CN111742242A

CN111742242A - Point cloud processing method, system, device and storage medium

Info

Publication number: CN111742242A
Application number: CN201980012219.4A
Authority: CN
Inventors: 刘寒颖; 李星河; 邱凡
Original assignee: SZ DJI Technology Co Ltd
Current assignee: SZ DJI Technology Co Ltd
Priority date: 2019-06-11
Filing date: 2019-06-11
Publication date: 2020-10-02
Also published as: WO2020248118A1

Abstract

A point cloud processing method, system, device and storage medium. The method comprises the steps of acquiring a three-dimensional point cloud obtained by detecting objects around a movable platform through a detection device, and determining a point cloud-free three-dimensional space corresponding to a preset two-dimensional area according to the preset two-dimensional area, wherein the height of the three-dimensional point cloud corresponding to an air obstacle is larger than that of a ground point cloud or a ground obstacle point cloud, and the point cloud-free three-dimensional space may be between the air obstacle and the ground, or between the air obstacle and the ground obstacle, or between the air obstacle and the movable platform, so that the point cloud of the air obstacle in the three-dimensional point cloud is determined according to the point cloud-free three-dimensional space, and the detection precision of the air obstacle is improved. In addition, the air obstacle is detected through the point-free cloud three-dimensional space, and the air obstacle can be accurately detected even on an ascending slope and a descending slope without being influenced by terrain because the air obstacle is not dependent on a fixed absolute height threshold value.

Description

Point cloud processing method, system, device and storage medium

Technical Field

The embodiment of the invention relates to the field of radar detection, in particular to a point cloud processing method, a point cloud processing system, point cloud processing equipment and a storage medium.

Background

Movable platform among the prior art, for example unmanned aerial vehicle, mobile robot, vehicle etc. are provided with detection equipment, and this detection equipment can survey movable platform obstacle around.

For example, the detection device may include a laser radar that emits a laser beam that, when scanned along a trajectory, results in a large number of laser points, i.e., a three-dimensional point cloud. Because the three-dimensional point cloud has height information, the laser radar can also detect air obstacles such as three-dimensional point cloud of branches, eaves, viaducts, tunnels, ceilings of underground garages and the like.

However, the prior art has low detection precision for the air obstacle, so that the ground under the air obstacle is determined as the ground obstacle, and the ground under the air obstacle is actually a feasible area, or the air obstacle cannot be accurately determined in a terrain such as an up slope and a down slope.

Disclosure of Invention

The embodiment of the invention provides a point cloud processing method, a point cloud processing system, a point cloud processing device and a storage medium, which are used for improving the detection precision of an air obstacle and accurately detecting the air obstacle even on an ascending slope and a descending slope.

A first aspect of an embodiment of the present invention provides a point cloud processing method, which is applied to a movable platform, where the movable platform is provided with a detection device, and the detection device is configured to detect an object around the movable platform to obtain a three-dimensional point cloud, where the method includes:

acquiring the three-dimensional point cloud;

determining a point cloud-free three-dimensional space corresponding to a preset two-dimensional area according to the preset two-dimensional area;

and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the point cloud-free three-dimensional space.

A second aspect of an embodiment of the present invention provides a point cloud processing system, including: a detection device, a memory, and a processor;

the detection equipment is used for detecting objects around the movable platform to obtain three-dimensional point cloud;

the memory is used for storing program codes;

the processor, invoking the program code, when executed, is configured to:

acquiring the three-dimensional point cloud;

A third aspect of an embodiment of the present invention is to provide a movable platform, including:

a body;

the power system is arranged on the machine body and used for providing moving power;

and the point cloud processing system of the second aspect.

A fourth aspect of embodiments of the present invention is to provide a computer-readable storage medium, on which a computer program is stored, the computer program being executed by a processor to implement the method of the first aspect.

The point cloud processing method, system, device and storage medium provided by the embodiment can acquire the three-dimensional point cloud obtained by detecting the object around the movable platform by the detection device, and determining a point cloud-free three-dimensional space corresponding to the preset two-dimensional area according to the preset two-dimensional area, since the height of the three-dimensional point cloud corresponding to the air obstacle is greater than the height of the ground point cloud or ground obstacle point cloud, there may be a point cloud free three dimensional space between the airborne obstacle and the ground, or between the airborne obstacle and the ground obstacle, or between the airborne obstacle and the movable platform, and therefore, according to the point cloud-free three-dimensional space, the point cloud of the air obstacle in the three-dimensional point cloud is determined, and compared with the method for detecting the air obstacle through a height threshold value in the prior art, the detection precision of the air obstacle is improved, so that the ground below the air obstacle is prevented from being judged as the ground obstacle. In addition, the air obstacle is detected through the point-free cloud three-dimensional space, and the air obstacle can be accurately detected even on an ascending slope and a descending slope without being influenced by terrain because the air obstacle is not dependent on a fixed absolute height threshold value.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without inventive labor.

Fig. 1 is a schematic diagram of an application scenario provided in an embodiment of the present invention;

FIG. 2 is a flowchart of a point cloud processing method according to an embodiment of the present invention;

fig. 3 is a schematic diagram of another application scenario provided in the embodiment of the present invention;

FIG. 4 is a schematic diagram of a predetermined two-dimensional area according to an embodiment of the present invention;

FIG. 5 is a flowchart of a point cloud processing method according to another embodiment of the present invention;

FIG. 6 is a diagram of denoising effects according to another embodiment of the present invention;

FIG. 7 is a flowchart of a point cloud processing method according to another embodiment of the present invention;

FIG. 8 is a schematic diagram of a three-dimensional mesh provided in accordance with another embodiment of the present invention;

fig. 9 is a structural diagram of a point cloud processing system according to an embodiment of the present invention.

Reference numerals:

11: a vehicle; 12: a server; 31: a ground barrier;

32: an airborne obstacle; 33: a circular region; 34: noise points;

35: a noise information space; 36: no three-dimensional point cloud space; 37: no three-dimensional point cloud space;

40: a two-dimensional grid; 41: three-dimensional meshes; 42: three-dimensional meshes;

43: three-dimensional meshes; 44: three-dimensional meshes;

45: three-dimensional meshes; 46: three-dimensional meshes;

461: three-dimensional points; 411: three-dimensional points; 47: three-dimensional meshes;

48: three-dimensional meshes; 49: three-dimensional meshes; 471: three-dimensional points;

90: a point cloud processing system; 91: a detection device; 92: a memory;

93: a processor.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It will be understood that when an element is referred to as being "secured to" another element, it can be directly on the other element or intervening elements may also be present. When a component is referred to as being "connected" to another component, it can be directly connected to the other component or intervening components may also be present.

Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.

Some embodiments of the invention are described in detail below with reference to the accompanying drawings. The embodiments described below and the features of the embodiments can be combined with each other without conflict.

The embodiment of the invention provides a point cloud processing method. The point cloud processing method is applied to a movable platform, the movable platform is provided with a detection device, and the detection device is used for detecting objects around the movable platform to obtain three-dimensional point cloud. Optionally, the movable platform comprises: unmanned aerial vehicle, mobile robot or vehicle.

The movable platform is a vehicle, which may be an unmanned vehicle, a vehicle equipped with an Advanced Driver Assistance Systems (ADAS), or the like. It is understood that the point cloud processing method may also be applied to an unmanned aerial vehicle or a mobile robot, for example, an unmanned aerial vehicle or a mobile robot equipped with a detection device for acquiring point cloud data. As shown in fig. 1, the vehicle 11 is a carrier carrying a detection device, which may be a binocular stereo camera, a Time of flight (TOF) camera, and/or a lidar. In the driving process of the vehicle 11, the detection device detects objects around the vehicle 11 in real time to obtain a three-dimensional point cloud. Taking the laser radar as an example, when a laser beam emitted by the laser radar irradiates the surface of an object, the surface of the object reflects the laser beam, and the laser radar can determine information such as the direction and the distance of the object relative to the laser radar according to the laser beam reflected by the surface of the object. If the laser beam emitted by the laser radar scans according to a certain track, for example, 360-degree rotation scanning, a large number of laser points are obtained, and thus laser point cloud data, i.e., three-dimensional point cloud, of the object can be formed.

In addition, the present embodiment does not limit the execution subject of the point cloud processing method, and the point cloud processing method may be executed by a vehicle-mounted processor, or may be executed by a device having a data processing function other than the vehicle-mounted processor, for example, as the server 12 shown in fig. 1, the vehicle 11 and the server 12 may perform wireless communication or wired communication, the vehicle 11 may transmit the three-dimensional point cloud to the server 12, and the server 12 may execute the point cloud processing method. The point cloud processing method provided by the embodiment of the invention is described by taking a vehicle as an example.

Fig. 2 is a flowchart of a point cloud processing method according to an embodiment of the present invention. As shown in fig. 2, the method in this embodiment may include:

step S201, obtaining the three-dimensional point cloud.

As shown in fig. 1, in the driving process of the vehicle 11, a detection device mounted on the vehicle 11 detects objects around the vehicle 11 in real time to obtain a three-dimensional point cloud, and an on-board processor on the vehicle 11 obtains the three-dimensional point cloud in real time, optionally, the three-dimensional point cloud obtained by the on-board processor may be a three-dimensional point cloud obtained by superimposing multiple frames of three-dimensional point clouds detected by the detection device.

Step S202, determining a point cloud-free three-dimensional space corresponding to a preset two-dimensional area according to the preset two-dimensional area.

Optionally, the point cloud-free three-dimensional space includes at least one of the following: and a three-dimensional point cloud space and a noise information space are not generated.

As shown in fig. 3, since the detection device mounted on the vehicle can detect not only the ground and ground obstacles in front of the vehicle, for example, the ground obstacle 31, but also an obstacle above the vehicle, for example, the air obstacle 32. The overhead barrier 32 may be a branch, an eave, a viaduct, a tunnel roof, an underground garage ceiling, or the like. Therefore, the three-dimensional point cloud detected by the detection device may include: at least one of a ground point cloud, a ground obstacle point cloud, and an air obstacle point cloud. Wherein the various point clouds respectively represent or approximate the areas of ground, ground obstacles and air obstacles in real space.

In order to determine an air obstacle above the vehicle, a plane area within a preset range around the vehicle may be taken as a preset two-dimensional area, and the preset two-dimensional area may be parallel to the ground. For example, the predetermined two-dimensional area may be a circular area 33 centered on the vehicle and having a predetermined length (e.g., 60 meters) as a radius, as shown in fig. 3. The preset two-dimensional area is only schematically illustrated here, and the specific shape of the preset two-dimensional area is not limited, and in other embodiments, the preset two-dimensional area may also be a rectangular area.

As shown in fig. 3, the airborne obstacle 32 is above the circular area 33, and therefore, when the vehicle is located within the circular area 33, the airborne obstacle 32 will be detected by the detection equipment carried on the vehicle. In addition, in other embodiments, if some other tiny objects float in the air of the vehicle or due to the detection accuracy of the detection device itself, the detection device may also detect a noise point 34 as shown in fig. 3, where a large number of noise points 34 may constitute a noise point cloud, and the space occupied by the noise point cloud may be denoted as a noise information space, such as the noise information space 35 shown in fig. 3. In the present embodiment, the noise point cloud may be regarded as an invalid point cloud, that is, the noise point cloud is a useless point cloud and may be regarded as no point cloud. In addition, the point cloud-free three-dimensional space described in this embodiment specifically refers to a three-dimensional space without effective point clouds. Thus, the noise information space may be used as a case of a point cloud-free three-dimensional space.

Since the three-dimensional points in the three-dimensional point cloud have height information, the height of the three-dimensional point cloud corresponding to the air obstacle 32 is larger than the height of the ground point cloud or the ground obstacle point cloud. There may be no three-dimensional point cloud space between the airborne obstacle 32 and the ground, or between the airborne obstacle 32 and the ground obstacle, or between the airborne obstacle 32 and the vehicle roof. The three-dimensional point cloud free space refers to a space without any three-dimensional point cloud. The three-dimensional point cloud free space 36 shown in fig. 3 is devoid of any three-dimensional point cloud. Therefore, a point cloud free three-dimensional space may be another case of the point cloud free three-dimensional space. Further, the point cloud-free three-dimensional space may also include both a noise information space and a three-dimensional point cloud-free space, and for example, the noise information space 35 and the three-dimensional point cloud-free space 36 as shown in fig. 3 may be regarded as the point cloud-free three-dimensional space.

And S203, determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the point cloud-free three-dimensional space.

If the point cloud free three-dimensional space includes only a three-dimensional point cloud free space, such as the three-dimensional point cloud free space 37 shown in fig. 3, the on-board processor may determine an overhead obstacle point cloud in the three-dimensional point cloud based on the three-dimensional point cloud free space, e.g., the three-dimensional point cloud above the three-dimensional point cloud free space 37 is the overhead obstacle point cloud.

If the point cloud-free three-dimensional space further includes a noise information space, such as the noise information space 35 and the three-dimensional point cloud-free space 36 shown in fig. 3, the on-board processor may determine the air obstacle point cloud in the three-dimensional point cloud according to the noise information space 35 and the three-dimensional point cloud-free space 36, for example, perform denoising processing on the noise information space 35 in the point cloud-free three-dimensional space, so that the three-dimensional point cloud-free space 36 in the point cloud-free three-dimensional space becomes the three-dimensional point cloud-free space 37 after denoising processing, and further, determine the air obstacle point cloud in the three-dimensional point cloud according to the three-dimensional point cloud-free space 37.

In the embodiment, the detection device is used for detecting the three-dimensional point cloud obtained by detecting objects around the movable platform, and the point cloud-free three-dimensional space corresponding to the preset two-dimensional area is determined according to the preset two-dimensional area, because the height of the three-dimensional point cloud corresponding to the air obstacle is larger than that of the ground point cloud or the ground obstacle point cloud, and the point cloud-free three-dimensional space may be formed between the air obstacle and the ground, or between the air obstacle and the ground obstacle, or between the air obstacle and the movable platform, the point cloud of the air obstacle in the three-dimensional point cloud is determined according to the point cloud-free three-dimensional space. In addition, the air obstacle is detected through the point-free cloud three-dimensional space, and the air obstacle can be accurately detected even on an ascending slope and a descending slope without being influenced by terrain because the air obstacle is not dependent on a fixed absolute height threshold value.

On the basis of the foregoing embodiment, assuming that the point cloud-free three-dimensional space further includes a noise information space, optionally, determining an aerial obstacle point cloud in the three-dimensional point cloud according to the point cloud-free three-dimensional space includes: removing the noise information space in the three-dimensional point cloud to obtain a denoised three-dimensional point cloud; determining a three-dimensional point cloud-free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area; and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the space without the three-dimensional point cloud corresponding to the preset two-dimensional area.

For example, after the on-board processor acquires the three-dimensional point cloud detected by the detection device, the noise information space in the three-dimensional point cloud is removed, and the three-dimensional point cloud after being denoised is obtained. In one possible implementation, the removing the noise information space in the three-dimensional point cloud includes: and removing the noise information space in the three-dimensional point cloud by adopting a K nearest neighbor algorithm.

Optionally, the removing the noise information space in the three-dimensional point cloud by using a K nearest neighbor algorithm includes: traversing each three-dimensional point in the three-dimensional point cloud, and determining K neighborhoods closest to the three-dimensional point; calculating the average distance from the three-dimensional point to the K neighborhoods; and if the average distance is larger than or equal to a preset distance threshold value, determining the three-dimensional point as a noise point, and removing the noise point.

For example, the three-dimensional point cloud acquired by the onboard processor includes a large number of three-dimensional points, and when the onboard processor removes the noise point cloud in the three-dimensional point cloud by using a K-nearest neighbor algorithm, a K-d tree is firstly generated according to the three-dimensional point cloud, and a topological relation of the three-dimensional point cloud is established. Secondly, traversing each three-dimensional point in the three-dimensional point cloud, and calculating K neighborhoods nearest to the three-dimensional point. Further, the average distance from the three-dimensional point to the K neighborhoods is calculated. And if the average distance is greater than or equal to a preset distance threshold value, determining the three-dimensional point as a noise point, removing the noise point from the three-dimensional point cloud, and traversing the next three-dimensional point. And if the average distance is smaller than the preset distance threshold, determining that the three-dimensional point is not a noise point, reserving the three-dimensional point and traversing the next three-dimensional point. And repeating the steps until all three-dimensional points in the three-dimensional point cloud are traversed, so that the noise point cloud in the three-dimensional point cloud is removed, and the denoised three-dimensional point cloud is obtained.

For example, as shown in fig. 3, when the three-dimensional point cloud detected by the detection device includes the noise information space 35, the three-dimensional point cloud after denoising can no longer include the noise information space 35 after denoising processing as described above. And the denoised three-dimensional point cloud may include an aerial obstacle point cloud above the vehicle.

In addition, after the denoising process as described above, the three-dimensional-free point cloud space above the circular region 33 is changed from the three-dimensional-free point cloud space 36 to the three-dimensional-free point cloud space 37.

The onboard processor may further determine an airborne obstacle point cloud above the circular area 33 from the three-dimensional point cloud free space 37 above the circular area 33, e.g., the three-dimensional point cloud above the three-dimensional point cloud free space 37 is the airborne obstacle point cloud.

In the embodiment, the denoised three-dimensional point cloud is obtained by acquiring the three-dimensional point cloud and removing the noise point cloud in the three-dimensional point cloud, determining a three-dimensional point cloud free space corresponding to a preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area, since the height of the three-dimensional point cloud corresponding to the air obstacle is greater than the height of the ground point cloud or ground obstacle point cloud, there may be no three-dimensional point cloud space between the airborne obstacle and the ground, or between the airborne obstacle and the ground obstacle, or between the airborne obstacle and the movable platform, and therefore, compared with the prior art that the air obstacles are detected by a height threshold value method, the detection precision of the air obstacles is improved, and therefore the situation that the ground below the air obstacles is judged as the ground obstacles is avoided. In addition, the influence of the noise point cloud on the detection of the air obstacle is reduced by removing the noise point cloud in the three-dimensional point cloud, and the detection precision of the air obstacle is further improved. In addition, the air obstacle is detected through the three-dimensional point cloud-free space, and the air obstacle can be accurately detected even on an ascending slope and a descending slope without depending on a fixed absolute height threshold value.

The embodiment of the invention provides a point cloud processing method. On the basis of the above embodiment, the preset two-dimensional area includes a plurality of two-dimensional grids.

As shown in fig. 4, the preset two-dimensional region is a rectangular region OABC, and the preset two-dimensional region includes a plurality of two-dimensional grids, for example, 40 denotes any one of the two-dimensional grids. Here, the description is only illustrative, and the specific shape of the preset two-dimensional region is not limited, and a method of dividing the preset two-dimensional region into a plurality of two-dimensional meshes is not limited. Alternatively, the size of each two-dimensional grid may be a preset size.

Before determining a three-dimensional point cloud free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area, the method further comprises: dividing a preset three-dimensional space area around the movable platform into a plurality of three-dimensional grids, wherein each three-dimensional grid in the plurality of three-dimensional grids is projected into one two-dimensional grid in the preset two-dimensional area.

As shown in fig. 4, the preset three-dimensional space may be a three-dimensional space in a three-dimensional coordinate system defined by XYZ as three-dimensional coordinate axes with the position of the vehicle as the origin O, and since the position of the vehicle is changed in real time, the origin O of the three-dimensional coordinate system is also changed in real time, that is, the three-dimensional coordinate system is changed in real time. Additionally, in other embodiments, the three-dimensional coordinate system may be an absolute coordinate system that does not change as the vehicle location changes, such as a world coordinate system.

As shown in fig. 4, the preset three-dimensional space around the vehicle may be a three-dimensional space above a preset two-dimensional area OABC, the three-dimensional point cloud corresponding to the preset two-dimensional area OABC may be a three-dimensional point cloud in the three-dimensional space above the preset two-dimensional area OABC, before determining a three-dimensional point cloud free space corresponding to the preset two-dimensional area OABC according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area OABC, the three-dimensional space above the preset two-dimensional area OABC may be divided into a plurality of three-dimensional grids, each three-dimensional grid has the same size, and each three-dimensional grid in the plurality of three-dimensional grids may be projected as a two-dimensional grid in the preset two-dimensional area OABC. Taking two-dimensional grid 40 as an example, the three-dimensional space above two-dimensional grid 40 is divided into three-dimensional grids 41-46, and all three-dimensional grids 41-46 may be projected as two-dimensional grid 40 within the predetermined two-dimensional area OABC. That is, two-dimensional grid 40 corresponds to three-dimensional grids 41-46. The three-dimensional point cloud corresponding to the two-dimensional grid 40 is a three-dimensional point cloud included in the three-dimensional grids 41-46. Similarly, the three-dimensional space above the other two-dimensional grids may also be divided into a plurality of three-dimensional grids, and the plurality of three-dimensional grids corresponding to each two-dimensional grid in the preset two-dimensional area OABC may form a plurality of three-dimensional grids included in the three-dimensional space above the preset two-dimensional area OABC.

As shown in fig. 4, after the three-dimensional space above the preset two-dimensional area OABC is divided into a plurality of three-dimensional grids, the three-dimensional grid to which each three-dimensional point in the three-dimensional point cloud corresponding to the preset two-dimensional area OABC belongs may be determined. For example, an index of each three-dimensional mesh may be established, and a three-dimensional mesh into which each three-dimensional point falls may be determined, taking one of the three-dimensional meshes as an example, when a three-dimensional point falls into the three-dimensional mesh, the number of the three-dimensional points in the three-dimensional mesh is counted plus 1, and the density of the three-dimensional point cloud in the three-dimensional mesh is correspondingly increased. Therefore, the number of the three-dimensional points included in each three-dimensional grid in the plurality of three-dimensional grids can be determined, and the number of the three-dimensional points included in each three-dimensional grid can be used for representing the three-dimensional point cloud density in each three-dimensional grid due to the fact that the three-dimensional grids are the same in size. Furthermore, a three-dimensional point cloud density histogram can be established according to the three-dimensional point cloud density in each three-dimensional grid. Meanwhile, counting the three-dimensional points with the maximum height and the three-dimensional points with the minimum height in the three-dimensional point cloud corresponding to each two-dimensional grid. As shown in fig. 4, the three-dimensional point with the highest degree in the three-dimensional point cloud corresponding to the two-dimensional grid 40 is a point 461, and the three-dimensional point with the lowest degree in the three-dimensional point cloud corresponding to the two-dimensional grid 40 is a point 411. Or, counting the three-dimensional grids to which the three-dimensional points with the maximum height belong and the three-dimensional grids to which the three-dimensional points with the minimum height belong in the three-dimensional point cloud corresponding to each two-dimensional grid.

In another possible implementation manner, the removing the noise information space in the three-dimensional point cloud includes: and removing the noise information space in the three-dimensional point cloud according to the three-dimensional point cloud density in each three-dimensional grid in the plurality of three-dimensional grids.

As shown in fig. 4, according to the number of three-dimensional points included in each of the plurality of three-dimensional grids included in the three-dimensional space above the preset two-dimensional area OABC, the three-dimensional point cloud density in each of the three-dimensional grids can be determined, and further according to the three-dimensional point cloud density in each of the three-dimensional grids, the noise point cloud in the three-dimensional space above the preset two-dimensional area OABC can be removed.

Optionally, the removing the noise information space in the three-dimensional point cloud according to the three-dimensional point cloud density in each of the plurality of three-dimensional grids includes the following steps as shown in fig. 5:

step S501, determining whether the three-dimensional point cloud in each three-dimensional grid is noise point cloud according to the three-dimensional point cloud density in each three-dimensional grid in the plurality of three-dimensional grids and the three-dimensional point cloud density in the adjacent grid of each three-dimensional grid.

As shown in fig. 4, taking the three-dimensional grid 41 as an example, according to the density of the three-dimensional point cloud in the three-dimensional grid 41 and the density of the three-dimensional point cloud in the adjacent grid of the three-dimensional grid 41, it can be determined whether the three-dimensional point cloud in the three-dimensional grid 41 is a noise point cloud.

Wherein adjacent meshes of the three-dimensional mesh include at least one of: a three-dimensional grid above the three-dimensional grid, a three-dimensional grid below the three-dimensional grid, a three-dimensional grid in front of the three-dimensional grid, a three-dimensional grid behind the three-dimensional grid, a three-dimensional grid on the left side of the three-dimensional grid, and a three-dimensional grid on the right side of the three-dimensional grid.

For example, the neighboring meshes of the three-dimensional mesh 41 may include 6 neighboring meshes of the upper, lower, left, right, front, and rear of the three-dimensional mesh 41.

Optionally, the determining whether the three-dimensional point cloud in each three-dimensional mesh is a noise point cloud according to the three-dimensional point cloud density in each three-dimensional mesh in the plurality of three-dimensional meshes and the three-dimensional point cloud density in the adjacent mesh of each three-dimensional mesh includes: and if the density of the three-dimensional point cloud in the three-dimensional grid is smaller than a first density threshold value and the sum of the densities of the three-dimensional point clouds in the adjacent grids of the three-dimensional grid is smaller than a second density threshold value, determining that the three-dimensional point cloud in the three-dimensional grid is the noise point cloud.

For example, let the density of the three-dimensional point cloud in the three-dimensional mesh 41 be d, and the sum of the densities of the three-dimensional point clouds in the adjacent meshes of the three-dimensional mesh 41 be d _ sum, and when d is smaller than a first density threshold and d _ sum is smaller than a second density threshold, then it is determined that the three-dimensional point cloud falling into the three-dimensional mesh 41 is a noise point cloud. The method for determining whether the three-dimensional point cloud in other three-dimensional grids is the noise point cloud is similar to this, and is not repeated here.

The first density threshold is recorded as threshold1, the second density threshold is recorded as threshold2, and in general, the threshold2 is smaller than the threshold1, so that the detected noise point cloud is an isolated point in the three-dimensional space. That is, it is determined that a three-dimensional point cloud in a certain three-dimensional grid is a noise point cloud, and two conditions need to be satisfied: one condition is that the three-dimensional point cloud density in the three-dimensional mesh is sufficiently small, and the other condition is that the sum of the three-dimensional point cloud densities in neighboring meshes of the three-dimensional mesh is sufficiently small.

Step S502, if the three-dimensional point cloud in the three-dimensional grid is the noise point cloud, removing the three-dimensional point cloud in the three-dimensional grid.

For example, if it is determined that the three-dimensional point cloud falling in the three-dimensional mesh 41 is a noise point cloud, the three-dimensional point cloud in the three-dimensional mesh 41 is removed. Compared with the method adopting the K nearest neighbor algorithm, the method has the advantages that the noise point cloud in the three-dimensional point cloud is removed, a K-d tree does not need to be generated, the topological relation of the three-dimensional point cloud is established, each three-dimensional point in the three-dimensional point cloud does not need to be traversed, K neighborhoods nearest to the three-dimensional point are calculated, and the calculation amount and the calculation complexity are reduced.

Fig. 6 shows a denoising effect graph provided in this embodiment, where the white point cloud is noise. The denoising method may adopt the method described in this embodiment, or may also adopt a K-nearest neighbor algorithm.

In addition, in other embodiments, the vehicle-mounted processor may specifically be a Graphics Processing Unit (GPU), and the GPU may read the three-dimensional point cloud density in each three-dimensional grid in parallel, and count the sum of the three-dimensional point cloud densities in adjacent grids of each three-dimensional grid, thereby improving the real-time performance of removing noise point cloud.

It can be understood that, after the noise point cloud in the three-dimensional point cloud is removed by using the method described in this embodiment or using the K nearest neighbor algorithm as described above, the number of three-dimensional points in a part of three-dimensional meshes may be reduced, that is, the density of the three-dimensional point cloud in a part of three-dimensional meshes may be reduced, and further, a three-dimensional point cloud density histogram may be established according to the density of the three-dimensional point cloud in each three-dimensional mesh after denoising.

Determining a three-dimensional point cloud free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area, wherein the determining comprises the following steps: and determining a three-dimensional point cloud free space corresponding to each two-dimensional grid according to the denoised three-dimensional point cloud corresponding to each two-dimensional grid and the three-dimensional grid corresponding to each two-dimensional grid.

As shown in fig. 4, when determining a three-dimensional-point-cloud-free space corresponding to a preset two-dimensional area OABC according to a denoised three-dimensional point cloud corresponding to the preset two-dimensional area OABC, the three-dimensional-point-cloud-free space corresponding to each two-dimensional grid may be determined specifically according to the denoised three-dimensional point cloud corresponding to each two-dimensional grid in the preset two-dimensional area OABC and the three-dimensional grid corresponding to each two-dimensional grid. A set formed by the three-dimensional-free point cloud space corresponding to each two-dimensional grid can be used as the three-dimensional-free point cloud space corresponding to the preset two-dimensional area OABC.

For example, according to the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 and the three-dimensional grid corresponding to the two-dimensional grid 40, the three-dimensional point cloud free space corresponding to the two-dimensional grid 40 is determined. The denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 may be specifically a denoised three-dimensional point cloud included in the three-dimensional grids 41 to 46 corresponding to the two-dimensional grid 40.

Correspondingly, the determining the aerial obstacle point cloud in the three-dimensional point cloud according to the three-dimensional point cloud free space corresponding to the preset two-dimensional area comprises: and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the space without the three-dimensional point cloud corresponding to each two-dimensional grid.

For example, when determining the aerial obstacle point cloud in the three-dimensional point cloud corresponding to the preset two-dimensional area OABC according to the three-dimensional point cloud free space corresponding to the preset two-dimensional area OABC, the aerial obstacle point cloud in the three-dimensional point cloud corresponding to each two-dimensional grid may be determined specifically according to the three-dimensional point cloud free space corresponding to each two-dimensional grid in the preset two-dimensional area OABC, and a set formed by the aerial obstacle point clouds in the three-dimensional point cloud corresponding to each two-dimensional grid may be used as the aerial obstacle point cloud in the three-dimensional point cloud corresponding to the preset two-dimensional area OABC.

Optionally, after determining the point cloud of the airborne obstacle in the three-dimensional point cloud, the method further includes: and correcting the category of the three-dimensional point cloud in each three-dimensional grid in the plurality of three-dimensional grids according to a region continuity principle.

Optionally, the categories of the three-dimensional point cloud include: ground point clouds, ground obstacle point clouds, or air obstacle point clouds.

For example, when it is determined that the three-dimensional point cloud in the three-dimensional mesh 46 is an obstacle point cloud in the air, the categories of the three-dimensional point clouds in the neighboring meshes of the three-dimensional mesh 46 may be further counted, and the categories of the three-dimensional point cloud in the three-dimensional mesh 46 may be corrected according to the categories of the three-dimensional point cloud in the neighboring meshes of the three-dimensional mesh 46. The neighboring meshes of the three-dimensional mesh 46 may specifically include 6 neighboring meshes of the three-dimensional mesh 46, i.e., upper, lower, left, right, front, and rear meshes, and if the three-dimensional point clouds in most of the 6 neighboring meshes are ground obstacle point clouds, the three-dimensional point clouds in the three-dimensional mesh 46 are modified into ground obstacle point clouds.

According to the embodiment, the category of the three-dimensional point cloud in each three-dimensional grid is corrected according to the category of the three-dimensional point cloud in the adjacent grid around each three-dimensional grid through the principle of regional continuity, so that the misjudgment probability of the category of the three-dimensional point cloud in each three-dimensional grid can be reduced.

The embodiment of the invention provides a point cloud processing method. Fig. 7 is a flowchart of a point cloud processing method according to another embodiment of the invention. As shown in fig. 7, on the basis of the foregoing embodiment, the determining, according to the denoised three-dimensional point cloud corresponding to each two-dimensional mesh and the three-dimensional mesh corresponding to each two-dimensional mesh, a three-dimensional point cloud free space corresponding to each two-dimensional mesh includes:

step S701, starting from the three-dimensional point with the minimum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grid, inquiring a first three-dimensional grid without the three-dimensional point cloud in the three-dimensional grid corresponding to the two-dimensional grid, and taking the first three-dimensional grid without the three-dimensional point cloud as a starting point of a space without the three-dimensional point cloud.

As shown in fig. 4, taking the two-dimensional grid 40 as an example, the three-dimensional point with the highest degree in the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 is a point 461, and the three-dimensional point with the lowest degree in the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 is a point 411.

Starting from the point 411 or starting from the three-dimensional grid 41 to which the point 411 belongs, the first three-dimensional grid without the three-dimensional point cloud in the three-dimensional grid corresponding to the two-dimensional grid 40 is queried, for example, the three-dimensional grid 42 is the first three-dimensional grid without the three-dimensional point cloud starting from the three-dimensional grid 41. In this case, the three-dimensional mesh 42 may be used as a starting point of the three-dimensional point cloud free space corresponding to the two-dimensional mesh 40.

Step S702, starting from the starting point, continuously inquiring a first three-dimensional grid with three-dimensional point cloud in three-dimensional grids corresponding to the two-dimensional grids, and taking a three-dimensional grid without three-dimensional point cloud before the first three-dimensional grid with three-dimensional point cloud as an end point of the space without three-dimensional point cloud.

Starting from the three-dimensional grid 42, the first three-dimensional grid with the three-dimensional point cloud in the three-dimensional grids corresponding to the two-dimensional grid 40 is continuously queried, as shown in fig. 4, the three-dimensional grid 46 is the first three-dimensional grid with the three-dimensional point cloud starting from the three-dimensional grid 42, and at this time, a three-dimensional grid without the three-dimensional point cloud before the three-dimensional grid 46, that is, the three-dimensional grid 45, may be used as an end point of the three-dimensional point cloud-free space corresponding to the two-dimensional. In this way, the three-dimensional point cloud free space is formed from the three-dimensional mesh 42 as the starting point of the three-dimensional point cloud free space to the three-dimensional mesh 45 as the ending point of the three-dimensional point cloud free space.

Step S703, if the three-dimensional point with the maximum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grid is not in the first three-dimensional grid with the three-dimensional point cloud, continuing to query the next three-dimensional point cloud free space from the first three-dimensional grid with the three-dimensional point cloud.

As shown in fig. 4, since the three-dimensional point 461 with the highest degree in the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 is already in the three-dimensional grid 46, which is the first three-dimensional grid with the three-dimensional point cloud, the three-dimensional point cloud free space formed from the three-dimensional grid 42 to the three-dimensional grid 45 can be used as the three-dimensional point cloud free space corresponding to the two-dimensional grid 40.

As shown in fig. 8, assuming that the three-dimensional point with the highest degree in the denoised three-dimensional point cloud corresponding to the two-dimensional mesh 40 is the point 471, since the point 471 is not in the three-dimensional mesh 46, the next space without the three-dimensional point cloud needs to be queried from the three-dimensional mesh 46, and the querying process is similar to the above method, and is not repeated here. For example, the next three-dimensional-free point cloud space is from three-dimensional grid 47 to three-dimensional grid 48.

Correspondingly, the determining the aerial obstacle point cloud in the three-dimensional point cloud according to the three-dimensional point cloud space corresponding to each two-dimensional grid comprises: and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the maximum three-dimensional-free point cloud space corresponding to each two-dimensional grid.

As shown in fig. 8, two-dimensional mesh 40 corresponds to two three-dimensional-free point cloud spaces, one from three-dimensional mesh 42 to three-dimensional mesh 45, and the other from three-dimensional mesh 47 to three-dimensional mesh 48. Specifically, a maximum three-dimensional-free point cloud space may be determined from the two three-dimensional-free point cloud spaces, for example, the three-dimensional-free point cloud space from three-dimensional grid 42 to three-dimensional grid 45 is the maximum, and based on the maximum three-dimensional-free point cloud space, the two-dimensional grid 40 is determined to correspond to the airborne obstacle point cloud in the three-dimensional point cloud.

Optionally, the determining the aerial obstacle point cloud in the three-dimensional point cloud according to the maximum three-dimensional-free point cloud space corresponding to each two-dimensional grid includes: and if the height of the maximum three-dimensional point cloud-free space corresponding to the two-dimensional grid is greater than or equal to a preset height threshold, determining the denoised three-dimensional point cloud in the three-dimensional grid above the end point of the maximum three-dimensional point cloud-free space as the air obstacle point cloud.

As shown in fig. 8, if the maximum three-dimensional-point-free space corresponding to the two-dimensional grid 40, that is, the height from the three-dimensional grid 42 to the three-dimensional grid 45, is greater than or equal to the preset height threshold, it is determined that the denoised three-dimensional point cloud in the three-dimensional grid above the three-dimensional grid 45, which is the end point of the maximum three-dimensional-point-free space, is an obstacle point cloud in the air. Similarly, the point cloud of the obstacle in the air in the denoised three-dimensional point cloud corresponding to other two-dimensional grids can be determined.

In the embodiment, the three-dimensional point cloud density features of each three-dimensional grid are determined through the three-dimensional point cloud density histogram, so that the three-dimensional point cloud in the three-dimensional grid is judged to be the noise point cloud or the air obstacle point cloud.

The embodiment of the invention provides a point cloud processing method. On the basis of the above embodiment, before determining the three-dimensional point cloud free space corresponding to the preset two-dimensional region according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional region, the method further includes: and determining whether the denoised three-dimensional point cloud corresponding to each two-dimensional grid comprises an obstacle point cloud or not according to the maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional grid.

As shown in fig. 8, before determining the three-dimensional point cloud-free space corresponding to the preset two-dimensional area OABC according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area OABC, that is, after establishing the three-dimensional point cloud density histogram according to the three-dimensional point cloud density in each denoised three-dimensional grid, the denoised three-dimensional point cloud corresponding to each two-dimensional grid in the preset two-dimensional area OABC may be read in parallel, the maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional grid in the preset two-dimensional area OABC is calculated, and whether the denoised three-dimensional point cloud corresponding to each two-dimensional grid includes the obstacle point cloud is determined according to the maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional grid in the preset. Taking the two-dimensional grid 40 as an example, assuming that the three-dimensional point with the maximum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 is point 471, and the three-dimensional point with the minimum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 is point 411, the maximum height difference of the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 is the height difference between the point 471 and the point 411. From the height difference between the point 471 and the point 411, it can be determined whether the denoised three-dimensional point cloud corresponding to the two-dimensional mesh 40 includes an obstacle point cloud.

Optionally, the determining, according to a maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional mesh, whether the denoised three-dimensional point cloud corresponding to each two-dimensional mesh includes an obstacle point cloud includes: and if the maximum height difference of the denoised three-dimensional point clouds corresponding to the two-dimensional grids is larger than or equal to a preset height threshold value, determining that the denoised three-dimensional point clouds corresponding to the two-dimensional grids comprise the obstacle point clouds.

For example, if the height difference between the point 471 and the point 411 is greater than or equal to the preset height threshold, it is determined that the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 includes an obstacle point cloud, which may be an air obstacle point cloud or a ground obstacle point cloud, and further, the denoised three-dimensional point cloud corresponding to the two-dimensional grid 40 may be detected according to the above-mentioned method for determining the air obstacle point cloud. The process of specifically detecting the point cloud of the obstacle in the air is not described here again.

The embodiment of the invention provides a point cloud processing system. Fig. 9 is a structural diagram of a point cloud processing system according to an embodiment of the present invention, and as shown in fig. 9, a point cloud processing system 90 includes: a detection device 91, a memory 92 and a processor 93. The point cloud processing system 90 may be a separate device, such as a driver assistance device mounted on the vehicle; it may also be a distributed system, such as a distributed vehicle-mounted autopilot system. The detection device 91 is used for detecting objects around the movable platform to obtain a three-dimensional point cloud; memory 92 is used to store program code; a processor 93, calling the program code, for performing the following when the program code is executed: acquiring the three-dimensional point cloud; determining a point cloud-free three-dimensional space corresponding to a preset two-dimensional area according to the preset two-dimensional area; and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the point cloud-free three-dimensional space.

Optionally, when determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the point cloud-free three-dimensional space, the processor is specifically configured to: removing the noise information space in the three-dimensional point cloud to obtain a denoised three-dimensional point cloud; determining a three-dimensional point cloud-free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area; and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the space without the three-dimensional point cloud corresponding to the preset two-dimensional area.

Optionally, the preset two-dimensional area includes a plurality of two-dimensional grids; the processor 93 is further configured to, before determining the three-dimensional point cloud free space corresponding to the preset two-dimensional region according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional region: dividing a preset three-dimensional space area around the movable platform into a plurality of three-dimensional grids, wherein each three-dimensional grid in the plurality of three-dimensional grids is projected into one two-dimensional grid in the preset two-dimensional area.

Optionally, when determining, by the processor 93, a three-dimensional point cloud free space corresponding to the preset two-dimensional region according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional region, the processor is specifically configured to: and determining a three-dimensional point cloud free space corresponding to each two-dimensional grid according to the denoised three-dimensional point cloud corresponding to each two-dimensional grid and the three-dimensional grid corresponding to each two-dimensional grid.

Optionally, when the processor 93 determines the point cloud of the overhead obstacle in the three-dimensional point cloud according to the three-dimensional point cloud free space corresponding to the preset two-dimensional area, the processor is specifically configured to: and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the space without the three-dimensional point cloud corresponding to each two-dimensional grid.

Optionally, the processor 93 is specifically configured to, when determining a three-dimensional point cloud free space corresponding to each two-dimensional mesh according to the denoised three-dimensional point cloud corresponding to each two-dimensional mesh and the three-dimensional mesh corresponding to each two-dimensional mesh: inquiring a first three-dimensional grid without the three-dimensional point cloud in the three-dimensional grids corresponding to the two-dimensional grids from a three-dimensional point with the minimum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grids, and taking the first three-dimensional grid without the three-dimensional point cloud as a starting point of a space without the three-dimensional point cloud; continuously inquiring a first three-dimensional grid with three-dimensional point cloud in three-dimensional grids corresponding to the two-dimensional grids from the starting point, and taking a three-dimensional grid without the three-dimensional point cloud before the first three-dimensional grid with the three-dimensional point cloud as an end point of the three-dimensional point cloud-free space; and if the three-dimensional point with the maximum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grid is not in the first three-dimensional grid with the three-dimensional point cloud, continuously inquiring the next space without the three-dimensional point cloud from the first three-dimensional grid with the three-dimensional point cloud.

Optionally, when determining the point cloud of the overhead obstacle in the three-dimensional point cloud according to the three-dimensional point cloud free space corresponding to each two-dimensional grid, the processor 93 is specifically configured to: and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the maximum three-dimensional-free point cloud space corresponding to each two-dimensional grid.

Optionally, when the processor 93 determines the point cloud of the overhead obstacle in the three-dimensional point cloud according to the maximum three-dimensional-free point cloud space corresponding to each two-dimensional grid, the processor is specifically configured to: and if the height of the maximum three-dimensional point cloud-free space corresponding to the two-dimensional grid is greater than or equal to a preset height threshold, determining the denoised three-dimensional point cloud in the three-dimensional grid above the end point of the maximum three-dimensional point cloud-free space as the air obstacle point cloud.

Optionally, before the processor 93 determines the three-dimensional point cloud free space corresponding to the preset two-dimensional region according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional region, the processor is further configured to: and determining whether the denoised three-dimensional point cloud corresponding to each two-dimensional grid comprises an obstacle point cloud or not according to the maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional grid.

Optionally, when the processor 93 determines whether the denoised three-dimensional point cloud corresponding to each two-dimensional mesh includes an obstacle point cloud according to the maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional mesh, the processor is specifically configured to: and if the maximum height difference of the denoised three-dimensional point clouds corresponding to the two-dimensional grids is larger than or equal to a preset height threshold value, determining that the denoised three-dimensional point clouds corresponding to the two-dimensional grids comprise the obstacle point clouds.

Optionally, when the processor 93 removes the noise information space in the three-dimensional point cloud, it is specifically configured to: and removing the noise information space in the three-dimensional point cloud according to the three-dimensional point cloud density in each three-dimensional grid in the plurality of three-dimensional grids.

Optionally, when the processor 93 removes the noise information space in the three-dimensional point cloud according to the three-dimensional point cloud density in each of the plurality of three-dimensional grids, the processor is specifically configured to: determining whether the three-dimensional point cloud in each three-dimensional grid is a noise point cloud according to the three-dimensional point cloud density in each three-dimensional grid in the plurality of three-dimensional grids and the three-dimensional point cloud density in the adjacent grid of each three-dimensional grid; and if the three-dimensional point cloud in the three-dimensional grid is the noise point cloud, removing the three-dimensional point cloud in the three-dimensional grid.

Optionally, when the processor 93 determines whether the three-dimensional point cloud in each three-dimensional mesh is a noise point cloud according to the three-dimensional point cloud density in each three-dimensional mesh of the plurality of three-dimensional meshes and the three-dimensional point cloud density in the adjacent mesh of each three-dimensional mesh, the processor is specifically configured to: and if the density of the three-dimensional point cloud in the three-dimensional grid is smaller than a first density threshold value and the sum of the densities of the three-dimensional point clouds in the adjacent grids of the three-dimensional grid is smaller than a second density threshold value, determining that the three-dimensional point cloud in the three-dimensional grid is the noise point cloud.

Optionally, the adjacent grids of the three-dimensional grid include at least one of: a three-dimensional grid above the three-dimensional grid, a three-dimensional grid below the three-dimensional grid, a three-dimensional grid in front of the three-dimensional grid, a three-dimensional grid behind the three-dimensional grid, a three-dimensional grid on the left side of the three-dimensional grid, and a three-dimensional grid on the right side of the three-dimensional grid.

Optionally, when the processor 93 removes the noise information space in the three-dimensional point cloud, it is specifically configured to: and removing the noise information space in the three-dimensional point cloud by adopting a K nearest neighbor algorithm.

Optionally, the processor 93 adopts a K nearest neighbor algorithm, and when the noise information space in the three-dimensional point cloud is removed, the K nearest neighbor algorithm is specifically configured to: traversing each three-dimensional point in the three-dimensional point cloud, and determining K neighborhoods closest to the three-dimensional point; calculating the average distance from the three-dimensional point to the K neighborhoods; and if the average distance is larger than or equal to a preset distance threshold value, determining the three-dimensional point as a noise point, and removing the noise point.

Optionally, after the processor 93 determines the aerial obstacle point cloud in the three-dimensional point cloud, the processor is further configured to: and correcting the category of the three-dimensional point cloud in each three-dimensional grid in the plurality of three-dimensional grids according to a region continuity principle.

Optionally, the movable platform comprises: unmanned aerial vehicle, mobile robot or vehicle.

Optionally, the processor 93 is a graphic processor, and the graphic processor is configured to perform parallel processing on each frame of three-dimensional point cloud in multiple frames of three-dimensional point cloud collected by the detection device 91.

The specific principle and implementation manner of the point cloud processing system provided by the embodiment of the invention are similar to those of the above embodiments, and are not described herein again.

The embodiment of the invention provides a movable platform. The movable platform comprises: the airframe, the power system and the point cloud processing system of the embodiment; wherein, a power system is arranged on the machine body and used for providing moving power. The specific implementation and principle of the point cloud processing system are the same as those of the above embodiments, and are not described herein again. Optionally, the movable platform comprises: unmanned aerial vehicle, mobile robot or vehicle. When the movable platform is a vehicle, the body may comprise a load-bearing part of the body, chassis, etc. of the vehicle. When the point cloud processing system is a stand alone device, it may be integrated on the movable platform, e.g., a vehicle, in front-loading or back-loading; when the point cloud processing system is a distributed system, various parts thereof such as the detection device, the processor, the memory, and the like may be installed at the same or different positions of the movable platform.

In addition, the present embodiment also provides a computer-readable storage medium on which a computer program is stored, the computer program being executed by a processor to implement the point cloud processing method described in the above embodiments.

In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, or in a form of hardware plus a software functional unit.

The integrated unit implemented in the form of a software functional unit may be stored in a computer readable storage medium. The software functional unit is stored in a storage medium and includes several instructions to enable a computer device (which may be a personal computer, a server, or a network device) or a processor (processor) to execute some steps of the methods according to the embodiments of the present invention. And the aforementioned storage medium includes: various media capable of storing program codes, such as a usb disk, a removable hard disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, or an optical disk.

It is obvious to those skilled in the art that, for convenience and simplicity of description, the foregoing division of the functional modules is merely used as an example, and in practical applications, the above function distribution may be performed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules to perform all or part of the above described functions. For the specific working process of the device described above, reference may be made to the corresponding process in the foregoing method embodiment, which is not described herein again.

Finally, it should be noted that: the above embodiments are only used to illustrate the technical solution of the present invention, and not to limit the same; while the invention has been described in detail and with reference to the foregoing embodiments, it will be understood by those skilled in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; and the modifications or the substitutions do not make the essence of the corresponding technical solutions depart from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A point cloud processing method is applied to a movable platform, the movable platform is provided with a detection device, the detection device is used for detecting objects around the movable platform to obtain three-dimensional point cloud, and the method comprises the following steps:

acquiring the three-dimensional point cloud;

2. The method of claim 1, wherein the point cloud-free three-dimensional space comprises at least one of:

and a three-dimensional point cloud space and a noise information space are not generated.

3. The method of claim 2, wherein said determining an airborne obstacle point cloud in said three-dimensional point cloud from said point cloud free three-dimensional space comprises:

removing the noise information space in the three-dimensional point cloud to obtain a denoised three-dimensional point cloud;

determining a three-dimensional point cloud-free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area;

and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the space without the three-dimensional point cloud corresponding to the preset two-dimensional area.

4. The method of claim 3, wherein the predetermined two-dimensional area comprises a plurality of two-dimensional meshes;

before determining a three-dimensional point cloud free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area, the method further comprises:

dividing a preset three-dimensional space area around the movable platform into a plurality of three-dimensional grids, wherein each three-dimensional grid in the plurality of three-dimensional grids is projected into one two-dimensional grid in the preset two-dimensional area.

5. The method according to claim 4, wherein the determining a three-dimensional point cloud free space corresponding to the preset two-dimensional area according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional area comprises:

and determining a three-dimensional point cloud free space corresponding to each two-dimensional grid according to the denoised three-dimensional point cloud corresponding to each two-dimensional grid and the three-dimensional grid corresponding to each two-dimensional grid.

6. The method according to claim 4 or 5, wherein the determining the aerial obstacle point cloud in the three-dimensional point cloud according to the three-dimensional point cloud free space corresponding to the preset two-dimensional area comprises:

and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the space without the three-dimensional point cloud corresponding to each two-dimensional grid.

7. The method of claim 5, wherein determining a three-dimensional-free point cloud space corresponding to each of the two-dimensional meshes according to the denoised three-dimensional point cloud corresponding to each of the two-dimensional meshes and the three-dimensional mesh corresponding to each of the two-dimensional meshes comprises:

inquiring a first three-dimensional grid without the three-dimensional point cloud in the three-dimensional grids corresponding to the two-dimensional grids from a three-dimensional point with the minimum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grids, and taking the first three-dimensional grid without the three-dimensional point cloud as a starting point of a space without the three-dimensional point cloud;

continuously inquiring a first three-dimensional grid with three-dimensional point cloud in three-dimensional grids corresponding to the two-dimensional grids from the starting point, and taking a three-dimensional grid without the three-dimensional point cloud before the first three-dimensional grid with the three-dimensional point cloud as an end point of the three-dimensional point cloud-free space;

and if the three-dimensional point with the maximum height in the denoised three-dimensional point cloud corresponding to the two-dimensional grid is not in the first three-dimensional grid with the three-dimensional point cloud, continuously inquiring the next space without the three-dimensional point cloud from the first three-dimensional grid with the three-dimensional point cloud.

8. The method of claim 6, wherein determining the airborne obstacle point cloud in the three-dimensional point cloud from the three-dimensional point cloud free space corresponding to each of the two-dimensional meshes comprises:

and determining the point cloud of the obstacle in the air in the three-dimensional point cloud according to the maximum three-dimensional-free point cloud space corresponding to each two-dimensional grid.

9. The method of claim 8, wherein determining the airborne obstacle point cloud in the three-dimensional point cloud from the largest three-dimensional-free point cloud space corresponding to each of the two-dimensional meshes comprises:

and if the height of the maximum three-dimensional point cloud-free space corresponding to the two-dimensional grid is greater than or equal to a preset height threshold, determining the denoised three-dimensional point cloud in the three-dimensional grid above the end point of the maximum three-dimensional point cloud-free space as the air obstacle point cloud.

10. The method according to any one of claims 5-9, wherein before determining the three-dimensional-free point cloud space corresponding to the predetermined two-dimensional area according to the denoised three-dimensional point cloud corresponding to the predetermined two-dimensional area, the method further comprises:

and determining whether the denoised three-dimensional point cloud corresponding to each two-dimensional grid comprises an obstacle point cloud or not according to the maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional grid.

11. The method of claim 10, wherein determining whether the denoised three-dimensional point cloud corresponding to each two-dimensional mesh includes an obstacle point cloud according to a maximum height difference of the denoised three-dimensional point cloud corresponding to each two-dimensional mesh comprises:

and if the maximum height difference of the denoised three-dimensional point clouds corresponding to the two-dimensional grids is larger than or equal to a preset height threshold value, determining that the denoised three-dimensional point clouds corresponding to the two-dimensional grids comprise the obstacle point clouds.

12. The method of any one of claims 4-11, wherein said removing the noise information space in the three-dimensional point cloud comprises:

and removing the noise information space in the three-dimensional point cloud according to the three-dimensional point cloud density in each three-dimensional grid in the plurality of three-dimensional grids.

13. The method of claim 12, wherein removing the noise information space in the three-dimensional point cloud according to a three-dimensional point cloud density in each of the plurality of three-dimensional meshes comprises:

determining whether the three-dimensional point cloud in each three-dimensional grid is a noise point cloud according to the three-dimensional point cloud density in each three-dimensional grid in the plurality of three-dimensional grids and the three-dimensional point cloud density in the adjacent grid of each three-dimensional grid;

and if the three-dimensional point cloud in the three-dimensional grid is the noise point cloud, removing the three-dimensional point cloud in the three-dimensional grid.

14. The method of claim 13, wherein determining whether the three-dimensional point cloud in each of the three-dimensional meshes is a noise point cloud based on the three-dimensional point cloud density in each of the three-dimensional meshes and the three-dimensional point cloud density in neighboring meshes of each of the three-dimensional meshes comprises:

and if the density of the three-dimensional point cloud in the three-dimensional grid is smaller than a first density threshold value and the sum of the densities of the three-dimensional point clouds in the adjacent grids of the three-dimensional grid is smaller than a second density threshold value, determining that the three-dimensional point cloud in the three-dimensional grid is the noise point cloud.

15. The method of claim 13 or 14, wherein adjacent ones of the three-dimensional meshes comprise at least one of:

a three-dimensional grid above the three-dimensional grid, a three-dimensional grid below the three-dimensional grid, a three-dimensional grid in front of the three-dimensional grid, a three-dimensional grid behind the three-dimensional grid, a three-dimensional grid on the left side of the three-dimensional grid, and a three-dimensional grid on the right side of the three-dimensional grid.

16. The method of any one of claims 3-11, wherein said removing the noise information space in the three-dimensional point cloud comprises:

and removing the noise information space in the three-dimensional point cloud by adopting a K nearest neighbor algorithm.

17. The method of claim 16, wherein said removing the noise information space in the three-dimensional point cloud using a K-nearest neighbor algorithm comprises:

traversing each three-dimensional point in the three-dimensional point cloud, and determining K neighborhoods closest to the three-dimensional point;

calculating the average distance from the three-dimensional point to the K neighborhoods;

and if the average distance is larger than or equal to a preset distance threshold value, determining the three-dimensional point as a noise point, and removing the noise point.

18. The method of claim 4, wherein after determining the airborne obstacle point cloud in the three-dimensional point cloud, the method further comprises:

and correcting the category of the three-dimensional point cloud in each three-dimensional grid in the plurality of three-dimensional grids according to a region continuity principle.

19. The method of claim 18, wherein the categories of the three-dimensional point cloud comprise: ground point clouds, ground obstacle point clouds, or air obstacle point clouds.

20. The method of any one of claims 1-19, wherein the movable platform comprises: unmanned aerial vehicle, mobile robot or vehicle.

21. A point cloud processing system, comprising: a detection device, a memory, and a processor;

the memory is used for storing program codes;

the processor, invoking the program code, when executed, is configured to:

acquiring the three-dimensional point cloud;

22. The point cloud processing system of claim 21, wherein the point cloud-free three-dimensional space comprises at least one of:

23. The point cloud processing system of claim 22, wherein the processor is configured to determine an airborne obstacle point cloud in the three-dimensional point cloud from the point cloud-free three-dimensional space, and is further configured to:

24. The point cloud processing system of claim 23, wherein the predetermined two-dimensional area comprises a plurality of two-dimensional meshes;

the processor is further configured to, before determining a three-dimensional point cloud free space corresponding to the preset two-dimensional region according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional region:

25. The point cloud processing system of claim 24, wherein the processor is configured to, when determining, according to the denoised three-dimensional point cloud corresponding to the preset two-dimensional region, a three-dimensional point cloud free space corresponding to the preset two-dimensional region:

26. The point cloud processing system of claim 24 or 25, wherein the processor is configured to, when determining the aerial obstacle point cloud in the three-dimensional point cloud according to the three-dimensional point cloud free space corresponding to the preset two-dimensional area, specifically:

27. The point cloud processing system of claim 25, wherein the processor is configured to determine a three-dimensional-free point cloud space corresponding to each of the two-dimensional grids according to the denoised three-dimensional point cloud corresponding to each of the two-dimensional grids and the three-dimensional grid corresponding to each of the two-dimensional grids, and is further configured to:

28. The point cloud processing system of claim 26, wherein the processor is configured to determine an airborne obstacle point cloud in the three-dimensional point cloud from the three-dimensional-free point cloud space corresponding to each of the two-dimensional grids, in particular:

29. The point cloud processing system of claim 28, wherein the processor is configured to determine an airborne obstacle point cloud in the three-dimensional point cloud from a maximum three-dimensional-free point cloud space corresponding to each of the two-dimensional grids, in particular:

30. The point cloud processing system of any of claims 25-29, wherein the processor is further configured to, before determining the three-dimensional-free point cloud space corresponding to the predetermined two-dimensional region according to the denoised three-dimensional point cloud corresponding to the predetermined two-dimensional region:

31. The point cloud processing system of claim 30, wherein the processor is configured to determine whether the denoised three-dimensional point cloud corresponding to each of the two-dimensional meshes includes an obstacle point cloud according to a maximum height difference of the denoised three-dimensional point cloud corresponding to each of the two-dimensional meshes, and is specifically configured to:

32. The point cloud processing system of any of claims 24-31, wherein the processor, when removing the noise information space in the three-dimensional point cloud, is specifically configured to:

33. The point cloud processing system of claim 32, wherein the processor is configured to, in removing the noise information space from the three-dimensional point cloud based on a three-dimensional point cloud density in each of the plurality of three-dimensional grids, in particular:

34. The point cloud processing system of claim 33, wherein the processor is further configured to determine whether the three-dimensional point cloud in each of the three-dimensional meshes is a noise point cloud based on the density of the three-dimensional point cloud in each of the plurality of three-dimensional meshes and the density of the three-dimensional point cloud in adjacent meshes of each of the three-dimensional meshes, and is further configured to:

35. The point cloud processing system of claim 33 or 34, wherein neighboring meshes of the three-dimensional mesh comprise at least one of:

36. The point cloud processing system of any of claims 23-31, wherein the processor, when removing the noise information space in the three-dimensional point cloud, is specifically configured to:

37. The point cloud processing system of claim 36, wherein the processor employs a K-nearest neighbor algorithm to remove the noise information space in the three-dimensional point cloud, and is specifically configured to:

38. The point cloud processing system of claim 24, wherein after the processor determines the cloud of airborne obstacle points in the three-dimensional point cloud, further configured to:

39. The point cloud processing system of claim 38, wherein the categories of the three-dimensional point cloud include: ground point clouds, ground obstacle point clouds, or air obstacle point clouds.

40. The point cloud processing system of any of claims 21-39, wherein the movable platform comprises: unmanned aerial vehicle, mobile robot or vehicle.

41. The point cloud processing system of any one of claims 21-40, wherein the processor is a graphics processor configured to process each of the plurality of frames of three-dimensional point clouds captured by the detection device in parallel.

42. A movable platform, comprising:

a body;

and the point cloud processing system of any of claims 21-41.

43. The movable platform of claim 42, comprising: unmanned aerial vehicle, mobile robot or vehicle.

44. A computer-readable storage medium, having stored thereon a computer program for execution by a processor to perform the method of any one of claims 1-20.