WO2021016751A1 - Method for extracting point cloud feature points, point cloud sensing system, and mobile platform - Google Patents

Abstract

A method for extracting point cloud feature points, a point cloud sensing system, and a mobile platform. The method comprises: acquiring a three-dimensional point cloud of an environment where a mobile platform is located (S201); projecting, in a height direction, the three-dimensional point cloud to a horizontal plane to obtain a two-dimensional projected point cloud (S202); performing rasterization processing on and compiling feature information statistics on the two-dimensional projected point cloud to generate a point cloud information map (S203); and determining point cloud feature points on the basis of the point cloud information map (S204). According to the embodiments of the present invention, the feature points in the point cloud can be effectively extracted on the basis of the point cloud information map.

Description

Point cloud feature point extraction method, point cloud sensing system and movable platform Technical field

This application relates to the field of data processing technology, and in particular to a point cloud feature point extraction method, a point cloud sensing system and a movable platform.

Background technique

Positioning technology can provide position and other information for the movable platform, which is a prerequisite for path planning, motion control and autonomous decision-making of the movable platform. The current more mature method is based on the point cloud sensor to locate the movable platform. Based on the point cloud sensor positioning, the complete point cloud collected by the point cloud sensor needs to be calculated online and matched based on all the point clouds in the effective range. Since the amount of complete point cloud data collected by the point cloud sensor is usually very large, the amount of calculation is large when performing online calculation of the complete point cloud, which results in the consumption of a large amount of computing resources and the high cost of algorithm application. In order to reduce the amount of calculations in the positioning process, it can be considered that only part of the point cloud in the complete point cloud is calculated, but how to extract the part of the point cloud used for positioning calculation in the complete point cloud is a problem to be solved.

Summary of the invention

The embodiment of the invention discloses a point cloud feature point extraction method, a point cloud sensing system and a movable platform, which can effectively extract feature points in a point cloud based on a point cloud information graph.

The first aspect of the embodiments of the present invention discloses a point cloud feature point extraction method, the method includes:

Obtain the 3D point cloud of the environment where the mobile platform is located;

Projecting the three-dimensional point cloud onto a horizontal plane along the height direction to obtain a two-dimensional projection point cloud;

Performing rasterization processing and feature information statistics on the two-dimensional projection point cloud to generate a point cloud information map;

Determine point cloud feature points based on the point cloud information map.

A second aspect of the embodiments of the present invention discloses a point cloud sensing system, including: a point cloud sensor, a memory, and a processor, and the memory is used to store program instructions;

The processor is configured to execute program instructions stored in the memory, and when the program instructions are executed, the processor is configured to:

Acquiring a three-dimensional point cloud of the environment in which the point cloud sensing system is located through the point cloud sensor;

Projecting the three-dimensional point cloud onto a horizontal plane along the height direction to obtain a two-dimensional projection point cloud;

Performing rasterization processing and feature information statistics on the two-dimensional projection point cloud to generate a point cloud information map;

Determine point cloud feature points based on the point cloud information map.

The third aspect of the embodiments of the present invention discloses a movable platform, including:

body;

A power system installed on the fuselage and used to provide power for the movable platform;

The point cloud sensing system as described in the second aspect above.

The third and fourth aspects of the embodiments of the present invention disclose a computer-readable storage medium in which a computer program is stored. When the computer program is executed by a processor, the method described in the first aspect is implemented. step.

The embodiment of the present invention obtains a three-dimensional point cloud of the environment where the movable platform is located, and projects the three-dimensional point cloud to a horizontal plane along the height direction to obtain a two-dimensional projection point cloud; performs rasterization processing and features on the two-dimensional projection point cloud Information statistics, generating a point cloud information map, and determining point cloud feature points based on the point cloud information map, so that the feature points in the point cloud can be effectively extracted based on the point cloud information map.

Description of the drawings

In order to explain the technical solutions in the embodiments of the present invention more clearly, the following will briefly introduce the drawings needed in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative labor, other drawings can be obtained from these drawings.

Figure 1 is a schematic structural diagram of a movable platform disclosed in an embodiment of the present invention;

2 is a schematic flowchart of a point cloud feature point extraction method disclosed in an embodiment of the present invention;

FIG. 3 is a schematic diagram of point cloud distribution disclosed in an embodiment of the present invention;

Fig. 4 is a schematic structural diagram of a point cloud sensing system disclosed in an embodiment of the present invention.

Detailed ways

The technical solutions in the embodiments of the present invention will be described clearly and completely in conjunction with the accompanying drawings in the embodiments of the present invention. In the case of no conflict, the following embodiments and features in the embodiments can be combined with each other.

Please also refer to FIG. 1, which is a schematic structural diagram of a movable platform provided by an embodiment of the present invention. As shown in FIG. 1, the movable platform includes: a fuselage 101, a power system 102, a point cloud sensing system 103 and a pan-tilt 104. The power system 102 is installed on the fuselage 101 to provide power for the movable platform. The point cloud sensing system includes a point cloud sensor 1031, a processor 1032, and a memory 1033. The point cloud sensor 1031 is a sensing device for collecting three-dimensional point cloud data of the environment where the mobile platform or the point cloud sensing system 103 is located. In some embodiments, the point cloud sensor 1031 may be a lidar. The point cloud sensor 1031 is carried on the fuselage 101 of the movable platform through the pan/tilt 104, specifically, the pan/tilt 104 is installed on the fuselage 101 of the movable platform, and the point cloud sensor 1031 is fixed on the pan/tilt 104; 104 can drive the point cloud sensor 1031 to rotate around one or more of the yaw axis, roll axis, and pitch axis, so as to adjust the attitude of the point cloud sensor 1031 to collect three-dimensional point cloud data. In addition, in some embodiments, the point cloud sensor 1031 may be directly carried on the body 101 of the movable platform.

Among them, the memory 1033 is used to store program instructions, and the processor 1032 is used to execute the program instructions stored in the memory 1033. When the program instructions are executed, the processor 1032 is used to: obtain the point cloud sensing system 103 or the point cloud sensor system 103 through the point cloud sensor 1031 The three-dimensional point cloud of the environment where the movable platform is located, and project the obtained three-dimensional point cloud to a horizontal plane along the height direction to obtain a two-dimensional projection point cloud; rasterize the two-dimensional projection point cloud and perform feature information statistics, Generate a point cloud information map, and determine the point cloud feature points based on the point cloud information map. With the above method, the feature points in the point cloud can be effectively extracted based on the point cloud information graph, so that the extracted feature points of the point cloud can be directly calculated when the point cloud sensor is positioned later, thereby effectively saving the amount of calculation.

It should be noted that the movable platform shown in FIG. 1 is described by taking a vehicle as an example. The movable platform in the embodiment of the present invention may also be an unmanned aerial vehicle (UAV), an unmanned ship, a mobile robot, etc. Removable equipment.

Please refer to FIG. 2, which is a schematic flowchart of a method for extracting feature points of a point cloud according to an embodiment of the present invention. The point cloud feature point extraction method described in the embodiment of the present invention can be applied to the movable platform shown in FIG. 1, and the point cloud feature point extraction method includes:

S201. Obtain a three-dimensional point cloud of the environment where the mobile platform is located.

In the embodiment of the present invention, the three-dimensional point cloud includes one or more of the complete three-dimensional point cloud of the environment where the movable platform is located, the ground three-dimensional point cloud, and the non-ground three-dimensional point cloud. The mobile platform collects the complete 3D point cloud of the environment where the mobile platform is located through its configured point cloud sensor, and can obtain the ground 3D point cloud and non-ground 3D points of the environment where the mobile platform is located based on the collected complete 3D point cloud cloud. Among them, the point cloud sensor is, for example, a sensing device based on lidar, and the point cloud described herein can carry coordinate information of a point, and can also carry reflectivity information of a point.

S202: Project the three-dimensional point cloud onto a horizontal plane along the height direction to obtain a two-dimensional projection point cloud.

In the embodiment of the present invention, the coordinate system corresponding to the three-dimensional coordinates of the three-dimensional point cloud is a spatial three-dimensional coordinate system, and the spatial three-dimensional coordinate system includes a horizontal axis (X axis), a vertical axis (Y axis), and a vertical axis (Z axis). If the Z axis is the height direction of the point cloud, the movable platform will project the 3D point cloud along the Z axis to the XY plane formed by the X axis and the Y axis to obtain a series of two-dimensional scatter points, which constitute two Dimensional projection point cloud.

S203: Perform rasterization processing and feature information statistics on the two-dimensional projection point cloud to generate a point cloud information map.

In the embodiment of the present invention, the rasterization processing includes performing rasterization and segmentation on the two-dimensional projection point cloud in the two-dimensional projection point cloud to obtain multiple grids; each grid corresponds to an area in the range of K*M. , K and M can be the same or different. The grid unit sizes K and M can be flexibly adjusted according to the positioning accuracy requirements, such as K=0.2m, M=0.2m.

When the acquired 3D point cloud includes the complete 3D point cloud of the environment where the mobile platform is located, the mobile platform performs rasterization processing and height information statistics on the 2D projection point cloud corresponding to the complete 3D point cloud to generate the point cloud Height map, and use the generated point cloud height map as a point cloud information map. In an embodiment, the generated point cloud height map includes a point cloud height average map. The mobile platform first performs rasterization processing on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids; then calculates the height average of the height values corresponding to the two-dimensional projection point cloud in each grid, and Express each grid with grid coordinates and average height to generate the average point cloud height map, that is, use a grid as a pixel of the average height map, and the grid coordinates are also the corresponding pixel points Coordinates, the average height of the two-dimensional projection point cloud in the grid is also the pixel value of the corresponding pixel. In another embodiment, the generated point cloud height map includes a point cloud height variance map. The movable platform first performs rasterization processing on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids; then calculates the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid, Each grid is represented by grid coordinates and height variance value to generate the point cloud height variance map, that is, a grid is used as a pixel of the height variance map, and the grid coordinates are the corresponding pixels. The coordinate of the point, the height variance value of the two-dimensional projected point cloud in the grid is also the pixel value of the corresponding pixel.

In another embodiment, when the movable platform calculates the height average value and/or the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid, it first obtains that the corresponding height value in the target grid is positive or negative. The target two-dimensional projection point cloud within a preset number of standard deviations is calculated, and then the height average value and/or the height variance value of the height value corresponding to the target two-dimensional projection point cloud in the target grid is calculated. Among them, the target grid is any one of the multiple grids obtained after the rasterization of the two-dimensional projection point cloud, and the standard deviation is determined based on the height values corresponding to all the two-dimensional projection points in the target grid; the preset number For example, 2. Using the above method, before calculating the height average and/or height variance value of the two-dimensional projected point cloud in the grid, first remove the points with the height value in the grid outside the preset number of standard deviations, which is beneficial to the realization of the algorithm Robustness.

When the acquired 3D point cloud includes the ground 3D point cloud of the environment where the movable platform is located, the movable platform performs rasterization processing and reflectivity information statistics on the 2D projection point cloud corresponding to the ground 3D point cloud to generate points Cloud reflectivity map, and use the generated point cloud reflectivity map as a point cloud information map. In an embodiment, the generated point cloud reflectivity map includes a point cloud maximum reflectivity map. The movable platform first performs rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids; then calculates the maximum reflectance value corresponding to the reflectivity value of the two-dimensional projection point cloud in each grid , And express each grid with grid coordinates and maximum reflectance value to generate the point cloud maximum reflectance map, that is, use a grid as a pixel of the point cloud maximum reflectance map, and the grid coordinates That is, the coordinates of the corresponding pixel point, and the maximum reflectivity value of the two-dimensional projection point cloud in the grid is also the pixel value of the corresponding pixel point. In another embodiment, the generated point cloud reflectance map includes a point cloud reflectance variance map. The movable platform first performs rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids; then calculates the reflectance variance of the reflectance value corresponding to the two-dimensional projection point cloud in each grid The value of each grid is expressed by grid coordinates and reflectance variance value to generate the point cloud reflectivity variance map, that is, a grid is used as a pixel of the point cloud reflectivity variance map. The grid coordinates are also the coordinates of the corresponding pixels, and the reflectance variance value of the two-dimensional projection point cloud in the grid is also the pixel value of the corresponding pixels. In the same way, before calculating the reflectance variance value of the two-dimensional projected point cloud in the grid, you can also remove the points with the reflectance value in the grid outside the preset number of standard deviations. The specific implementation can refer to the previous description. , I won’t repeat it here.

S204: Determine point cloud feature points based on the point cloud information map.

In the embodiment of the present invention, when the point cloud information map includes a point cloud height average map, the movable platform may determine the height jump feature points in the point cloud based on the point cloud height average map. The movable platform obtains the adjacent grid adjacent to the first grid in the average point cloud height map. The first grid is any grid or pixel in the average point cloud height map. The adjacent grid can also be Say it is an adjacent pixel; then determine the second grid from the adjacent grid, the average height of the two-dimensional projection point cloud in the second grid and the average height of the two-dimensional projection point cloud in the first grid The absolute value of the difference between is greater than or equal to the first value; further, it is detected whether the number of second grids is greater than or equal to the second value, and if the number of second grids is greater than or equal to the second value, the first The two-dimensional projection point in a grid is determined as the height jump feature point. It should be noted that the first value and the second value may be preset values, or may be based on the average height of the two-dimensional projection point cloud in the first grid, and the adjacent grid adjacent to the first grid. Determined by the average height of the two-dimensional projection point cloud in the grid. The second value may also be determined based on the number of adjacent grids adjacent to the first grid.

For example, given a window with a size of 3*3, a window size of 3*3 means that the length and width of the window are all three grids in length. According to a given window, the height difference between a grid and its 8 neighborhood grids is counted, and the number of grids with different heights in the neighborhood grid is analyzed. For example, in one embodiment, may be _{n = sum (| p o -p} i | ≥d 1) to statistics, if n ≧ 3, i.e., the lattice 8 and its surrounding neighborhood raster least If the height difference of the three neighborhood grids is greater than the predetermined difference, the grid is considered to be a height jump grid, and the two-dimensional projection point in the grid _po is determined as the height jump feature point. Among them, the sum function is the summation function; p _o is the grid in the center of the given window, p _i is any one of the 8 grids adjacent to the grid p _{o in the} window; d ₁ can be based on The value determined by the height average value of the two-dimensional projection point cloud in each grid in the window may also be a preset value. Among them, the window size can be optimized and adjusted according to the actual area size of the grid, and the neighborhood pattern can be flexibly adjusted according to actual needs. After determining the height jump feature points in a given window, move the given window in the point cloud height average map according to the preset step length, and determine the height jump feature points in the moved window according to the above method, until it is determined All height jump feature points in the point cloud are outputted. The height jump feature points can quickly and accurately describe the height jump structure in the point cloud, such as static objects such as railings and walls in a road scene.

In another embodiment, when the point cloud information map includes a point cloud maximum reflectance map and a point cloud reflectance variance map, the movable platform may determine the line tracking feature based on the point cloud maximum reflectivity map and the point cloud reflectance variance map The feature point of the line inspection can be the feature point of the lane line and so on. The mobile platform divides the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud based on the multiple grids obtained by the rasterization process to obtain multiple point cloud regions, and each point cloud region includes multiple grids; further , Based on the point cloud maximum reflectance map and the point cloud reflectance variance map, determine the two-dimensional Gaussian mixture model of each point cloud area in the multiple point cloud areas, and obtain the two-dimensional Gaussian mixture model from the multiple point cloud areas based on the two-dimensional Gaussian mixture model of each point cloud area. Identify the characteristic points of the patrol in the point cloud area.

For example, because the material of the lane line is different from the material of the road surface, the reflectivity value of the lidar point cloud is greatly affected by the material, and the reflectivity of the lane line area is usually greater than that of the road surface area. Therefore, the lane line and the ground area can be distinguished by the reflectance value information. First, the ground three-dimensional point cloud is divided into regions to obtain multiple point cloud regions; each point cloud region includes multiple grids, and the area of each point cloud region is as 2m*2m. Further, a two-dimensional Gaussian mixture model of each point cloud area is determined based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map. Suppose the mathematical expression of a two-dimensional Gaussian mixture model of a certain point cloud area is N ₁ ～(u ₁ , sigma ₁ ) and N ₂ ～(u ₂ , sigma ₂ ), if |u ₁ –u ₂ |>d ₂ , then It can be considered that there is a lane line in the certain point cloud area, and the point in the certain point cloud area whose reflectance value is within one standard deviation is set as the lane line feature point. For example, the reflectance value of the lane line feature point is greater than (u ₂ -sigma ₂ ) or greater than (u ₁ -sigma ₁ ). Among them, u ₁ and u ₂ represent the mean reflectance values, and sigma ₁ and sigma ₂ represent the standard deviation of the reflectance. In the above method, the lane line feature points in the point cloud can be determined, so that the lane line position in the online laser point cloud can be determined based on the lane line feature points during subsequent positioning, and the lane line position in the online laser point cloud can be compared with the height The lane line position in the refined map can be used to obtain the current lateral positioning of the movable platform more accurately.

In the embodiment of the present invention, since the two-dimensional projection point cloud of the ground area has a smaller average height, neighborhood height difference, and point cloud height variance in the grid, etc., the two-dimensional object with complex geometric structures such as wall edges and trees The projected point cloud has a larger average height, neighborhood height difference, and point cloud height variance within the raster. Therefore, the feature point analysis of the two-dimensional projection point cloud based on the height map such as the point cloud height average map and the point cloud height variance map can well extract the stable height map feature points, such as extracting the point cloud feature points of the ground area and the tree cluster. Corresponding point cloud feature points, etc. In some embodiments, the movable platform can extract high-density feature points in the two-dimensional projection point cloud. The total number of point clouds in the area corresponding to the high-density feature points is larger, and the distance between the high-density feature points is greater than a certain threshold. The physical meaning of high-density feature points corresponds to static objects with obvious geometric structures such as traffic light poles, tree trunks, pillars, etc., such as traffic light poles, which are usually slender objects with a certain height, concentrated projection positions, and relatively short distance between two objects. Great features. In some embodiments, the movable platform can extract sparse feature points in the two-dimensional projection point cloud. The sum of the number of point clouds in the area corresponding to the sparse feature points is small. The sparse feature points correspond to static objects in the actual environment such as bushes and distant buildings. Due to the low angular resolution of the point cloud sensor, the distant objects scan the point cloud sensor for fewer positions and are distributed in a sparse state; while the sparse structure of trees and other objects determines the feature of sparse point cloud distribution.

In some embodiments, the movable platform can extract non-road feature points, and specifically can extract non-road feature points based on a non-ground three-dimensional point cloud. Non-road feature points correspond to points that remove all areas except ground points. Non-road feature points such as highway piers, buildings on both sides of the road, trees and other corresponding points have high geometric stability. In some embodiments, the movable platform can extract high reflectivity feature points of non-ground areas. In the point cloud of the non-ground area, the objects with high reflectivity are usually static metal objects such as railings, traffic signs, and billboards on both sides of the road. Therefore, extracting high reflectivity feature points in non-ground areas can obtain more static object features in non-ground areas.

In the embodiment of the present invention, after the movable platform determines the point cloud feature points, the position of the movable platform can be determined in the high-precision map based on the determined point cloud feature points. Currently, positioning based on a complete three-dimensional point cloud requires a large amount of online computing resources, and it is difficult to realize chip applications. This solution first effectively extracts the feature points in the point cloud based on the point cloud information map, and then performs calculations based on the extracted feature points of the point cloud to achieve positioning. Since the number of point cloud feature points extracted by the above method is much smaller than the total number of points of the complete three-dimensional point cloud, positioning based on the extracted point cloud feature points can greatly reduce the amount of calculation and effectively save computing resources. In addition, the point cloud feature points extracted by the above method include a large number of feature points corresponding to static objects in physical meaning. Since static objects can better express the structural stability of the environment and the similarity of the scene, it is based on the extracted point cloud Feature point positioning can also effectively ensure positioning accuracy. In addition, the extracted feature points of the point cloud can be quickly unitized and parallelized, and the processing process can be performed on a graphics processor (Graphics Processing Unit, GPU), field-programmable gate array (Field-Programmable Gate Array, FPGA) Realization, so the calculation and processing of the extracted point cloud feature points has good parallelism, good real-time performance, can meet the needs of high-speed processing on mobile platforms, and can realize chip applications.

In another embodiment, since the uniform distribution of the point cloud feature points directly affects the accuracy and stability of the positioning result, it is possible to first determine the position of the movable platform based on the determined point cloud feature points. The point cloud feature points are equalized. The embodiment of the present invention provides a point cloud feature point equalization method based on uniform distribution or non-uniform distribution. Specifically, the movable platform first converts the acquired point cloud feature points to polar coordinates to obtain points in polar coordinates Cloud feature points; then the point cloud feature points in polar coordinates are divided proportionally based on the angle and radial length to obtain the point cloud feature points after the proportional division, thereby completing the equalization of the point cloud feature points. Please also refer to Figure 3, which is a schematic diagram of the point cloud distribution obtained through experiments. The left image in Figure 3 shows the distribution of the acquired point cloud feature points after equalization; the right image in Figure 3 shows the original distribution of the acquired point cloud feature points. It can be seen that the unbalanced point cloud feature points are mainly concentrated in the area closer to the center position, the point cloud feature points in the area far from the center position are sparsely distributed, and the unbalanced point cloud feature points are unevenly distributed; and The point cloud feature points after equalization are more evenly distributed in each area. It can be seen that the point cloud feature point equalization method provided by the embodiment of the present invention can well balance the number of point cloud feature points at a distance and an intermediate position. Further, the movable platform determines the position of the movable platform in the high-precision map based on the point cloud feature points divided in proportion. The positioning based on the point cloud feature points after the equalization process can improve the positioning accuracy to a certain extent.

The embodiment of the present invention obtains a three-dimensional point cloud of the environment where the movable platform is located, and projects the three-dimensional point cloud to a horizontal plane along the height direction to obtain a two-dimensional projection point cloud; performs rasterization processing and features on the two-dimensional projection point cloud Information statistics, generating a point cloud information map, and determining point cloud feature points based on the point cloud information map, so that the feature points in the point cloud can be effectively extracted based on the point cloud information map.

Please refer to FIG. 4, which is a schematic structural diagram of a point cloud sensing system according to an embodiment of the present invention. The point cloud sensing system described in the embodiment of the present invention includes: a processor 401, a point cloud sensor 402, and a memory 403. Among them, the processor 401, the point cloud sensor 402, and the memory 403 may be connected through a bus or in other ways. The embodiment of the present invention takes the connection through a bus as an example.

The processor 401 may be a central processing unit (CPU), a graphics processing unit (Graphics Processing Unit, GPU), or a combination of a CPU and a GPU. The processor 401 may also be a multi-core CPU or a core in a multi-core GPU for implementing communication identification binding. The processor 401 may be a hardware chip. The hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD) or a combination thereof. The PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a general array logic (generic array logic, GAL) or any combination thereof.

The point cloud sensor 402 can be used to collect a three-dimensional point cloud of the environment in which the point cloud sensing system is located. The memory 403 may mainly include a storage program area and a storage data area. The storage program area may store an operating system and a storage program required by at least one function (such as a text storage function, a location storage function, etc.); the storage data area may store Data (such as image data, text data) created according to the use of the device, etc., and may include application storage programs, etc. In addition, the memory 403 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

The memory 403 is also used to store program instructions. The processor 401 is configured to execute program instructions stored in the memory 403, and when the program instructions are executed, the processor 401 is configured to: obtain the point cloud sensor system through the point cloud sensor 402 A three-dimensional point cloud of the environment; project the three-dimensional point cloud to a horizontal plane along the height direction to obtain a two-dimensional projection point cloud; perform rasterization processing and feature information statistics on the two-dimensional projection point cloud to generate a point cloud Information graph; determining point cloud feature points based on the point cloud information graph.

The methods executed by the processor in the embodiments of the present invention are all described from the perspective of the processor. It can be understood that the processor in the embodiments of the present invention requires the cooperation of other hardware structures to execute the foregoing methods. The embodiments of the present invention do not make detailed descriptions and restrictions on the specific implementation process.

In an embodiment, the three-dimensional point cloud includes one or more of a complete three-dimensional point cloud, a ground three-dimensional point cloud, and a non-ground three-dimensional point cloud.

In an embodiment, the three-dimensional point cloud includes a complete three-dimensional point cloud, and the processor 401 performs rasterization processing and feature information statistics on the two-dimensional projection point cloud, and is specifically used for generating a point cloud information map. : Perform rasterization processing and height information statistics on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud, generate a point cloud height map, and use the point cloud height map as a point cloud information map.

In an embodiment, the point cloud height map includes a point cloud height average map, and the processor 401 performs rasterization processing and height information statistics on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud, and generates points When the cloud height map is used, it is specifically used to: rasterize the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids; calculate the height value corresponding to the two-dimensional projection point cloud in each grid Height average value; each grid is represented by grid coordinates and height average value to generate the point cloud height average value map.

In an embodiment, when the processor 401 calculates the height average value of the height values corresponding to the two-dimensional projection point clouds in each grid, it is specifically used to: obtain the predetermined number of positive or negative height values corresponding to the target grid. A target two-dimensional projection point cloud within a standard deviation, where the target grid is any one of the multiple grids, and the standard deviation is based on height values corresponding to all two-dimensional projection points in the target grid Determined; calculating the height average value of the height values corresponding to the two-dimensional projection point cloud of the target in the target grid.

In an embodiment, the point cloud height map includes a point cloud height variance map, and the processor 401 performs rasterization processing and height information statistics on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to generate points When the cloud height map is used, it is specifically used to: rasterize the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids; calculate the height value corresponding to the two-dimensional projection point cloud in each grid Height variance value; each grid is represented by grid coordinates and height variance value to generate the point cloud height variance map.

In an embodiment, when the processor 401 calculates the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid, it is specifically used to: obtain the height value corresponding to the target grid in positive or negative preset A target two-dimensional projection point cloud within a number of standard deviations, where the target grid is any one of the multiple grids, and the standard deviation is based on the heights corresponding to all two-dimensional projection points in the target grid The value is determined; the height variance value of the height value corresponding to the two-dimensional projection point cloud of the target in the target grid is calculated.

In an embodiment, the three-dimensional point cloud includes a ground three-dimensional point cloud, and the processor 401 performs rasterization processing and feature information statistics on the two-dimensional projection point cloud, and when generating a point cloud information map, it is specifically used for : Perform rasterization processing and reflectivity information statistics on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud, generate a point cloud reflectivity map, and use the point cloud reflectivity map as a point cloud information map.

In an embodiment, the point cloud reflectivity map includes a point cloud maximum reflectivity map, and the processor 401 performs rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud and statistics of reflectivity information , When generating the point cloud reflectance map, it is specifically used to: rasterize the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids; calculate the location of the two-dimensional projection point cloud in each grid The maximum reflectance value corresponding to the reflectance value; each grid is represented by the grid coordinates and the maximum reflectance value to generate the point cloud maximum reflectance map.

In an embodiment, the point cloud reflectance map includes a point cloud reflectance variance map, and the processor 401 performs rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud and reflectance information statistics. , When generating the point cloud reflectance map, it is specifically used to: rasterize the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids; calculate the location of the two-dimensional projection point cloud in each grid The reflectance variance value corresponding to the reflectance value; each grid is represented by the grid coordinates and the reflectance variance value to generate the point cloud reflectance variance map.

In an embodiment, when the processor 401 determines the point cloud feature point based on the point cloud information map, it is specifically configured to determine the height jump feature point based on the point cloud height average map.

In an embodiment, when the processor 401 determines the height jump feature points based on the point cloud height average map, it is specifically configured to: obtain the adjacent point cloud height average map adjacent to the first grid Grid, the first grid is any grid in the point cloud height average map; a second grid is determined from the adjacent grids, and the two-dimensional projected point cloud in the second grid The absolute value of the difference between the average height of the two-dimensional projection point cloud in the first grid and the average height of the two-dimensional projection point cloud is greater than or equal to the first value; if the number of the second grids is greater than or equal to the second Numerical value, the two-dimensional projection point in the first grid is determined as the height jump feature point.

In an embodiment, the point cloud reflectivity map includes a point cloud maximum reflectivity map and a point cloud reflectivity variance map. When the processor 401 determines point cloud feature points based on the point cloud information map, it is specifically configured to : Determine line-following feature points based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map.

In an embodiment, when the processor 401 determines the line-following feature points based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map, it is specifically configured to: The three-dimensional projection point cloud is divided into regions to obtain multiple point cloud regions; based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map, the two-dimensional mixture of each point cloud region in the multiple point cloud regions is determined Gaussian model; determining line-following feature points from the multiple point cloud regions based on the two-dimensional Gaussian mixture model.

In an embodiment, after the processor 401 determines point cloud feature points based on the point cloud information map, it is further configured to: determine the point cloud sensing system's value in a high-precision map based on the point cloud feature points position.

In an embodiment, when the processor 401 determines the position of the point cloud sensing system in a high-precision map based on the point cloud feature points, it is specifically configured to: convert the point cloud feature points to polar coordinates Next, the point cloud feature points in polar coordinates are divided proportionally based on the angle and radial length; the position of the point cloud sensing system is determined on the high-precision map based on the point cloud feature points after the proportional division.

In specific implementation, the processor 401, the point cloud sensor 402, and the memory 403 described in the embodiment of the present invention can execute the implementation of the movable platform described in the method for extracting feature points of a point cloud provided by the embodiment of the present invention. I will not repeat them here.

The above-mentioned point cloud sensing system obtains the three-dimensional point cloud of the environment in which the point cloud sensing system is located through the point cloud sensor, and projects the three-dimensional point cloud to a horizontal plane along the height direction to obtain a two-dimensional projection point cloud; for the two-dimensional projection point cloud Perform rasterization and feature information statistics to generate a point cloud information map, and determine point cloud feature points based on the point cloud information map, so that the feature points in the point cloud can be effectively extracted based on the point cloud information map to facilitate subsequent point cloud based When the sensor is positioned, it directly calculates the extracted point cloud feature points, which effectively saves the amount of calculation.

Based on the above description of the point cloud feature point extraction method and the point cloud sensing system, an embodiment of the present invention provides a movable platform. The movable platform includes a fuselage, a power system, and the aforementioned point cloud sensing system; wherein the power system is installed on the fuselage of the movable platform to provide power for the movable platform. The point cloud sensing system includes a point cloud sensor, a processor and a memory. The point cloud sensor is directly carried on the fuselage of the movable platform, or is carried on the fuselage of the movable platform through the pan-tilt of the movable platform; the memory is used to store program instructions, and the processor is used to execute the program instructions stored in the memory. When the program instructions are executed, the processor is used to: obtain the three-dimensional point cloud of the environment in which the point cloud sensing system or the movable platform is located through the point cloud sensor, and project the obtained three-dimensional point cloud to the horizontal plane along the height direction to obtain Two-dimensional projection point cloud; rasterize the two-dimensional projection point cloud and perform feature information statistics to generate a point cloud information map, and determine the point cloud feature points based on the point cloud information map. For the specific implementation of the above steps, please refer to the foregoing description, which will not be repeated here.

An embodiment of the present invention also provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the point cloud feature point extraction method described in the above method embodiment is implemented .

The embodiment of the present invention also provides a computer program product containing instructions, which when run on a computer, causes the computer to execute the point cloud feature point extraction method described in the foregoing method embodiment.

It should be noted that for the foregoing method embodiments, for the sake of simple description, they are all expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited by the described sequence of actions. Because according to the present invention, certain steps can be performed in other order or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the involved actions and modules are not necessarily required by the present invention.

The steps in the method of the embodiment of the present invention can be adjusted, merged, and deleted in order according to actual needs.

The modules in the device of the embodiment of the present invention can be combined, divided, and deleted according to actual needs.

A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above-mentioned embodiments can be completed by a program instructing relevant hardware. The program can be stored in a computer-readable storage medium, and the storage medium can include: Flash disk, read-only memory (Read-Only Memory, ROM), random access device (Random Access Memory, RAM), magnetic disk or optical disk, etc.

The above describes in detail a point cloud feature point extraction method, a point cloud sensing system, and a movable platform provided by the embodiments of the present invention. In this article, specific examples are used to explain the principles and implementation of the present invention. The description of the embodiments is only used to help understand the method and core idea of the present invention; at the same time, for those of ordinary skill in the art, according to the idea of the present invention, there will be changes in the specific implementation and the scope of application. As mentioned above, the contents of this specification should not be construed as limiting the present invention.

Claims (34)

1. A point cloud feature point extraction method, characterized in that, the method includes:

   Obtain the 3D point cloud of the environment where the mobile platform is located;

   Projecting the three-dimensional point cloud onto a horizontal plane along the height direction to obtain a two-dimensional projection point cloud;

   Performing rasterization processing and feature information statistics on the two-dimensional projection point cloud to generate a point cloud information map;

   Determine point cloud feature points based on the point cloud information map.
2. The method according to claim 1, wherein the three-dimensional point cloud includes one or more of a complete three-dimensional point cloud, a ground three-dimensional point cloud, and a non-ground three-dimensional point cloud.
3. The method according to claim 1 or 2, wherein the three-dimensional point cloud comprises a complete three-dimensional point cloud, and the two-dimensional projection point cloud is rasterized and feature information statistics are performed to generate point cloud information Figures, including:

   Perform rasterization processing and height information statistics on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud, generate a point cloud height map, and use the point cloud height map as a point cloud information map.
4. The method according to claim 3, wherein the point cloud height map comprises a point cloud height average map, and the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud is rasterized and height information Statistics, generate point cloud height map, including:

   Performing rasterization processing on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids;

   Calculate the height average of the height values corresponding to the two-dimensional projection point cloud in each grid;

   Each grid is represented by grid coordinates and a height average value, and the point cloud height average value map is generated.
5. The method according to claim 4, wherein the calculating the height average value of the height values corresponding to the two-dimensional projection point clouds in each grid comprises:

   Obtain a target two-dimensional projection point cloud with a corresponding height value within a predetermined number of standard deviations within the target grid, where the target grid is any one of the multiple grids, and the standard deviation is based on The height values corresponding to all the two-dimensional projection points in the target grid are determined;

   Calculate the height average value of the height values corresponding to the two-dimensional projection point cloud of the target in the target grid.
6. The method according to claim 3, wherein the point cloud height map comprises a point cloud height variance map, and the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud is rasterized and height information Statistics, generate point cloud height map, including:

   Performing rasterization processing on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids;

   Calculate the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid;

   Each grid is represented by grid coordinates and a height variance value to generate the point cloud height variance map.
7. The method according to claim 6, wherein the calculating the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid comprises:

   Obtain a target two-dimensional projection point cloud with a corresponding height value within a predetermined number of standard deviations within the target grid, where the target grid is any one of the multiple grids, and the standard deviation is based on The height values corresponding to all the two-dimensional projection points in the target grid are determined;

   Calculate the height variance value of the height value corresponding to the two-dimensional projection point cloud of the target in the target grid.
8. The method according to claim 1 or 2, wherein the three-dimensional point cloud includes a ground three-dimensional point cloud, and the two-dimensional projection point cloud is rasterized and feature information statistics are performed to generate point cloud information Figures, including:

   Perform rasterization processing and reflectivity information statistics on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to generate a point cloud reflectivity map, and use the point cloud reflectivity map as a point cloud information map.
9. The method according to claim 8, wherein the point cloud reflectivity map comprises a point cloud maximum reflectivity map, the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud is rasterized, and Reflectance information statistics, generate point cloud reflectance map, including:

   Performing rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids;

   Calculate the maximum reflectance value of the reflectance value corresponding to the two-dimensional projection point cloud in each grid;

   Each grid is represented by grid coordinates and a maximum reflectance value, and the maximum reflectance map of the point cloud is generated.
10. The method according to claim 8, wherein the point cloud reflectance map comprises a point cloud reflectance variance map, the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud is rasterized, and Reflectance information statistics, generate point cloud reflectance map, including:

    Performing rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids;

    Calculate the reflectance variance value of the reflectance value corresponding to the two-dimensional projection point cloud in each grid;

    Each grid is represented by grid coordinates and a reflectance variance value, and the point cloud reflectivity variance map is generated.
11. The method according to claim 4 or 5, wherein the determining point cloud feature points based on the point cloud information graph comprises:

    Determine the height jump feature points based on the point cloud height average map.
12. The method according to claim 11, wherein the determining a height jump feature point based on the point cloud height average map comprises:

    Acquiring an adjacent grid adjacent to a first grid in the mean point cloud height map, where the first grid is any grid in the mean point cloud height map;

    Determine a second grid from the adjacent grids, and the difference between the average height of the two-dimensional projection point cloud in the second grid and the average height of the two-dimensional projection point cloud in the first grid The absolute value of the difference is greater than or equal to the first value;

    If the number of the second grid is greater than or equal to the second value, the two-dimensional projection point in the first grid is determined as the height jump feature point.
13. The method according to claim 8, wherein the point cloud reflectivity map comprises a point cloud maximum reflectivity map and a point cloud reflectivity variance map, and the point cloud feature points are determined based on the point cloud information map, include:

    Determine line-following feature points based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map.
14. The method according to claim 13, wherein the determining the line-following feature points based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map comprises:

    Performing area division on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple point cloud areas;

    Determining a two-dimensional Gaussian mixture model of each of the multiple point cloud areas based on the point cloud maximum reflectance map and the point cloud reflectance variance map;

    Based on the two-dimensional Gaussian mixture model, a line-following feature point is determined from the multiple point cloud regions.
15. The method according to any one of claims 1 to 14, wherein after the point cloud feature points are determined based on the point cloud information graph, the method further comprises:

    The position of the movable platform is determined in a high-precision map based on the point cloud feature points.
16. The method according to claim 15, wherein the determining the position of the movable platform in a high-precision map based on the feature points of the point cloud comprises:

    Converting the point cloud feature points to polar coordinates, and dividing the point cloud feature points in polar coordinates proportionally based on angle and radial length;

    The position of the movable platform is determined on the high-precision map based on the feature points of the point cloud divided in proportion.
17. A point cloud sensing system, characterized by comprising: a point cloud sensor, a memory and a processor, the memory being used for storing program instructions;

    The processor is configured to execute program instructions stored in the memory, and when the program instructions are executed, the processor is configured to:

    Acquiring a three-dimensional point cloud of the environment in which the point cloud sensing system is located through the point cloud sensor;

    Projecting the three-dimensional point cloud onto a horizontal plane along the height direction to obtain a two-dimensional projection point cloud;

    Performing rasterization processing and feature information statistics on the two-dimensional projection point cloud to generate a point cloud information map;

    Determine point cloud feature points based on the point cloud information map.
18. The point cloud sensing system according to claim 17, wherein the three-dimensional point cloud includes one or more of a complete three-dimensional point cloud, a ground three-dimensional point cloud, and a non-ground three-dimensional point cloud.
19. The point cloud sensing system according to claim 17 or 18, wherein the three-dimensional point cloud comprises a complete three-dimensional point cloud, and the processor performs rasterization processing and characteristic information on the two-dimensional projection point cloud Statistics, when generating a point cloud information graph, are specifically used for:

    Perform rasterization processing and height information statistics on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud, generate a point cloud height map, and use the point cloud height map as a point cloud information map.
20. The point cloud sensing system according to claim 19, wherein the point cloud height map comprises a point cloud height average map, and the processor rasterizes the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud. Grid processing and height information statistics, when generating a point cloud height map, are specifically used for:

    Performing rasterization processing on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids;

    Calculate the height average of the height values corresponding to the two-dimensional projection point cloud in each grid;

    Each grid is represented by grid coordinates and a height average value, and the point cloud height average value map is generated.
21. The point cloud sensing system according to claim 20, wherein when the processor calculates the height average value of the height values corresponding to the two-dimensional projection point clouds in each grid, it is specifically used for:

    Obtain a target two-dimensional projection point cloud with a corresponding height value within a predetermined number of standard deviations within the target grid, where the target grid is any one of the multiple grids, and the standard deviation is based on The height values corresponding to all the two-dimensional projection points in the target grid are determined;

    Calculate the height average value of the height values corresponding to the two-dimensional projection point cloud of the target in the target grid.
22. The point cloud sensing system according to claim 19, wherein the point cloud height map comprises a point cloud height variance map, and the processor rasterizes the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud. Grid processing and height information statistics, when generating a point cloud height map, are specifically used for:

    Performing rasterization processing on the two-dimensional projection point cloud corresponding to the complete three-dimensional point cloud to obtain multiple grids;

    Calculate the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid;

    Each grid is represented by grid coordinates and a height variance value to generate the point cloud height variance map.
23. The point cloud sensing system according to claim 22, wherein when the processor calculates the height variance value of the height value corresponding to the two-dimensional projection point cloud in each grid, it is specifically used for:

    Obtain a target two-dimensional projection point cloud with a corresponding height value within a predetermined number of standard deviations within the target grid, where the target grid is any one of the multiple grids, and the standard deviation is based on The height values corresponding to all the two-dimensional projection points in the target grid are determined;

    Calculate the height variance value of the height value corresponding to the two-dimensional projection point cloud of the target in the target grid.
24. The point cloud sensing system according to claim 17 or 18, wherein the three-dimensional point cloud comprises a ground three-dimensional point cloud, and the processor performs rasterization processing and characteristic information on the two-dimensional projection point cloud Statistics, when generating a point cloud information graph, are specifically used for:

    Perform rasterization processing and reflectivity information statistics on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to generate a point cloud reflectivity map, and use the point cloud reflectivity map as a point cloud information map.
25. The point cloud sensing system according to claim 24, wherein the point cloud reflectance map comprises a point cloud maximum reflectance map, and the processor performs a two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud When performing rasterization processing and reflectivity information statistics, when generating a point cloud reflectivity map, it is specifically used for:

    Performing rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids;

    Calculate the maximum reflectance value of the reflectance value corresponding to the two-dimensional projection point cloud in each grid;

    Each grid is represented by grid coordinates and a maximum reflectance value, and the maximum reflectance map of the point cloud is generated.
26. The point cloud sensing system according to claim 24, wherein the point cloud reflectivity map comprises a point cloud reflectivity variance map, and the processor performs a calculation of the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud When performing rasterization processing and reflectivity information statistics, when generating a point cloud reflectivity map, it is specifically used for:

    Performing rasterization processing on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple grids;

    Calculate the reflectance variance value of the reflectance value corresponding to the two-dimensional projection point cloud in each grid;

    Each grid is represented by grid coordinates and a reflectance variance value, and the point cloud reflectivity variance map is generated.
27. The point cloud sensing system according to claim 20 or 21, wherein when the processor determines point cloud feature points based on the point cloud information graph, it is specifically configured to:

    Determine the height jump feature points based on the point cloud height average map.
28. The point cloud sensing system according to claim 27, wherein when the processor determines the height jump feature points based on the point cloud height average map, it is specifically configured to:

    Acquiring an adjacent grid adjacent to a first grid in the mean point cloud height map, where the first grid is any grid in the mean point cloud height map;

    Determine a second grid from the adjacent grids, and the difference between the average height of the two-dimensional projection point cloud in the second grid and the average height of the two-dimensional projection point cloud in the first grid The absolute value of the difference is greater than or equal to the first value;

    If the number of the second grid is greater than or equal to the second value, the two-dimensional projection point in the first grid is determined as the height jump feature point.
29. The point cloud sensing system according to claim 24, wherein the point cloud reflectivity map comprises a point cloud maximum reflectivity map and a point cloud reflectivity variance map, and the processor is based on the point cloud information map When determining point cloud feature points, they are specifically used to:

    Determine line-following feature points based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map.
30. The point cloud sensing system according to claim 29, wherein when the processor determines the line-following feature points based on the point cloud maximum reflectivity map and the point cloud reflectivity variance map, it is specifically used for:

    Performing area division on the two-dimensional projection point cloud corresponding to the ground three-dimensional point cloud to obtain multiple point cloud areas;

    Determining a two-dimensional Gaussian mixture model of each of the multiple point cloud areas based on the point cloud maximum reflectance map and the point cloud reflectance variance map;

    Based on the two-dimensional Gaussian mixture model, a line-following feature point is determined from the multiple point cloud regions.
31. The point cloud sensing system according to any one of claims 17 to 30, wherein after the processor determines the point cloud feature points based on the point cloud information graph, it is further configured to:

    The position of the point cloud sensing system is determined in a high-precision map based on the point cloud feature points.
32. The point cloud sensing system according to claim 31, wherein the processor is specifically configured to: when determining the position of the point cloud sensing system in a high-precision map based on the point cloud feature points:

    Converting the point cloud feature points to polar coordinates, and dividing the point cloud feature points in polar coordinates proportionally based on angle and radial length;

    The position of the point cloud sensing system is determined on the high-precision map based on the point cloud feature points divided in proportion.
33. A movable platform, characterized in that it comprises:

    body;

    A power system installed on the fuselage and used to provide power for the movable platform;

    The point cloud sensing system according to any one of claims 17 to 32.
34. A computer-readable storage medium in which a computer program is stored, characterized in that: when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 16 are implemented .

Images