Point cloud filtering method and device and computer storage medium

Abstract

The invention provides a point cloud filtering method and device and a computer storage medium. The point cloud filtering method comprises the steps of obtaining an original point cloud; rasterizing the original point cloud to obtain a plurality of grid point clouds; based on the bottom area data point of each grid point cloud, determining a rough ground point cloud; acquiring a ground plane parameter of each grid point cloud by using the rough ground point cloud; and filtering ceiling point clouds of the original point clouds by using a ground plane model determined by the ground plane parameters of each grid point cloud. In this way, interference of ceiling drooping obstacles is effectively dealt with, and the problem of obstacle false detection is solved.

Description

Technical Field

[0001] The present application relates to the field of point cloud recognition technology, and in particular to a point cloud filtering method, device and computer storage medium. Background Art

[0002] With the rapid development of autonomous driving and mobile robot autonomous navigation, LiDAR, as a relatively important sensing element, plays a very important role in it. Indoor scenes are a relatively important application scenario. For indoor mobile robots and autonomous driving vehicles equipped with LiDAR sensors, ceiling filtering is an important task to prevent ceiling targets from being mistakenly identified as obstacles by vehicles.

[0003] Because indoor scenes such as underground garages usually have many drooping obstacles and various pipes on the ceiling, there is a problem of false detection of drooping obstacles in the existing technology. Summary of the invention

[0004] The present application provides a point cloud filtering method, device and computer storage medium.

[0005] In order to solve the above technical problems, the present application proposes a point cloud filtering method, which includes: obtaining an original point cloud; rasterizing the original point cloud to obtain a number of grid point clouds; determining a coarse ground point cloud based on the bottom area data points of each grid point cloud; using the coarse ground point cloud, obtaining the ground plane parameters of each grid point cloud; and filtering the ceiling point cloud of the original point cloud using a ground plane model determined by the ground plane parameters of each grid point cloud.

[0006] Wherein, after acquiring the original point cloud, the point cloud filtering method includes: performing through filtering on the original point cloud according to a preset region of interest.

[0007] The step of rasterizing the original point cloud to obtain a plurality of raster point clouds includes: evenly dividing the original point cloud into a plurality of raster point clouds in the X-axis direction according to a preset number of grids in the X-axis direction and in the Y-axis direction according to a preset format in the Y-axis direction.

[0008] Among them, determining the coarse ground point cloud based on the bottom area data points of each grid point cloud includes: sorting the data points in the grid point cloud along the Z-axis direction to obtain several bottom area data points with smaller Z-axis values; determining the coarse ground point cloud of the grid point cloud based on the average height determined by the several bottom area data points and a preset height threshold.

[0009] Among them, using the rough ground point cloud to obtain the ground plane parameters of each grid point cloud includes: calculating the plane parameters of the rough ground point cloud through a random sampling consistency algorithm; and using the plane parameters of the rough ground point cloud as the ground plane parameters of the grid point cloud.

[0010] Among them, after obtaining the ground plane parameters of each grid point cloud, the point cloud filtering method also includes: obtaining the unit vector of the ground plane parameters; obtaining the vector angle between the unit vector and the virtual ground plane normal vector; and using the ground plane parameters of the adjacent grid point clouds to update the ground plane parameters of the target grid point cloud whose vector angle is less than a preset threshold.

[0011] Among them, the method of using the ground plane parameters of the adjacent grid point cloud to update the ground plane parameters of the target grid point cloud whose vector angle is less than a preset threshold includes: replacing the ground plane parameters of the target grid point cloud with the ground plane parameters of the adjacent grid point cloud; or obtaining the parameter mean of the ground plane parameters of the adjacent grid point cloud and the ground plane parameters of the target grid point cloud as the updated ground plane parameters of the target grid point cloud.

[0012] Among them, after filtering the ceiling point cloud of the original point cloud using the ground plane model determined by the ground plane parameters of each grid point cloud, the point cloud filtering method also includes: obtaining the bottom point cloud of each grid point cloud; obtaining the central data point of each bottom point cloud; obtaining the height difference between two adjacent grid point clouds based on the coordinates of the central data points of two adjacent grid point clouds; and filtering the bottom point cloud of the grid point cloud with a larger central data point coordinate Z value in two adjacent grid point clouds whose height difference is greater than a preset threshold.

[0013] In order to solve the above technical problems, the present application proposes a point cloud filtering device, which includes a memory and a processor coupled to the memory; wherein the memory is used to store program data, and the processor is used to execute the program data to implement the above point cloud filtering method.

[0014] In order to solve the above technical problems, the present application proposes a computer storage medium, which is used to store program data. When the program data is executed by a computer, it is used to implement the above point cloud filtering method.

[0015] Different from the prior art, the beneficial effects of the present application are as follows: the point cloud filtering device obtains the original point cloud; the original point cloud is rasterized to obtain a plurality of raster point clouds; based on the bottom area data points of each raster point cloud, a rough ground point cloud is determined; using the rough ground point cloud, the ground plane parameters of each raster point cloud are obtained; and the ceiling point cloud of the original point cloud is filtered using the ground plane model determined by the ground plane parameters of each raster point cloud. In the above manner, the interference of ceiling sagging obstacles is effectively dealt with and the problem of false obstacle detection is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. 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 creative work.

[0017] Figure 1 It is a flowchart of the first embodiment of the point cloud filtering method provided by the present application;

[0018] Figure 2 is a flow chart of a second embodiment of the point cloud filtering method provided by the present application;

[0019] Figure 3 It is a structural schematic diagram of an embodiment of a point cloud filtering device provided by the present application;

[0020] Figure 4 It is a structural diagram of an embodiment of a computer storage medium provided by the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0022] The point cloud filtering method of the present application is applied to a point cloud filtering device, wherein the point cloud filtering device of the present application can be a server, or a system composed of a server and a local terminal cooperating with each other. Accordingly, the various parts of the point cloud filtering device, such as various units, sub-units, modules, and sub-modules, can all be set in the server, or can be set in the server and the local terminal respectively.

[0023] Furthermore, the above-mentioned server can be hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or it can be implemented as a single server. When the server is software, it can be implemented as multiple software or software modules, such as software or software modules used to provide distributed servers, or it can be implemented as a single software or software module, which is not specifically limited here. In some possible implementations, the point cloud filtering method of the embodiment of the present application can be implemented by a processor calling computer-readable instructions stored in a memory.

[0024] In one method of the prior art, the ceiling filtering method is usually based on a simple ROI filtering algorithm, and the target is filtered through the height information of the vehicle. This method is relatively easy to use for most floor scenes, but there is a problem that some sagging signs, exhaust ducts and other objects cannot be completely filtered. Due to the inaccurate grasp of the height threshold in the prior art, a height threshold that is too large will result in incomplete filtering, and a height threshold that is too small will filter out normal moving targets. Based on the prior point cloud map, filtering the ceiling point cloud information usually requires a map construction operation, and then the ceiling point cloud and ceiling sagging static obstacles in the map are removed according to the prior map information.

[0025] Effective filtering requires clear and effective obstacle information. In indoor scenes such as underground parking lots, there are usually drooping signs and ventilation ducts, which are located at low positions. At the same time, in cross-floor scenes, the slope of indoor ramps is usually relatively large, so the height information of the distant ground point cloud and the nearby ceiling point cloud in the laser coordinate system is consistent, which will filter out the ground information.

[0026] In another method of the prior art, a scene map is constructed and filtered through the ceiling prior information in the map. This method usually involves a better mapping algorithm and requires a certain update frequency of the map, otherwise short-term scene changes may affect the performance of the algorithm.

[0027] Usually, the scene offline mapping algorithm is more dependent on algorithm performance. If a high-performance algorithm is not used, there will be mapping errors, especially when crossing floors, the drift phenomenon will increase, affecting the subsequent ceiling filtering. If a drooping sign is suddenly installed in the scene and the map is not updated, there will be a false detection problem.

[0028] The present application can solve the problems encountered in the above-mentioned prior art methods, fully consider the ground information from the perspective of the lidar, filter by analyzing the relationship between the ceiling and sagging obstacles and the ground, avoid misfiltering, and solve the problem of ceilings across floors. Constraints can be performed through the ground information under the ceiling to effectively avoid misfiltering.

[0029] At the same time, this application does not require complex mapping algorithms, and can process scene information detected by laser point cloud in real time. It does not rely on map prior information, and can better ensure real-time performance.

[0030] This application proposes a point cloud filtering method, see Figure 1 , Figure 1 It is a flowchart of the first embodiment of the point cloud filtering method provided by the present application; Figure 2 It is a schematic diagram of the overall process of the point cloud filtering method provided by this application.

[0031] like Figure 1 As shown, the specific steps are as follows:

[0032] Step S11: Obtain the original point cloud.

[0033] Specifically, the point cloud filtering device obtains the original point cloud, which is an unprocessed initial point cloud and can be obtained through a laser radar.

[0034] Further, in an embodiment of the present application, after the point cloud filtering device obtains the original point cloud, the point cloud filtering method includes: performing through filtering on the original point cloud according to a preset region of interest.

[0035] Specifically, the point cloud is filtered through ROI to obtain P roi Point cloud. In a specific embodiment of the present application, the ROI area is set to {X: [-30, 30], Y: [-20, 20], Z: [-2.0, -2.5]} to ensure that the ground point cloud information is relatively sufficient.

[0036] It should be noted that similar numerical values ​​may also be used in other embodiments of the present application, and the present application does not limit the specific numerical values.

[0037] Step S12: rasterizing the original point cloud to obtain a plurality of raster point clouds.

[0038] In an embodiment of the present application, the point cloud filtering device rasterizes the point cloud, divides the scene point cloud from the perspective of the laser radar into a number of rectangular columns, and then operates each rectangular column individually to achieve the effect of converting a plane into a curved surface, solving the problems of uneven ground and ramp scenes.

[0039] Specifically, in the embodiment of the present application, the point cloud filtering device evenly divides the original point cloud into a plurality of grid point clouds in the X-axis direction according to a preset number of grids in the X-axis direction and in the Y-axis direction according to a preset format in the Y-axis direction.

[0040] Usually, the speed of vehicles in indoor scenes is slow, so a smaller ROI can be set to achieve the best effect. Dividing the cells into 2m rectangular grids along the X and Y directions can obtain a 30x20 fan-shaped grid point cloud and columnar Pillar information.

[0041] In other embodiments of the present application, other reference values ​​may also be used, and the present application does not make any specific limitation on the specific values.

[0042] This application does not directly limit the Z height information to filter ceiling obstacles. Directly limiting the Z height information to filter ceiling obstacles will cause vehicles, people and other targets connected to the ground to be truncated to a certain extent. At the same time, it cannot solve the steep slope problem. For up and down steep slope scenes, the calculation method of limiting the height will filter out the ground and wall information in the distance, causing false filtering. Therefore, Z should not be set too small. If it is too large, it cannot completely filter the hanging objects.

[0043] Step S13: Determine a rough ground point cloud based on the bottom area data points of each grid point cloud.

[0044] Among them, in one embodiment of the present application, the point cloud filtering device sorts the data points in the grid point cloud along the Z-axis direction to obtain several bottom area data points with smaller Z-axis values; and determines the coarse ground point cloud of the grid point cloud based on the average height determined by the several bottom area data points and a preset height threshold.

[0045] Specifically, the point cloud filtering device traverses each Pillar and sorts the grid point cloud from small to large according to the numerical value of the Z value. Usually the ground point is the lowest point, so that the plane information at the bottom can be guaranteed first.

[0046] Furthermore, the point cloud filtering device selects N points with small Z values, determines the position of the bottom area, and calculates the average height Z of the N points. mean According to the average height Z of the selected seed point cloud mean And the height threshold H, select the coarse ground point cloud P bottom .

[0047] Step S14: using the coarse ground point cloud, obtaining ground plane parameters of each grid point cloud.

[0048] In one embodiment of the present application, the point cloud filtering device calculates the plane parameters of the rough ground point cloud through a random sampling consistency algorithm; and uses the plane parameters of the rough ground point cloud as the ground plane parameters of the grid point cloud.

[0049] Specifically, the point cloud filtering device calculates P by using the random sampling consensus algorithm RANSAC bottomThe plane parameters S: Ax+By+CZ+D = 0. Among them, random sampling consistency is an iterative algorithm that estimates the parameters of a mathematical model from a set of observation data containing outliers, in which case the outliers have no effect on the estimated values.

[0050] Furthermore, the present application proposes an embodiment for updating the ground plane parameters. For details, please refer to Figure 2 , Figure 2 It is a flowchart of the second embodiment of the point cloud filtering method provided by the present application.

[0051] like Figure 2 As shown, the details are as follows:

[0052] Step S21: Obtain the unit vector of the ground plane step parameter.

[0053] The plane normal vector is a unit vector composed of normalized coefficients n1 = [A, B, C], and the reference virtual ground plane normal vector is n2 = [A, B, C].

[0054] Step S22: Obtain the vector angle between the unit vector and the virtual ground plane normal vector.

[0055] Specifically, the point cloud filtering device calculates the angle theta between n1 and n2.

[0056] Step S23: using the ground plane parameters of the adjacent grid point clouds to update the ground plane parameters of the target grid point cloud whose vector angle is less than a preset threshold.

[0057] If the angle is greater than a certain threshold, the possibility that the plane is the ground plane is relatively low, and a comprehensive evaluation is required based on the ground plane parameters of adjacent frames to ensure the validity of the ground plane and prevent segmentation into the wall.

[0058] In one embodiment of the present application, the point cloud filtering device replaces the ground plane parameters of the target grid point cloud with the ground plane parameters of the adjacent grid point cloud. Alternatively, in another embodiment of the present application, the point cloud filtering device obtains the parameter mean of the ground plane parameters of the adjacent grid point cloud and the ground plane parameters of the target grid point cloud as the updated ground plane parameters of the target grid point cloud.

[0059] Specifically, the plane parameters of the adjacent left and right planes are used for average calculation or the adjacent ground plane parameters are directly assigned. In order to speed up the calculation efficiency, the plane parameters of the bottom point cloud of the grid Pillar are calculated by computer multithreading. Because the grids do not affect each other, the number of threads can be designed according to the computer performance, which greatly speeds up the calculation efficiency.

[0060] The present application marks and recalculates each grid ground plane parameter. If the ground plane normal vector of the current grid point cloud does not meet the verticality requirement and the angle between it and the reference normal vector is too large, it is necessary to use the left and right adjacent ground planes with better quality to replace it. Usually, the left and right adjacent ground planes can approximately reflect the plane parameters of the current plane. Therefore, in the embodiment of the present application, the current ground plane parameters are updated through the adjacent grid point clouds.

[0061] The existing technology usually does not perform special processing when using the RANSAC algorithm, such as seed point deletion. In a voxel, the upper and lower surfaces are both planes. If no preprocessing is performed, there will be erroneous segmentation.

[0062] Other methods in the prior art usually perform plane fitting based on a global plane, while the present application can greatly adapt to the ups and downs of the plane by rasterizing the point cloud, and is more in line with the real scene. Furthermore, the present application is accelerated through a multi-threaded approach, with high calculation efficiency and accurate calculation results.

[0063] The existing technology usually filters directly according to the Z value height, or filters through the prior height information of the floor of the map, which is not accurate enough, time-consuming and costly. The present application accelerates the process through multi-threading, with high calculation efficiency and accurate calculation results.

[0064] Step S15: Filtering the ceiling point cloud of the original point cloud using the ground plane model determined by the ground plane parameters of each grid point cloud.

[0065] Specifically, the point cloud filtering device uses the ground plane model determined by the ground plane parameters of each grid point cloud to filter the ceiling point cloud of the original point cloud through the ground plane model.

[0066] The point cloud filtering device filters the point cloud through the ground plane model S in the bottom area of ​​the grid, traversing P cell , calculate point p i Distance from plane S The point cloud of the ceiling in the grid is filtered out according to the distance threshold, and finally the non-ceiling point cloud is obtained.

[0067] In the embodiment of the present application, after filtering the ceiling point cloud of the original point cloud using the ground plane model determined by the ground plane parameters of each grid point cloud, the point cloud filtering method further includes:

[0068] The point cloud filtering device obtains the bottom point cloud of each grid point cloud. The center data point of each bottom point cloud is obtained; based on the coordinates of the center data points of two adjacent grid point clouds, the height difference between the two adjacent grid point clouds is obtained; and the bottom point cloud of the grid point cloud with the larger center data point coordinate Z value in the two adjacent grid point clouds whose height difference is greater than a preset threshold is filtered.

[0069] Specifically, the point cloud filtering device calculates the center point coordinates p of each bottom grid point cloud center , along the X direction, each column grid is constrained by the adjacent grids in front and behind, where ΔZ and ΔX represent the height difference and level difference of the bottom area of ​​adjacent grids, respectively. When ΔZ gradually increases or decreases with the change of X, it can be judged that the point cloud is uphill or downhill. Then, when the ΔZ of the adjacent grids in front and behind exceeds a certain threshold, the distant grids can be truncated.

[0070] Furthermore, the point cloud filtering device performs ceiling point cloud filtering on the corrected grid, and filters the point cloud through the ground plane model S in the bottom area of ​​the grid, traversing P cell , calculate point p i Distance from plane S The point cloud of the ceiling in the grid is filtered out according to the distance threshold, and the non-ceiling point cloud is finally obtained.

[0071] This application proposes a filtering method for the top ceiling interference point cloud during the application of LiDAR in indoor scenes. The method can effectively deal with the interference of obstacles hanging from the ceiling and solve the problem of false detection of obstacles. The ceiling information is identified and filtered through point cloud rasterization and ground parameter constraint relationships between adjacent grid point clouds. The ground plane parameters of all grid block domains in the rasterized ground can also be quickly calculated.

[0072] In order to implement the point cloud filtering method of the above embodiment, the present application also provides a point cloud filtering device. Figure 3 , Figure 3 It is a structural schematic diagram of an embodiment of a point cloud filtering device provided in the present application.

[0073] like Figure 3 As shown, the point cloud filtering device 600 of this embodiment includes a processor 61 , a memory 62 , an input and output device 63 and a bus 64 .

[0074] The processor 61 , the memory 62 , and the input / output device 63 are respectively connected to the bus 64 . The memory 62 stores a computer program, and the processor 61 is used to execute the computer program to implement the point cloud filtering method of the above embodiment.

[0075] In this embodiment, the processor 61 may also be referred to as a CPU (Central Processing Unit). The processor 61 may be an integrated circuit chip having the ability to process signals. The processor 61 may also be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The processor 61 may also be a GPU (Graphics Processing Unit), also known as a display core, a visual processor, or a display chip, which is a microprocessor that is specifically used for image computing on computers, workstations, game consoles, and some mobile devices (such as tablet computers, smart phones, etc.). The purpose of the GPU is to convert and drive the display information required by the computer system, and to provide a line scan signal to the display to control the correct display of the display. It is an important component that connects the display and the computer motherboard. As an important component of the computer host, the graphics card is responsible for outputting display graphics. The general-purpose processor may be a microprocessor or the processor 61 may also be any conventional processor, etc.

[0076] The present application also provides a computer storage medium, such as Figure 4 As shown, the computer storage medium 700 is used to store a computer program 71. When the computer program 71 is executed by a processor, it is used to implement the method described in the embodiment of the point cloud filtering method of the present application.

[0077] The method involved in the point cloud filtering method embodiment of the present application, when implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a device, such as a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor (processor) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program code.

[0078] The above description is only an implementation mode of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A point cloud filtering method, characterized in that: The point cloud filtering method comprises: Get the original point cloud; Rasterizing the original point cloud to obtain a plurality of raster point clouds; Determine a coarse ground point cloud based on the bottom area data points of each grid point cloud; Using the coarse ground point cloud, obtaining ground plane parameters of each grid point cloud; The ceiling point cloud of the original point cloud is filtered using a ground plane model determined by ground plane parameters of each grid point cloud.

2. The point cloud filtering method according to claim 1, characterized in that: After obtaining the original point cloud, the point cloud filtering method includes: The original point cloud is subjected to through filtering according to a preset region of interest.

3. The point cloud filtering method according to claim 1 or 2, characterized in that: The step of rasterizing the original point cloud to obtain a plurality of raster point clouds includes: The original point cloud is evenly divided into a number of grid point clouds in the X-axis direction according to a preset number of grids in the X-axis direction and in the Y-axis direction according to a preset format in the Y-axis direction.

4. The point cloud filtering method according to claim 1, characterized in that: The step of determining a rough ground point cloud based on the bottom area data points of each grid point cloud comprises: Sort the data points in the grid point cloud according to the Z-axis direction to obtain a number of bottom area data points with smaller Z-axis values; A coarse ground point cloud of the grid point cloud is determined according to an average height determined by the plurality of bottom area data points and a preset height threshold.

5. The point cloud filtering method according to claim 4, characterized in that: The step of obtaining the ground plane parameters of each grid point cloud by using the rough ground point cloud includes: The rough ground point cloud is subjected to a random sampling consistency algorithm to calculate the plane parameters of the rough ground point cloud; The plane parameters of the rough ground point cloud are used as the ground plane parameters of the grid point cloud.

6. The point cloud filtering method according to claim 1, characterized in that: After obtaining the ground plane parameters of each grid point cloud, the point cloud filtering method further includes: Obtaining a unit vector of the ground plane parameter; Obtaining the vector angle between the unit vector and the virtual ground plane normal vector; The ground plane parameters of the target grid point cloud whose vector angle is less than a preset threshold are updated using the ground plane parameters of the adjacent grid point cloud.

7. The point cloud filtering method according to claim 6, characterized in that: The updating of the ground plane parameters of the target grid point cloud whose vector angle is less than a preset threshold by using the ground plane parameters of the adjacent grid point cloud comprises: Replacing the ground plane parameters of the target grid point cloud with the ground plane parameters of the adjacent grid point cloud; Alternatively, a parameter average of the ground plane parameters of the adjacent grid point cloud and the ground plane parameters of the target grid point cloud is obtained as the updated ground plane parameter of the target grid point cloud.

8. The point cloud filtering method according to claim 1, characterized in that: After filtering the ceiling point cloud of the original point cloud using the ground plane model determined by the ground plane parameters of each grid point cloud, the point cloud filtering method further includes: Obtaining the bottom point cloud of each grid point cloud; Get the central data point of each bottom point cloud; Based on the coordinates of the central data points of two adjacent grid point clouds, the height difference between the two adjacent grid point clouds is obtained; The bottom point cloud of the grid point cloud with a larger central data point coordinate Z value among two adjacent grid point clouds whose height difference is greater than a preset threshold is filtered.

9. A point cloud filtering device, characterized in that: The point cloud filtering device includes a memory and a processor coupled to the memory; The memory is used to store program data, and the processor is used to execute the program data to implement the point cloud filtering method according to any one of claims 1 to 8.

10. A computer storage medium, characterized in that: The computer storage medium is used to store program data, and when the program data is executed by a computer, it is used to implement the point cloud filtering method according to any one of claims 1 to 8.

Images