance between adjacent contour points is less than a preset second distance threshold, and the difference in curvature of adjacent contour points is less than a preset first curvature difference threshold or greater than a preset second curvature difference threshold, the adjacent contour points are classified into the same category;

[0122] The first curvature difference threshold is smaller than the second curvature difference threshold.

[0123] As a possible implementation manner, after dividing the point cloud data in the grid where the contour points of the same type are located into the same point cloud cluster, the clustering module 64 is further used to:

[0124] Each point cloud cluster is clustered again using the neighbor propagation clustering algorithm;

[0125] Based on the results of the secondary clustering, the point cloud clustering results are updated.

[0126] As a possible implementation manner, after acquiring the point cloud data around the vehicle detected by the laser radar, the acquisition module 61 is also used to filter out noise points from the point cloud data.

[0127] Noise filtering of point cloud data, including

[0128] Each grid is layered in height to obtain multiple levels;

[0129] The ratio of the number of point cloud data in each level to the number of point cloud data in the grid where the level is located is calculated, and the point cloud data in the level with a ratio less than a preset ratio threshold is filtered out as noise.

[0130] Noise filtering of point cloud data also includes at least one of the following:

[0131] Filter out point cloud data with intensity values ​​less than a preset first intensity threshold and point cloud data with intensity values ​​greater than a preset second intensity threshold as noise points;

[0132] Calculate the number of point cloud data in each grid, and filter out the point cloud data in the grid whose number of point cloud data is less than a preset first number threshold as noise points;

[0133] Calculate the number of different laser radar beams in each grid, and filter out the point cloud data in the grid where the number of different laser radar beams is less than a preset second number threshold as noise points;

[0134] The number of valid grids corresponding to each grid is calculated, and the point cloud data in the grids whose corresponding valid grid number is less than a preset third number threshold is filtered out as noise points; the number of valid grids corresponding to each grid is the number of grids with point cloud data within a preset range around the center of each grid.

[0135] Figure 7 FIG. 1 is a schematic diagram of an electronic device 70 provided by an embodiment of the present invention. Figure 7 As shown, the electronic device 70 of this embodiment includes: a processor 71, a memory 72, and a computer program 73 stored in the memory 72 and executable on the processor 71, such as a laser radar point cloud clustering program. When the processor 71 executes the computer program 73, the steps in the above-mentioned laser radar point cloud clustering method embodiments are implemented, such as Figure 1 Alternatively, when the processor 71 executes the computer program 73, the functions of each module in the above-mentioned device embodiments are implemented, for example Figure 6 The functions of modules 61 to 64 are shown.

[0136] Exemplarily, the computer program 73 may be divided into one or more modules / units, one or more modules / units are stored in the memory 72, and executed by the processor 71 to implement the present invention. One or more modules / units may be a series of computer program instruction segments capable of implementing specific functions, and the instruction segments are used to describe the execution process of the computer program 73 in the electronic device 70.

[0137] The electronic device 70 may be a computing device such as a desktop computer, a notebook, a PDA, or a cloud server. The electronic device 70 may include, but is not limited to, a processor 71 and a memory 72. Those skilled in the art will appreciate that Figure 7 It is only an example of the electronic device 70 and does not constitute a limitation of the electronic device 70. It may include more or fewer components than shown in the figure, or a combination of certain components, or different components. For example, the electronic device 70 may also include input and output devices, network access devices, buses, etc.

[0138] The processor 71 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0139] The memory 72 may be an internal storage unit of the electronic device 70, such as a hard disk or memory of the electronic device 70. The memory 72 may also be an external storage device of the electronic device 70, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the electronic device 70. Further, the memory 72 may also include both an internal storage unit of the electronic device 70 and an external storage device. The memory 72 is used to store computer programs and other programs and data required by the electronic device 70. The memory 72 may also be used to temporarily store data that has been output or is to be output.

[0140] The technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In practical applications, the above-mentioned function allocation can be completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated in a processing unit, or each unit can exist physically separately, or two or more units can be integrated in one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, which will not be repeated here.

[0141] An embodiment of the present invention further provides a vehicle, which includes the electronic device as described above.

[0142] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0143] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0144] In the embodiments provided by the present invention, it should be understood that the disclosed devices / electronic devices and methods can be implemented in other ways. For example, the device / electronic device embodiments described above are only schematic. For example, the division of modules or units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0145] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0146] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0147] If the integrated module / unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the present invention implements all or part of the processes in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned various method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. Computer-readable media can include: any entity or device that can carry computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0148] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. Such modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the protection scope of the present invention.

Claims

1. A laser radar point cloud clustering method, characterized in that: include: Obtain point cloud data within a preset range of the vehicle detected by the laser radar; Projecting the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data; According to the distance between the projection point and the vehicle, the projection point at each yaw angle of the vehicle that is closest to the vehicle is selected to obtain a plurality of contour points; The plurality of contour points are clustered, and point cloud data in the grid where the contour points of the same type are located are divided into the same point cloud cluster to obtain a point cloud clustering result.

2. The laser radar point cloud clustering method according to claim 1, characterized in that: The clustering of the plurality of contour points comprises: According to the coordinate information of each contour point, the distance between adjacent contour points and the difference of the curvature of adjacent contour points are calculated; The plurality of contour points are clustered according to the distances between the adjacent contour points and the differences in the curvatures of the adjacent contour points.

3. The laser radar point cloud clustering method according to claim 2, characterized in that: According to the coordinate information of each contour point, the difference of the curvature of adjacent contour points is calculated, including: For each contour point, determine whether there is an adjacent contour point whose distance thereto is less than a preset first distance threshold; If it exists, the curvature of the contour point is calculated according to the coordinate information of the contour point and the coordinate information of the adjacent contour points whose distance thereto is less than a preset first distance threshold; If it does not exist, the curvature of the contour point is set to infinity; According to the curvature of each contour point, the difference of the curvatures of adjacent contour points is calculated.

4. The laser radar point cloud clustering method according to claim 2, characterized in that: The clustering of the plurality of contour points according to the difference between the distances between the adjacent contour points and the curvatures of the adjacent contour points comprises: If the distance between the adjacent contour points is less than a preset second distance threshold, and the difference in curvature of the adjacent contour points is less than a preset first curvature difference threshold or greater than a preset second curvature difference threshold, the adjacent contour points are classified into the same category; The first curvature difference threshold is smaller than the second curvature difference threshold.

5. The laser radar point cloud clustering method according to any one of claims 1 to 4, characterized in that: After dividing the point cloud data in the grid where the contour points of the same type obtained by clustering are located into the same point cloud cluster, the method further includes: Each point cloud cluster is clustered again using the neighbor propagation clustering algorithm; Based on the result of the secondary clustering, the point cloud clustering result is updated.

6. The laser radar point cloud clustering method according to any one of claims 1 to 4, characterized in that: After acquiring the point cloud data within a preset range of the vehicle detected by the laser radar, the method further includes: Each grid is layered in height to obtain multiple levels; The ratio of the number of point cloud data in each level to the number of point cloud data in the grid where the level is located is calculated, and the point cloud data in the level with a ratio less than a preset ratio threshold is filtered out as noise.

7. The laser radar point cloud clustering method according to claim 6, characterized in that: After acquiring the point cloud data within the preset range of the vehicle detected by the laser radar, the method further includes at least one of the following: Filter out point cloud data with intensity values ​​less than a preset first intensity threshold and point cloud data with intensity values ​​greater than a preset second intensity threshold as noise points, wherein the first intensity threshold is less than the second intensity threshold; Calculate the number of point cloud data in each grid, and filter out the point cloud data in the grid whose number of point cloud data is less than a preset first number threshold as noise points; Calculate the number of different laser radar beams in each grid, and filter out the point cloud data in the grid where the number of different laser radar beams is less than a preset second number threshold as noise points; The number of valid grids corresponding to each grid is calculated, and the point cloud data in the grids whose corresponding valid grid number is less than a preset third number threshold is filtered out as noise points; the number of valid grids corresponding to each grid is the number of grids with point cloud data within a preset range around the center of each grid.

8. A laser radar point cloud clustering device, characterized in that: include: An acquisition module is used to acquire point cloud data within a preset range of the vehicle detected by the laser radar; A projection module, used for projecting the point cloud data onto a two-dimensional grid to obtain projection points corresponding to the point cloud data; A screening module, for screening out the projection points closest to the vehicle at each yaw angle of the vehicle according to the distance between the projection points and the vehicle, to obtain a plurality of contour points; The clustering module is used to cluster the multiple contour points, divide the point cloud data in the grid where the same type of contour points are located into the same point cloud cluster, and obtain a point cloud clustering result.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A vehicle, characterized in that: Comprising the electronic device as claimed in claim 9.

Images