CN118037790A

CN118037790A - Point cloud processing method and device, computer equipment and storage medium

Info

Publication number: CN118037790A
Application number: CN202410045432.8A
Authority: CN
Inventors: 张煜东; 张磊
Original assignee: Beijing Jidu Technology Co Ltd
Current assignee: Beijing Jidu Technology Co Ltd
Priority date: 2024-01-11
Filing date: 2024-01-11
Publication date: 2024-05-14

Abstract

The disclosure provides a point cloud processing method, a point cloud processing device, computer equipment and a storage medium, and relates to the technical field of computers. According to the method, the point cloud map segments and the target point cloud data can be divided into grids which are convenient to process, then, the point cloud distribution information of each grid can be determined, and the dynamic point cloud detection of the grids can be realized by comparing the point cloud distribution information of the matched grids, so that the dynamic point cloud in the grids can be removed, and ghosts formed by dynamic objects are eliminated; and the mean value and the variance of the coordinates of the point cloud in the vertical axis direction of the grid are used for detecting the dynamic point cloud, so that the accuracy of the detection of the dynamic point cloud can be improved, and the misprinting and the less misprinting of the point cloud are reduced.

Description

Point cloud processing method and device, computer equipment and storage medium

Technical Field

The disclosure relates to the technical field of computers, and in particular relates to a point cloud processing method, a point cloud processing device, computer equipment and a storage medium.

Background

Lidar is a sensing technique that acquires target position and distance information by transmitting a laser beam and measuring its reflection. It is widely used in various applications, such as a vehicle on which a laser radar may be mounted, a vehicle that may be positioned and mapped using a solid state laser radar mounted directly in front of the vehicle, or a 360 degree rotation laser radar mounted above the vehicle.

Although the laser radar-based mapping system can construct a map of a surrounding environment through a 3D point cloud, the interference of dynamic obstacles, such as moving pedestrians, shuttling bicycles and running vehicles, can be inevitably encountered in the mapping process, and the dynamic objects can form ghosts in the point cloud map, so that the subsequent positioning and navigation are not facilitated.

Disclosure of Invention

The embodiment of the disclosure at least provides a point cloud processing method, a point cloud processing device, computer equipment and a storage medium.

In a first aspect, an embodiment of the present disclosure provides a point cloud processing method, including:

Acquiring target point cloud data acquired in a target place and a point cloud map segment matched with the target point cloud data; the corresponding area of the target point cloud data in the target place is the same as the corresponding area of the point cloud map segment in the target place;

Respectively rasterizing the point cloud map segment and the target point cloud data to obtain a plurality of first grids in the point cloud map segment and a plurality of second grids in the target point cloud data;

Determining first point cloud distribution information of a first grid and second point cloud distribution information of a second grid corresponding to the first grid aiming at any first grid; the point cloud distribution information comprises the mean value and variance of coordinates of point cloud in the grid in the vertical axis direction;

Performing dynamic point cloud detection on the first grid based on the first point cloud distribution information and the second point cloud distribution information;

And under the condition that the dynamic point cloud exists in the first grid, removing the dynamic point cloud in the first grid to obtain an updated point cloud map segment.

According to the point cloud processing method provided by the aspect, the point cloud map segments and the target point cloud data can be divided into grids which are convenient to process, then, the point cloud distribution information of each grid can be determined, and the dynamic point cloud detection of the grids can be realized by comparing the point cloud distribution information of the matched grids, so that the dynamic point cloud in the grids can be removed, and the ghosts formed by the dynamic objects are eliminated; and the mean value and the variance of the coordinates of the point cloud in the vertical axis direction of the grid are used for detecting the dynamic point cloud, so that the accuracy of the detection of the dynamic point cloud can be improved, and the misprinting and the less misprinting of the point cloud are reduced.

In an optional implementation manner, the performing dynamic point cloud detection on the first grid based on the first point cloud distribution information and the second point cloud distribution information includes:

Determining difference information between the first grid and the corresponding second grid based on the first point cloud distribution information and the second point cloud distribution information;

And under the condition that the difference information meets the preset condition, determining that the first grid has a dynamic point cloud.

According to the embodiment, the difference information between grids can be judged through the point cloud distribution information, so that whether the first grid has the dynamic point cloud or not is judged according to the difference of the point clouds of the grids on the vertical axis coordinates.

In an optional implementation manner, the determining, based on the first point cloud distribution information and the second point cloud distribution information, difference information between the first grid and the corresponding second grid includes:

Determining a first difference factor based on a difference between a second mean value in the second point cloud distribution information and a first mean value in the first point cloud distribution information, and a first variance in the first point cloud distribution information;

determining a second difference factor based on the first difference factor and a second variance in the first point cloud distribution information;

determining a third difference factor based on a logarithm of a ratio between the first variance and the second variance;

and determining difference information between the first grid and the corresponding second grid based on the sum of the second difference factor and the third difference factor.

According to the embodiment, the difference factors can be determined according to the mean value and the variance of the coordinates of the point clouds of the grids in the vertical axis direction, and then the difference information among the grids is determined according to the difference factors, so that the accuracy of the difference information is improved, and the accuracy of judging the dynamic point clouds is further improved.

In an alternative embodiment, the removing the dynamic point cloud in the first grid includes:

Identifying ground point clouds from the first grid and the second grid, and separating the ground point clouds from the first grid and the second grid to obtain a first intermediate point cloud corresponding to the first grid and a second intermediate point cloud corresponding to the second grid;

constructing a first octree map based on the first intermediate point cloud; constructing a second octree map based on the second intermediate point cloud; wherein the octree map comprises a plurality of cubes, and any one cube comprises at most one point cloud point;

performing dynamic point cloud filtering on the first intermediate point cloud by using the first octree map and the second octree map to obtain a filtered first intermediate point cloud;

And splicing the first intermediate point cloud with the corresponding ground point cloud to obtain a first grid from which the dynamic point cloud is removed.

According to the embodiment, the ground point cloud can be separated from the grid, so that the calculation amount of determining the dynamic point cloud is reduced, and the point cloud points in the middle point cloud can be separated through cube segmentation by creating the octree map of the middle point cloud, so that the difference between the middle point clouds is judged through the octree map, the dynamic point cloud filtering is performed, and the accuracy of the dynamic point cloud filtering is improved.

In an alternative embodiment, a ground point cloud is identified from the first grid by:

determining a plurality of target point cloud points with minimum coordinates in the vertical axis direction in the first grid;

Establishing a point cloud matrix corresponding to the target point cloud point, and determining a first characteristic value of the point cloud matrix in the horizontal direction and a second characteristic value of the point cloud matrix in the vertical axis direction;

And taking the point cloud formed by the target point cloud points as the ground point cloud in the first grid under the condition that the difference value between the first characteristic value and the second characteristic value is larger than a preset threshold value.

According to the embodiment, the characteristic comparison can be carried out on the point cloud matrix formed by the plurality of target point cloud points with the minimum coordinates in the vertical axis direction, so that when the horizontal direction characteristic is sufficiently larger than the vertical axis direction characteristic, the point cloud formed by the target point cloud is used as the ground point cloud, and the accuracy of ground point cloud identification is improved.

In an alternative embodiment, before acquiring the target point cloud data acquired in the target location and the point cloud map segment matched with the target point cloud data, the method further includes:

Acquiring multi-frame point cloud data acquired in a target place, and generating a point cloud map of the target place by utilizing the point cloud data;

the acquiring the target point cloud data acquired in the target place and the point cloud map segment matched with the target point cloud data comprises the following steps:

Taking any point cloud data in the multi-frame point cloud data as the target point cloud data, and determining a shooting pose corresponding to the target point cloud data;

And based on the shooting pose and a preset segmentation range, a point cloud map segment matched with the target point cloud data is segmented from the point cloud map.

In this embodiment, a point cloud map can be generated using the point cloud data, and the dynamic point cloud removal of the area (i.e., the point cloud map segment) in the point cloud map can be achieved using the collected original point cloud data (i.e., the target point cloud data).

In an alternative embodiment, the method further comprises:

acquiring updated point cloud map segments corresponding to the point cloud data of each frame;

Generating a target point cloud map of the target place, from which the dynamic point cloud is removed, based on the obtained point cloud map segment; the target point cloud map is used for positioning services for the target place.

According to the embodiment, the point cloud map segments corresponding to the point cloud data of each frame can be used for generating the target point cloud map for removing the dynamic point cloud, and the target point cloud map is used for providing positioning service for the target place.

In a second aspect, an embodiment of the present disclosure further provides a point cloud processing apparatus, including:

The acquisition module is used for acquiring target point cloud data acquired in a target place and a point cloud map segment matched with the target point cloud data; the corresponding area of the target point cloud data in the target place is the same as the corresponding area of the point cloud map segment in the target place;

the rasterizing module is used for respectively rasterizing the point cloud map segment and the target point cloud data to obtain a plurality of first grids in the point cloud map segment and a plurality of second grids in the target point cloud data;

The determining module is used for determining first point cloud distribution information of any first grid and second point cloud distribution information of a second grid corresponding to the first grid; the point cloud distribution information comprises the mean value and variance of coordinates of point cloud in the grid in the vertical axis direction;

the detection module is used for carrying out dynamic point cloud detection on the first grid based on the first point cloud distribution information and the second point cloud distribution information;

And the processing module is used for removing the dynamic point cloud in the first grid under the condition that the dynamic point cloud exists in the first grid, so as to obtain an updated point cloud map segment.

In an alternative embodiment, the detection module is specifically configured to:

In an optional implementation manner, the detection module is specifically configured to, when determining difference information between the first grid and the corresponding second grid based on the first point cloud distribution information and the second point cloud distribution information:

In an alternative embodiment, the processing module is specifically configured to:

In an alternative embodiment, the processing module is further configured to:

In an alternative embodiment, the obtaining module is further configured to:

the acquisition module is specifically configured to, when acquiring target point cloud data acquired in a target location and a point cloud map segment matched with the target point cloud data:

In an alternative embodiment, the apparatus further comprises a generating module configured to:

In a third aspect, an optional implementation manner of the disclosure further provides a computer device, a processor, and a memory, where the memory stores machine-readable instructions executable by the processor, and the processor is configured to execute the machine-readable instructions stored in the memory, where the machine-readable instructions, when executed by the processor, perform the steps in the first aspect, or any possible implementation manner of the first aspect, when executed by the processor.

In a fourth aspect, an alternative implementation of the present disclosure further provides a computer readable storage medium having stored thereon a computer program which when executed performs the steps of the first aspect, or any of the possible implementation manners of the first aspect.

The description of the effects of the point cloud processing apparatus, the computer device, and the computer readable storage medium is referred to the description of the point cloud processing method, and is not repeated here.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the aspects of the disclosure.

The foregoing objects, features and advantages of the disclosure will be more readily apparent from the following detailed description of the preferred embodiments taken in conjunction with the accompanying drawings.

Drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, the drawings required for the embodiments are briefly described below, which are incorporated in and constitute a part of the specification, these drawings showing embodiments consistent with the present disclosure and together with the description serve to illustrate the technical solutions of the present disclosure. It is to be understood that the following drawings illustrate only certain embodiments of the present disclosure and are therefore not to be considered limiting of its scope, for the person of ordinary skill in the art may admit to other equally relevant drawings without inventive effort.

FIG. 1 illustrates a flow chart of a point cloud processing method provided by some embodiments of the present disclosure;

FIG. 2 illustrates a schematic diagram of a rasterized point cloud provided by some embodiments of the present disclosure;

FIG. 3 illustrates a flow chart of another point cloud processing method provided by some embodiments of the present disclosure;

FIG. 4 illustrates a flowchart of steps provided by some embodiments of the present disclosure to generate a point cloud map;

FIG. 5 illustrates a schematic diagram of a point cloud processing device provided by some embodiments of the present disclosure;

Fig. 6 illustrates a schematic diagram of a computer device provided by some embodiments of the present disclosure.

Detailed Description

For the purposes of making the objects, technical solutions and advantages of the embodiments of the present disclosure more apparent, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are only some embodiments of the present disclosure, but not all embodiments. The components of the disclosed embodiments generally described and illustrated herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present disclosure is not intended to limit the scope of the disclosure, as claimed, but is merely representative of selected embodiments of the disclosure. All other embodiments, which can be made by those skilled in the art based on the embodiments of this disclosure without making any inventive effort, are intended to be within the scope of this disclosure.

According to research, in the process of scene mapping by using a laser radar, the interference of dynamic obstacles such as moving pedestrians, shuttling bicycles and running vehicles can be encountered, and the dynamic objects can form ghosts in a point cloud map, so that the subsequent positioning and navigation by using the point cloud map are not facilitated. If the neural network is used to filter the dynamic point cloud in real time in the process of collecting the point cloud data, more computing resources are needed, the learning method can only identify the type of the trained dynamic object, other types cannot be identified, and expensive graphic processing equipment is needed to provide hardware support for the dynamic object, so that a lot of cost is increased.

Based on the above study, the disclosure provides a point cloud processing method, a device, a computer device and a storage medium, which can divide a point cloud map segment and target point cloud data into grids convenient to process after generating a point cloud map, then can determine the point cloud distribution information of each grid, and can realize the dynamic point cloud detection of the grids by comparing the point cloud distribution information of the matched grids, thereby removing the dynamic point cloud in the grids and eliminating ghosts formed by dynamic objects; and the mean value and the variance of the coordinates of the point cloud in the vertical axis direction of the grid are used for detecting the dynamic point cloud, so that the accuracy of the detection of the dynamic point cloud can be improved, and the misprinting and the less misprinting of the point cloud are reduced.

The present invention is directed to a method for manufacturing a semiconductor device, and a semiconductor device manufactured by the method.

It should be noted that: like reference numerals and letters denote like items in the following figures, and thus once an item is defined in one figure, no further definition or explanation thereof is necessary in the following figures.

For the sake of understanding the present embodiment, first, a detailed description will be given of a point cloud processing method disclosed in an embodiment of the present disclosure, where an execution body of the point cloud processing method provided in the embodiment of the present disclosure is generally a computer device with a certain computing capability, where the computer device includes, for example: a terminal device or server or other processing device. In some possible implementations, the point cloud processing method may be implemented by a processor invoking computer readable instructions stored in a memory.

The point cloud processing method provided by the embodiment of the present disclosure is described below by taking an execution subject as an in-vehicle computer device as an example.

Referring to fig. 1, a flowchart of a point cloud processing method according to an embodiment of the disclosure is shown, where the method includes steps S101 to S104, where:

S101, acquiring target point cloud data acquired in a target place and a point cloud map segment matched with the target point cloud data; and the corresponding area of the target point cloud data in the target place is the same as the corresponding area of the point cloud map segment in the target place.

In this step, the target location may be a location where a point cloud map needs to be established, and the location may be an outdoor location or an indoor location, such as a parking lot.

When the vehicle runs on the target site, the point cloud data in the target site can be acquired through a laser radar deployed on the vehicle. The laser radar can be deployed at the front end of a vehicle or at the top of the vehicle. The vehicle can collect point cloud data of the environment near the vehicle in the target place, and in the collecting process, multiple frames of point cloud data can be collected.

After the vehicle collects enough point cloud data, a point cloud map of the target site can be established by using the collected point cloud data, and when the point cloud map is established, a simultaneous localization and mapping technology (Simultaneous Localization AND MAPPING, SLAM) can be adopted.

The point cloud map segment can be obtained by segmentation from the established point cloud map.

In the implementation process, the vehicle-mounted computer equipment can firstly collect multi-frame point cloud data in the target place and generate a point cloud map of the target place by utilizing the collected multi-frame point cloud data. Then, any point cloud data in the multi-frame point cloud data can be used as target point cloud data, shooting pose corresponding to the target point cloud data is determined, then, the position matched with the shooting pose in the point cloud map is determined, the point cloud map is segmented by taking the position as an origin and taking a preset segmentation range as a reference, and a point cloud map segment is obtained, so that the area in a target place corresponding to the point cloud map is consistent with the area in a target place in the target point cloud data.

For example, feature points, such as ground features, wall features, etc., may be extracted from the laser point cloud data, which may be used to locate features of the target site in a subsequent process, and to match the point cloud data at a specific point in the map building.

According to the laser point cloud data and the motion trail of the vehicle, a point cloud map of the surrounding environment can be constructed. According to the continuous laser radar data, the characteristic points in different frames of laser radar data can be matched, so that the motion trail and the gesture change of the laser radar device (or the vehicle carrying the laser radar device) can be estimated. The method can integrate the previous positioning information, sensor measurement and motion estimation through an optimization algorithm, and update the vehicle pose estimation and the point cloud map in real time, so as to output the target point cloud data and the corresponding point cloud map segment.

In the process of dynamic point cloud removal, the point cloud data can be used as the target point cloud data frame by frame according to the acquisition time sequence of the point cloud data.

S102, rasterizing the point cloud map segment and the target point cloud data respectively to obtain a plurality of first grids in the point cloud map segment and a plurality of second grids in the target point cloud data.

After the target point cloud data and the matched point cloud map segments are obtained, the point cloud map segments and the target point cloud data can be subjected to rasterization, and the point cloud map segments and the target point cloud data can be longitudinally segmented into a plurality of grids during the rasterization. The grid may be a cuboid, perpendicular to the ground plane.

Before rasterization processing, the point cloud segment and the target point cloud data can be cut out, and an interest area where dynamic point cloud possibly exists is cut out. For example, if the target location is an outdoor scene, an area within a certain range (for example, within 80 meters, or within 30 meters to 60 meters) from the origin of the coordinate system may be used as the region of interest; if the target place is an indoor scene, the value range on the vertical axis can be further limited to be between-1 meter and +3 meters.

In the indoor scene, the point cloud corresponding to the roof part can be removed in advance.

Referring to fig. 2, a schematic diagram of a rasterized point cloud according to an embodiment of the present disclosure is shown. A cartesian coordinate system may be established in the rasterized point cloud, and the horizontal direction may be defined by a horizontal axis X and a vertical axis Y, the vertical axis may be Z, and the origin may be O. Fig. 2 is a grid-formed point cloud in a plan view.

Illustratively, assume that the original point cloud scan (i.e., target point cloud data) at time t is Pose in map coordinate system (i.e. world coordinate system) is/>Then can be in pose points/>Centered, a local map submap (i.e., a point cloud map segment) is circled in the point cloud map M at a particular radius, denoted/>Then there are:

Wherein the origin positions of the target point cloud data scan and the point cloud map segment submap are overlapped, i.e., known And/>Is coincident. Considering that the point cloud of each frame of the laser radar is sparse at a far position, by setting the interest area, the phenomenon of misjudging the far laser radar point cloud as a dynamic point and deleting the far laser radar point cloud by mistake can be avoided. For indoor scenes, the separated roof point cloud can be recorded as/>Then there are:

Wherein scan may represent a lidar scan frame (i.e., target point cloud data), W represents a world coordinate system, sub represents a point cloud map segment, and root represents a roof.

S103, determining first point cloud distribution information of a first grid and second point cloud distribution information of a second grid corresponding to the first grid aiming at any first grid; the point cloud distribution information comprises the mean value and the variance of coordinates of point clouds in the grid in the vertical axis direction.

After the point cloud map segment and the target point cloud data are divided into grids, the point cloud distribution information corresponding to each grid can be calculated.

The point cloud distribution information of the first grid (i.e., the grid in the point cloud map segment) may be first point cloud distribution information; the point cloud distribution information of the second grid (i.e., the grid in the target point cloud data) may be second point cloud distribution information.

The above-mentioned point cloud distribution information may include the mean and variance of coordinates of the point cloud in the grid in the vertical axis (i.e., Z-axis) direction.

The first average value in the first point cloud distribution information may be:

The first variance in the first point cloud distribution information may be:

similarly, the second average value in the second point cloud distribution information may be:

the second variance in the second point cloud distribution information may be:

Wherein bin _i denotes an ith grid, and P denotes coordinates of a point cloud point in the grid in the vertical axis direction.

After rasterizing the point cloud, since the point cloud map segment and the target point cloud data have the same coordinate system and the corresponding scene ranges are the same, the grids between the point cloud map segment and the target point cloud data have a correspondence. For example, the grids (1, 1) in the point cloud map segment and the grids (1, 1) in the target point cloud data correspond to each other.

S104, carrying out dynamic point cloud detection on the first grid based on the first point cloud distribution information and the second point cloud distribution information.

After the first point cloud distribution information and the second point cloud distribution information are obtained, dynamic point cloud detection can be performed on the first grid.

When the dynamic point cloud detection is performed, the difference information between the first grid and the corresponding second grid can be determined by using the first point cloud distribution information and the second point cloud distribution information. And judging whether the dynamic point cloud exists in the first grid according to the difference information.

Specifically, because the point cloud data is obtained frame by frame, if an object is in a motion state in the process of obtaining the point cloud data, the positions of the object in the point cloud data of different frames are different, but the point cloud map is formed by splicing multiple frames of point cloud data, and the positions of a moving object in the point cloud data of different frames are also different, a plurality of corresponding images exist in the point cloud map, so that a continuous track, namely, a ghost image, is formed. In the point cloud map segment, there may be a plurality of images of the moving object, but only one image exists in the original point cloud data, and by comparing whether the image of the same object exists in the point cloud map segment and the target point cloud data at a position, it can be determined whether the object is a moving object, that is, whether the point cloud at the position is a dynamic point cloud.

In order to determine whether a dynamic point cloud exists in the grid, the point cloud distribution information of the grid needs to be compared to determine difference information.

When determining the difference information, a first difference factor may be determined based on a difference between a second average value in the second point cloud distribution information and a first average value in the first point cloud distribution information, and a first variance in the first point cloud distribution information; determining a second difference factor based on the first difference factor and a second variance in the first point cloud distribution information; determining a third difference factor based on a logarithm of the ratio between the first variance and the second variance; and finally, determining the difference information between the first grid and the corresponding second grid based on the sum of the second difference factor and the third difference factor.

When the first difference factor is determined, a difference value between the second average value and the first average value can be determined, the difference value is squared, and the square of the difference value and the first difference value are summed to obtain the first difference factor. Then, the square of the second variance can be subtracted by 1 to obtain a reference value, and the ratio between the first difference factor and the reference value is determined to obtain a second difference factor.

Then, the square of the logarithm of the ratio between the first variance and the second variance may be determined, and the square of the logarithm may be weighted to obtain a third difference factor.

Finally, the sum between the second and third difference factors may be taken as the above-mentioned difference information.

Exemplary, the distribution of each grid i in submap and scan point clouds can be obtained according to the above-mentioned step of determining the point cloud distribution information

To determine whether grid i in submap contains dynamic points, the value of KL divergence may be used as difference information, a criterion for the difference in point cloud distribution in the z-direction for submap and scan may be measured. The difference information may be:

In this way, under the condition that the difference information meets the preset condition, it can be determined that the first grid has a dynamic point cloud.

For example, if the value of KL (P (x) |q (x)) is less than the threshold, it may be considered that no dynamic point is contained in the current grid. When the value of KL (P (x) |q (x)) is greater than or equal to the threshold, then submap in the current grid may be considered to contain dynamic points, requiring filtering.

S105, under the condition that the dynamic point cloud exists in the first grid, removing the dynamic point cloud in the first grid to obtain an updated point cloud map segment.

When the dynamic point cloud exists in the first grid, the dynamic point cloud in the first grid can be removed, and therefore updated point cloud map segments are obtained. In this step, the dynamic point cloud in the first grid may be first identified, and then the identified dynamic point cloud may be deleted.

Before the dynamic point cloud is identified, the ground point cloud which is not necessarily the dynamic point cloud in the first grid can be identified first and separated from the first grid, so that whether the part of ground point cloud is the dynamic point cloud is not needed to be judged, and the calculated amount is reduced.

When the ground point cloud is identified, a plurality of target point cloud points with minimum coordinates in the vertical axis direction in the first grid can be determined; then, a point cloud matrix corresponding to the target point cloud point is established, and a first characteristic value of the point cloud matrix in the horizontal direction and a second characteristic value of the point cloud matrix in the vertical axis direction are determined.

The first eigenvalue and the second eigenvalue of the point cloud matrix may be performed by singular value decomposition (Singular Value Decomposition, SVD). SVD can decompose a matrix into products of three matrices, resulting in features in the X, Y, and Z axes.

For example, the point cloud data in the scan grid may be segmented on the X-Y plane, the minimum height Zmin in the grid may be found, all Zmin in the grid may be combined into a minimum Gao Chengmian, and the point cloud points in the minimum elevation plane may be used to form the point cloud matrix a. Then, the eigenvalue decomposition can be performed on the matrix A through SVD to obtain the eigenvalues of the X axis, the Y axis and the Z axis.

Then, in a case where a difference between the first characteristic value and the second characteristic value is greater than a preset threshold value (i.e., in a case where the first characteristic value is far greater than the second characteristic value), a point cloud composed of target point cloud points may be regarded as the ground point cloud in the first grid.

Illustratively, if the eigenvalues in the X-axis and Y-axis directions are much greater than the eigenvalues in the Z-axis directions, then the minimum elevation surface point is considered to be a surface point cloud

Similarly, a similar operation can be performed on the second grid to remove the ground point cloud in the target point cloud data

After the ground point cloud is removed, a first intermediate point cloud of the ground point cloud removed by the first grid and a second intermediate point cloud of the ground point cloud removed by the second grid can be obtained.

In order to compare the first intermediate point cloud and the second intermediate point cloud, the first intermediate point cloud and the second intermediate point cloud may be normalized, and the intermediate point cloud may be converted into an octree map.

Octree Map (Octree Map) is a data structure for representing a three-dimensional space, and is generally used in the fields of computer graphics, three-dimensional Map modeling, physical simulation, computer games, and the like. It is a tree structure in which each node can have a maximum of eight children, which represent sub-regions in space, and the children can continue to divide into further children. Through the octree map, the point cloud points in the point cloud can be respectively divided into an independent cube, namely, at most one point cloud point is contained in any cube, or 0 point cloud points are contained in any cube.

In this way, the first intermediate point cloud can be dynamically filtered by utilizing the octree map by dividing the intermediate point cloud into a plurality of cubes, so as to obtain the filtered first intermediate point cloud.

Illustratively, the first octree map of the point cloud map segment submap may be represented as: octoMap ^sub a second octree map of the target point cloud data scan may be represented as: octoMap ^scan.

A grid (cube) containing a point cloud in OctoMap ^scan can be expressed as occupying Odd(s) =1, and a grid not containing a point cloud is expressed as Odd(s) =0. Then OctoMap ^scan can be used as a mask (mask) with which OctoMap ^sub is filtered.

Specifically, in the grid portion corresponding to OctoMap ^scan in OctoMap ^sub, if Odd(s) =1 of OctoMap ^scan, it is indicated that the target point cloud data and the point cloud map fragment simultaneously contain the point cloud at the position, the corresponding point cloud in the grid of OctoMap ^sub remains, and conversely, when Odd(s) =0 of OctoMap ^scan, it is indicated that the point cloud does not exist at the position in the target point cloud data, the corresponding point cloud in the grid of OctoMap ^sub may be removed.

By comparing the second intermediate point cloudAnd a first intermediate point cloud/>Find/>The dynamic point clouds in the network are marked and recorded as point clouds/>

If the environment is outdoor, the point cloud map segment after removing the dynamic point cloud may be:

if the indoor environment is present, the roof point cloud temporarily separated before is restored, and the point cloud map segment after the dynamic point cloud is removed can be:

After recovering the point cloud of the roof And outputting the map as a final static point cloud map.

In the above step, for the calculation of the point cloud data at the time T, the above operation is performed on the set [ T ] at all times, so that a static 3D point cloud map without dynamic objects can be obtained. That is, updated point cloud map segments corresponding to the point cloud data of each frame can be obtained; and generating a target point cloud map of the target place, from which the dynamic point cloud is removed, based on the acquired point cloud map segment. The target point cloud map may be used for positioning services and navigation services for the target location.

According to the point cloud processing method provided by the embodiment of the disclosure, the point cloud map segments and the target point cloud data can be divided into grids convenient to process, then, the point cloud distribution information of each grid can be determined, and the dynamic point cloud detection of the grids can be realized by comparing the point cloud distribution information of the matched grids, so that the dynamic point cloud in the grids can be removed, and ghosts formed by dynamic objects are eliminated; and the mean value and the variance of the coordinates of the point cloud in the vertical axis direction of the grid are used for detecting the dynamic point cloud, so that the accuracy of the detection of the dynamic point cloud can be improved, and the misprinting and the less misprinting of the point cloud are reduced.

Referring to fig. 3, a schematic diagram of another point cloud processing method according to an embodiment of the disclosure is shown. The method comprises the following steps:

s110, performing point cloud registration by using a laser SLAM technology to obtain a point cloud map containing dynamic objects.

S120, judging whether the scene is an outdoor scene, if so, jumping to S130, and if not, separating point clouds of the roof part, and then jumping to S130.

S130, reading target point cloud data scan and a point cloud map segment submap, dividing the target point cloud data scan and the point cloud map segment into grids in the same mode, and calculating descriptors (namely the point cloud distribution information).

S140, comparing the difference between the scan and the submap, and judging whether dynamic points exist in submap according to the KL divergence value.

S150, under the condition of dynamic points, using an octree technology to construct ocTree map (octree map) of a scan point cloud, taking the ocTree map as a mask, removing the dynamic point cloud in submap, and retaining the static point cloud in submap.

S160, splicing the static point cloud map.

Referring to fig. 4, a flowchart of steps for generating a point cloud map according to an embodiment of the present disclosure is shown. This step may be a specific embodiment of the step S110 described above. Wherein S110 includes:

S111, acquiring laser point cloud data in the environment by using the 3D laser radar equipment.

S122, extracting characteristic points, such as ground characteristics, wall characteristics and the like, from the laser point cloud data. These feature points can be used for subsequent localization and mapping.

S113, constructing an environment map according to the laser point cloud data and the motion trail of the robot (namely the laser radar device).

S114, estimating the motion track and the posture change of the robot by matching with the previous characteristic points according to the continuous laser radar data.

S115, fusing the prior positioning information, sensor measurement and motion estimation through an optimization algorithm, and updating the pose estimation and the environment map of the robot in real time to obtain a point cloud map.

S116, outputting the pose pose of the scan of each frame and the constructed point cloud map.

It will be appreciated by those skilled in the art that in the above-described method of the specific embodiments, the written order of steps is not meant to imply a strict order of execution but rather should be construed according to the function and possibly inherent logic of the steps.

Based on the same inventive concept, the embodiment of the disclosure further provides a point cloud processing device corresponding to the point cloud processing method, and since the principle of solving the problem by the device in the embodiment of the disclosure is similar to that of the point cloud processing method in the embodiment of the disclosure, the implementation of the device can refer to the implementation of the method, and the repetition is omitted.

Referring to fig. 5, a schematic diagram of a point cloud processing device according to an embodiment of the disclosure is shown, where the device includes:

the acquisition module 510 is configured to acquire target point cloud data acquired in a target location, and a point cloud map segment matched with the target point cloud data; the corresponding area of the target point cloud data in the target place is the same as the corresponding area of the point cloud map segment in the target place;

The rasterizing module 520 is configured to perform rasterizing processing on the point cloud map segment and the target point cloud data, respectively, to obtain a plurality of first grids in the point cloud map segment and a plurality of second grids in the target point cloud data;

A determining module 530, configured to determine, for any first grid, first point cloud distribution information of the first grid, and second point cloud distribution information of a second grid corresponding to the first grid; the point cloud distribution information comprises the mean value and variance of coordinates of point cloud in the grid in the vertical axis direction;

A detection module 540, configured to perform dynamic point cloud detection on the first grid based on the first point cloud distribution information and the second point cloud distribution information;

And the processing module 550 is configured to, when detecting that the dynamic point cloud exists in the first grid, remove the dynamic point cloud in the first grid, and obtain an updated point cloud map segment.

In an alternative embodiment, the detection module 540 is specifically configured to:

In an alternative embodiment, the detection module 540 is specifically configured to, when determining the difference information between the first grid and the corresponding second grid based on the first point cloud distribution information and the second point cloud distribution information:

In an alternative embodiment, the processing module 550 is specifically configured to:

In an alternative embodiment, the processing module 550 is further configured to:

In an alternative embodiment, the obtaining module 510 is further configured to:

The acquiring module 510 is specifically configured to, when acquiring target point cloud data acquired in a target location and a point cloud map segment matched with the target point cloud data:

The process flow of each module in the apparatus and the interaction flow between the modules may be described with reference to the related descriptions in the above method embodiments, which are not described in detail herein.

The embodiment of the disclosure further provides a computer device, as shown in fig. 6, which is a schematic structural diagram of the computer device provided by the embodiment of the disclosure, including:

A processor 61 and a memory 62; the memory 62 stores machine readable instructions executable by the processor 61, the processor 61 being configured to execute the machine readable instructions stored in the memory 62, the machine readable instructions when executed by the processor 61, the processor 61 performing the steps of:

In an alternative embodiment, the method further comprises:

The memory 62 includes a memory 621 and an external memory 622; the memory 621 is also referred to as an internal memory, and is used for temporarily storing operation data in the processor 61 and data exchanged with the external memory 622 such as a hard disk, and the processor 61 exchanges data with the external memory 622 via the memory 621.

The specific execution process of the above instruction may refer to the steps of the point cloud processing method described in the embodiments of the present disclosure, which is not described herein again.

The disclosed embodiments also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the point cloud processing method described in the above method embodiments. Wherein the storage medium may be a volatile or nonvolatile computer readable storage medium.

The embodiments of the present disclosure further provide a computer program product, where the computer program product carries program code, and instructions included in the program code may be used to perform the steps of the point cloud processing method described in the foregoing method embodiments, and specifically refer to the foregoing method embodiments, which are not described herein in detail.

Wherein the above-mentioned computer program product may be realized in particular by means of hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied as a computer storage medium, and in another alternative embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.

It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system and apparatus may refer to corresponding procedures in the foregoing method embodiments, which are not described herein again. In the several embodiments provided in the present disclosure, it should be understood that the disclosed systems, devices, and methods may be implemented in other manners. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be other manners of division in actual implementation, and for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, device or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown 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 may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

In addition, each functional unit in each embodiment of the present disclosure may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.

The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a non-volatile computer readable storage medium executable by a processor. Based on such understanding, the technical solution of the present disclosure may be embodied in essence or a part contributing to the prior art or a part of the technical solution, or in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in the embodiments of the present disclosure. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.

Finally, it should be noted that: the foregoing examples are merely specific embodiments of the present disclosure, and are not intended to limit the scope of the disclosure, but the present disclosure is not limited thereto, and those skilled in the art will appreciate that while the foregoing examples are described in detail, it is not limited to the disclosure: any person skilled in the art, within the technical scope of the disclosure of the present disclosure, may modify or easily conceive changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features thereof; such modifications, changes or substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the disclosure, and are intended to be included within the scope of the present disclosure. Therefore, the protection scope of the present disclosure shall be subject to the protection scope of the claims.

Claims

1. A method of point cloud processing, comprising:

2. The method of claim 1, wherein the performing dynamic point cloud detection on the first grid based on the first point cloud distribution information and the second point cloud distribution information comprises:

3. The method of claim 2, wherein the determining difference information between the first grid and the corresponding second grid based on the first point cloud distribution information and the second point cloud distribution information comprises:

4. The method of claim 1, wherein the removing the dynamic point cloud in the first grid comprises:

5. The method of claim 4, wherein the ground point cloud is identified from the first grid by:

6. The method of claim 1, wherein prior to acquiring the target point cloud data collected in the target site and the point cloud map segment that matches the target point cloud data, the method further comprises:

Acquiring multi-frame point cloud data acquired in the target place, and generating a point cloud map of the target place by utilizing the point cloud data;

7. The method of claim 6, wherein the method further comprises:

8. A point cloud processing apparatus, comprising:

9. A computer device, comprising: a processor, a memory storing machine readable instructions executable by the processor for executing the machine readable instructions stored in the memory, which when executed by the processor, perform the steps of the point cloud processing method according to any of claims 1 to 7.

10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when run by a computer device, performs the steps of the point cloud processing method according to any of claims 1 to 7.