Mobile robot navigation map generation method, device, medium and mobile robot

Abstract

Embodiments of the present application are applicable to the technical field of mobile robots, and provide a mobile robot navigation map generation method, device, medium and mobile robot, the method comprising: acquiring an original point cloud map and an original point cloud trajectory, the original point cloud map and the original point cloud trajectory being obtained by a simultaneous localization and mapping module of a collection device; performing parameter processing on the original point cloud map and the original point cloud trajectory to obtain an intermediate state point cloud map and an intermediate state point cloud trajectory, the intermediate state point cloud map and the intermediate state point cloud trajectory having a characteristic parameter for filtering out residual images; performing residual image filtering processing on the intermediate state point cloud map according to the intermediate state point cloud trajectory and a preset residual image filtering rule to obtain a target point cloud map and / or a target grid map, the preset residual image filtering rule being associated with the characteristic parameter. Through the above method, residual images in the point cloud map can be removed, thereby improving the accuracy of the map.

Description

Technical Field

[0001] This application belongs to the field of mobile robot technology, and in particular relates to a method, device, medium and mobile robot for generating navigation maps for mobile robots. Background Technology

[0002] With the development of autonomous driving technology, the application of 3D laser simultaneous localization and mapping (SLAM) technology is becoming more and more widespread. As a key input data of the SLAM localization module, the accuracy and noise of the 3D point cloud map play a crucial role in the accuracy of localization.

[0003] Currently, in the process of building 3D maps using SLAM, it is generally necessary for personnel to remotely control the movement of a vehicle for map recording. During map recording, due to the unprofessionalism of the recording personnel, they may remain in a certain position on the road for an extended period of time while remotely controlling the vehicle, resulting in residual images of the personnel in the map. The mapping module has difficulty filtering out these personnel images from the map using probabilistic methods, thus resulting in the presence of these images in the final generated map, which reduces the accuracy of the map.

[0004] In relatively narrow scenarios such as office building corridors, long straight corridors, office interiors, factory passageways, and substation indoor passageways, remote-controlled vehicles recording maps can significantly obstruct the passageway cross-section, making it easier for mapping algorithms to record personnel ghosting in the point cloud map. If ghosting exists in locations that should be road areas on the map, these locations may be identified as obstacles in the generated map, reducing the map's positioning accuracy. Summary of the Invention

[0005] In view of this, embodiments of this application provide a method, device, medium, and mobile robot for generating navigation maps for mobile robots, in order to remove ghost points in point cloud maps, thereby improving the accuracy of the maps.

[0006] The first aspect of this application provides a method for generating navigation maps for mobile robots, including:

[0007] The original point cloud map and the original point cloud trajectory are obtained by the acquisition device through the real-time positioning and map building module.

[0008] The original point cloud map and the original point cloud trajectory are processed to obtain an intermediate point cloud map and an intermediate point cloud trajectory, which contain feature parameters for filtering out afterimages.

[0009] Based on the intermediate point cloud trajectory and the preset afterimage filtering rules, the intermediate point cloud map is subjected to afterimage filtering processing to obtain the target point cloud map and / or the target raster map. The preset afterimage filtering rules are associated with the feature parameters.

[0010] A second aspect of this application provides a mobile robot navigation map generation apparatus, comprising:

[0011] The acquisition module is used to acquire the original point cloud map and the original point cloud trajectory, which are obtained by the acquisition device through the real-time positioning and map building module.

[0012] The processing module is used to perform parameter processing on the original point cloud map and the original point cloud trajectory to obtain an intermediate point cloud map and an intermediate point cloud trajectory, wherein the intermediate point cloud map and the intermediate point cloud trajectory contain feature parameters for filtering out afterimages.

[0013] The filtering module is used to perform afterimage filtering processing on the intermediate point cloud map according to the intermediate point cloud trajectory and the preset afterimage filtering rules to obtain the target point cloud map and / or the target raster map. The preset afterimage filtering rules are associated with the feature parameters.

[0014] A third aspect of this application provides a mobile robot that generates a target point cloud map and / or a target grid map by means of the method described in the first aspect above.

[0015] A fourth aspect of this application provides a terminal device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method described in the first aspect above.

[0016] A fifth aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method described in the first aspect above.

[0017] A sixth aspect of this application provides a computer program product that, when run on a terminal device, causes the terminal device to execute the method described in the first aspect.

[0018] Compared with the prior art, the embodiments of this application have the following advantages:

[0019] In this embodiment, the acquisition device can acquire original point cloud maps and original trajectory maps within a preset area through a real-time positioning and mapping module. The original point cloud maps may contain afterimages that affect map accuracy. To remove these afterimages, the original point cloud maps and original point cloud trajectories can be parameter-processed to obtain intermediate point cloud maps and intermediate point cloud trajectories. This allows the intermediate point cloud maps and intermediate point cloud trajectories to contain afterimage-filtering feature parameters. Based on these afterimage-filtering feature parameters in the intermediate point cloud trajectories, and preset afterimage-filtering rules, afterimages can be filtered out from the intermediate point cloud maps. This results in a more accurate target point cloud map and / or target raster map based on the afterimage-filtered point cloud map. This embodiment can automatically remove afterimages from the point cloud map based on the point cloud trajectory, eliminating the need for manual editing of the afterimages. The method is simple and can produce a clearer and more accurate point cloud map or raster map based on the afterimage-filtered point cloud map, improving positioning accuracy. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0021] Figure 1 This is a flowchart illustrating the steps of a mobile robot navigation map generation method provided in an embodiment of this application;

[0022] Figure 2 This is a schematic diagram of a point cloud trajectory provided in an embodiment of this application;

[0023] Figure 3 This application provides an embodiment of a method for generating a two-grid map using existing technology;

[0024] Figure 4 This is a two-dimensional raster map generated using the scheme in this application, as provided in an embodiment of this application;

[0025] Figure 5 This is a flowchart illustrating the steps of another mobile robot navigation map generation method provided in this application embodiment;

[0026] Figure 6 This is a schematic diagram of a tilted point cloud map provided in an embodiment of this application;

[0027] Figure 7 This is a schematic diagram of a tilt-corrected point cloud map provided in an embodiment of this application;

[0028] Figure 8 This application provides an embodiment of a tilted point cloud map;

[0029] Figure 9This is a schematic diagram of a tilt-corrected point cloud map provided in an embodiment of this application;

[0030] Figure 10 This is a flowchart of a ghosting filtering method provided in an embodiment of this application;

[0031] Figure 11 This is a flowchart illustrating the generation process of a raster map according to an embodiment of this application;

[0032] Figure 12 This is a schematic diagram of a mobile robot navigation map generation device provided in an embodiment of this application;

[0033] Figure 13 This is a schematic diagram of a terminal device provided in an embodiment of this application. Detailed Implementation

[0034] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0035] If the point cloud map contains ghosting, using the point cloud map as input to the positioning module may result in a large matching error in areas with ghosting of people, thus affecting the final positioning accuracy. At the same time, the two-dimensional raster map generated using this point cloud map will also contain noise from ghosting of people. If it is not manually edited, the parts with noise from ghosting of people will be judged as obstacles by the global planning algorithm, thus making it impossible to plan the target path. This problem is particularly obvious in indoor scenes.

[0036] Based on this, this application proposes a method for generating navigation maps for mobile robots, which is used to remove ghosting in point cloud maps.

[0037] The technical solution of this application will be described below through specific embodiments.

[0038] This application embodiment can filter out ghosting in point cloud maps based on trajectory, thereby improving map accuracy. When applying this embodiment, for areas requiring navigation map determination, a data acquisition device can be used to collect point cloud maps and point cloud trajectories. Based on the acquired point cloud maps and point cloud trajectories, the terminal device can generate a navigation map using the method described in this application embodiment.

[0039] The mobile robot navigation map generation method provided in this application embodiment can be executed by terminal devices such as mobile phones, tablets, wearable devices, in-vehicle devices, augmented reality (AR) / virtual reality (VR) devices, laptops, ultra-mobile personal computers (UMPCs), netbooks, and personal digital assistants (PDAs). This application embodiment does not impose any restrictions on the specific type of terminal device.

[0040] For example, the mobile robot navigation map generation method in this application embodiment can be applied to an in-vehicle terminal. The in-vehicle terminal can obtain the original point cloud map and the original point cloud trajectory through the vehicle's real-time positioning and map building module. Then, based on the original point cloud map and the original point cloud trajectory, the mobile robot navigation map generation method in this application embodiment is executed to obtain a map. Based on the obtained map, the vehicle's positioning module can perform positioning or navigation.

[0041] For example, the mobile robot navigation map generation method in this application embodiment can be applied to a mobile robot. The mobile robot can obtain the original point cloud map and the original point cloud trajectory in the target area, and then execute the mobile robot navigation map generation method in this application embodiment based on the original point cloud map and the original point cloud trajectory to obtain a navigation map. The navigation map can be used for the mobile robot to locate itself.

[0042] In the following description of the solution, the terminal device is a mobile robot.

[0043] Reference Figure 1 The diagram illustrates a step-by-step flowchart of a mobile robot navigation map generation method according to an embodiment of this application, which may include the following steps:

[0044] S101, Obtain the original point cloud map and the original point cloud trajectory, which are obtained by the acquisition device through the real-time positioning and map building module.

[0045] The method in this embodiment can be applied to mobile robots, which can generate navigation maps based on the method in this embodiment.

[0046] The aforementioned data acquisition device can be a mobile robot or vehicle, etc. This device can move within a preset area to collect data, and the collected data can be output through a real-time localization and mapping (RTL) module. The RTL module can output a raw point cloud map of the preset area. This raw point cloud map can include multiple point cloud points, and these points can interact to delineate the shapes of buildings or other objects within the preset area. While collecting data within the preset area, the data acquisition device can travel along a specific trajectory. During this travel, the device can record the coordinates of its center position at preset intervals. Each coordinate can serve as a trajectory point, and the recorded trajectory points form the raw point cloud trajectory, which represents the device's movement during the data acquisition process.

[0047] In one possible implementation, the data acquisition device may include a real-time positioning and mapping module. The data acquisition device can receive control signals and move within a preset area in response to the control signals. During the movement, the data acquisition device can collect data through other data acquisition devices such as radar devices, camera devices, and sensor devices, and input the data into the real-time positioning and mapping module. The real-time positioning and mapping module can output a point cloud map and point cloud trajectory based on the input data and stored algorithms.

[0048] In one possible implementation, the mobile robot can obtain raw point cloud map data and raw point cloud trajectory data through its own real-time localization and mapping module, or it can receive raw point cloud map data and raw point cloud trajectory data collected by other devices.

[0049] It should be noted that the original point cloud map and the original point cloud trajectory are generally acquired by the same acquisition device in the same time and space. The original point cloud trajectory can be inside the original point cloud map, that is, the original point cloud trajectory and the original point cloud map can be in the same coordinate system.

[0050] S102, perform parameter processing on the original point cloud map and the original point cloud trajectory to obtain an intermediate point cloud map and an intermediate point cloud trajectory, wherein the intermediate point cloud map and the intermediate point cloud trajectory contain feature parameters for filtering out afterimages.

[0051] The aforementioned feature parameters are used for ghosting removal in point cloud maps. The feature parameters of the intermediate point cloud map can include the local volume density values ​​of each point in the intermediate point cloud map. The feature parameters of the intermediate point cloud trajectory can include the three-dimensional pose of each trajectory point in the intermediate point cloud trajectory.

[0052] In the original point cloud map, multiple point cloud points can constitute buildings or other objects within a preset area, and the afterimages in the point cloud map also exist in the form of point cloud points. For personnel afterimages, the density of point cloud points is generally low, while the density of buildings or other fixed devices within the area is generally high. Therefore, afterimage filtering can be performed based on the local volume density of point cloud points. Based on this, in this embodiment of the application, the mobile robot can perform parameter processing on the original point cloud points, calculate the local volume density of each point cloud point, and add the local volume density value as a feature parameter to the original point cloud map to obtain an intermediate point cloud map.

[0053] The local volume density of a point cloud can be determined by the number of points in the neighboring area of ​​that point cloud. For example, fixed obstacles such as walls are generally static and may be scanned multiple times, so the local volume density value of the corresponding point cloud is relatively large. In contrast, people are dynamic and are scanned less frequently, so the local volume density value of the point cloud corresponding to the person's afterimage is relatively small.

[0054] In one possible implementation, the mobile robot can calculate the local volume density value corresponding to each point in the original point cloud map using the following formula:

[0055] ρ=N / (4 / 3*π*R 3 )

[0056] R = (W + L) / 2

[0057] Where ρ is the local volume density value of the point cloud, R is the radius of the sphere centered on the point cloud, W is the length of the acquisition device, L is the width of the acquisition device, and N is the number of point cloud points in the sphere centered on the point cloud.

[0058] The original point cloud map is acquired by personnel-controlled acquisition equipment. During the acquisition process, personnel may stand in the trajectory of the acquisition equipment, resulting in personnel ghosting in the point cloud map. However, areas that the acquisition equipment can pass through indicate areas without point clouds. Therefore, personnel ghosting in the point cloud map can be determined based on the locations where the acquisition equipment can pass through. The position of the acquisition equipment in the point cloud map can be determined by the 3D coordinates of its corresponding trajectory points. Therefore, the mobile robot can perform parameter processing on the original point cloud trajectory, adding 3D pose as a feature parameter to each trajectory point in the original point cloud trajectory, thereby obtaining an intermediate point cloud map.

[0059] It should be noted that the steps for calculating the parameters of the original point cloud map and the steps for calculating the trajectory of the original point cloud can be performed concurrently by the mobile robot. The intermediate trajectory point cloud is the travel trajectory of the data acquisition device, and therefore, it can be determined based on the intermediate trajectory point cloud.

[0060] When performing image retention filtering based on areas that the acquisition device can pass through in a point cloud map, it is necessary to determine all possible areas that the acquisition device can pass through. This determination is based on trajectory points in the intermediate point cloud trajectory. Each trajectory point corresponds to a region for one acquisition device. If the distance between two adjacent trajectory points is less than or equal to the length of the acquisition device, the regions corresponding to the acquisition device determined by these two adjacent trajectory points can be connected, thus covering all areas that the acquisition device can pass through. However, if the distance between two adjacent trajectory points is greater than the length of the acquisition device, the regions corresponding to the acquisition device determined by these two adjacent trajectory points cannot be connected, potentially resulting in unidentified areas that the acquisition device could pass through.

[0061] Therefore, based on the distance between adjacent trajectory points in the intermediate point cloud trajectory, it can be determined whether interpolation processing is needed. The mobile robot can determine the distance between two adjacent trajectory points in the intermediate point cloud trajectory; if the distance between two trajectory points is greater than the length of the acquisition device, interpolation processing can be performed on the intermediate point cloud trajectory, that is, inserting a new trajectory point between the two trajectory points whose distance is greater than the length of the acquisition device. The purpose of interpolation processing is to make the distance between each trajectory point in the intermediate point cloud trajectory less than or equal to the length of the acquisition device, thereby allowing the identification of areas that the acquisition device can traverse in the point cloud map based on each trajectory point in the intermediate point cloud trajectory. The distance between two connected trajectory points in the intermediate point cloud trajectory can be determined based on the three-dimensional poses corresponding to the two trajectory points.

[0062] Figure 2 This is a schematic diagram of a point cloud trajectory provided in an embodiment of this application. Figure 2 The black dots in the image are the trajectory points collected by the acquisition device, for example... Figure 2 In this diagram, a, b, and d represent the trajectory points acquired by the acquisition device. Since the distance between trajectory points a and b is short (less than the length of the acquisition device), no interpolation is needed between them. However, the distance between trajectory points b and c is greater than the length of the acquisition device; therefore, interpolation is required between them. Figure 2 The hollow triangle in the diagram represents the new trajectory point to be inserted.

[0063] S103, perform afterimage filtering on the intermediate point cloud map according to the intermediate point cloud trajectory and the preset afterimage filtering rules to obtain the target point cloud map and / or target raster map, wherein the preset afterimage filtering rules are associated with the feature parameters.

[0064] Based on the aforementioned feature parameters and preset ghosting filtering rules, ghosting can be filtered out from intermediate point cloud maps.

[0065] The trajectory points in the intermediate point cloud trajectory have a three-dimensional pose, and the position corresponding to the trajectory point is the center position of the acquisition device. Therefore, based on the three-dimensional pose of the trajectory points and the model parameters of the acquisition device, the corresponding 3D bounding box of the acquisition device can be determined. For example, based on the model parameters of the acquisition device, the length, width, and height of the acquisition device can be determined, and a 3D bounding box with the same length, width, and height as the acquisition device can be determined with the trajectory point as the center and according to the three-dimensional pose of the trajectory point.

[0066] Reference Figure 2 The corresponding 3D bounding box can be determined based on the original trajectory points and the newly inserted trajectory points. Figure 2 The image only shows the case where the edges of the 3D frames defined by two connected trajectory points meet. The 3D frames defined by two connected trajectory points may also partially overlap.

[0067] Mobile robots can filter out ghosting from intermediate point cloud maps based on 3D bounding boxes and local volume density values, thereby obtaining the target point cloud map.

[0068] The area corresponding to the 3D frame is the area that the acquisition device can pass through in the point cloud map. Areas that the acquisition device can pass through should be roads in the map and should not be identified as obstacles. Therefore, it can be determined that the point cloud points within the area corresponding to the 3D frame are afterimages, and the point cloud points within the area corresponding to the 3D frame can be filtered out from the intermediate point cloud map.

[0069] Furthermore, afterimage points may also exist within a certain range of the data acquisition device. While the area near the data acquisition device's travel zone is generally a road area, there are also cases where the data acquisition zone is close to buildings. Therefore, afterimage points can be identified near the data acquisition device's travel zone based on local volume density values. The mobile robot can identify point cloud points with local volume density values ​​less than a preset threshold within the vicinity of the bounding box as afterimage points.

[0070] For example, the mobile robot can identify each target point cloud point whose distance from the outer surface of the 3D frame is less than a preset length, and then identify point cloud points whose local volume density value is less than a preset threshold as afterimage points.

[0071] For example, the mobile robot can also determine the circumscribed sphere corresponding to the 3D bounding box, then determine the target point cloud points that exist inside the circumscribed sphere but not inside the 3D bounding box, and identify point cloud points whose local volume density value is less than a preset threshold as afterimage points. That is:

[0072] Mobile robots can identify point cloud points in intermediate point cloud maps that meet the following conditions as afterimage points:

[0073]

[0074] Where, ρ n Let ρ be the local volume density of the point cloud. lmin R is the volume density threshold. n R is the distance between the center of the circumsphere corresponding to the 3D frame and the point cloud points. sbox R is a preset radius threshold. scan Let be the radius of the circumscribed sphere of the solid frame.

[0075] After filtering out afterimage points inside and around the 3D bounding box, the mobile robot can obtain a target point cloud map with afterimages removed. Based on the target point cloud map, the mobile robot can perform localization and navigation.

[0076] In one possible implementation, based on the target point cloud map, the mobile robot can also determine the target grid map. The target grid map can be a two-dimensional map.

[0077] Mobile robots can extract target point clouds within a preset height range from a target point cloud map. This preset height can be the passage height required by a typical mobile robot or vehicle. For example, in a navigation map used for vehicle positioning and navigation, the extracted height range can be the highest point from the ground to the vehicle. Extracting target point clouds within a preset height range from the target point cloud map avoids points at higher positions affecting the accuracy of the raster map. For instance, in situations where buildings at higher elevations are prominent (i.e., the distance between higher buildings is shorter than the distance between lower buildings), directly projecting the point cloud map onto a two-dimensional plane would result in a narrower perceived road. Therefore, to improve the accuracy of the two-dimensional raster map, points within a height range that the mobile robot or vehicle can pass through can be extracted from the target point cloud map.

[0078] Based on the 3D pose of each trajectory point in the intermediate trajectory point cloud, the height value corresponding to each trajectory point can be determined, thereby allowing the calculation of the average height value of the trajectory points. The height corresponding to a trajectory point is the height of the center position of the acquisition device from the ground, which is equivalent to determining the height of the ground.

[0079] If the horizontal plane in the target point cloud is parallel to the actual horizontal plane, meaning the point cloud is not tilted, the height range to be intercepted can be directly determined from the target point cloud. The intercepted height range can be determined based on the average height corresponding to the trajectory points. For example, if the center of the acquisition device is at a height of h above the ground, the average height corresponding to the trajectory points is x, and the height limit within the current area is y, then the point cloud to be intercepted has a height range of (xh, x-h+y).

[0080] If the horizontal plane in the target point cloud is not parallel to the actual horizontal plane, meaning the point cloud is tilted, the minimum and maximum height values ​​to be cropped can be determined based on the required height range and radar calibration parameters. The point cloud within this range can then be selected as the target point cloud. The radar calibration parameters can be used to calibrate the correspondence between the corresponding heights in the point cloud map and the actual scene heights.

[0081] After determining the target point cloud, the mobile robot can project it onto a two-dimensional plane to obtain a projected point cloud. Points in the target point cloud will have corresponding points in the projected point cloud. In the two-dimensional plane, the mobile robot can divide the projected point cloud into multiple square grids according to a preset resolution. Based on whether each grid contains point points, the mobile robot determines whether to mark it as an obstacle. If a grid contains point points, the mobile robot can mark it as an obstacle; otherwise, it can choose not to. After marking the grids, the mobile robot can perform image closing operations on each marked grid to generate a grid map.

[0082] In this application, ghosting in point cloud maps can be filtered out, thereby improving the accuracy of the point cloud maps. Based on the point cloud map after removing ghosting, a more accurate two-dimensional raster map can be obtained.

[0083] To visually demonstrate the technical effects of this embodiment, this embodiment provides a raster map established using existing technology and a raster map established using the solution in this application, for reference. Figure 3 and Figure 4 . Figure 3 This application provides an embodiment of a method for generating a two-grid map using existing technology. Figure 4 This is a two-dimensional raster map generated using the scheme described in this application, as provided in an embodiment of this application. (Comparison) Figure 3 and Figure 4 It is evident that the two-dimensional raster map generated using the scheme described in this application is clearer and more accurate.

[0084] Reference Figure 5 This illustration shows a flowchart of another mobile robot navigation map generation method provided in an embodiment of this application, which may specifically include the following steps:

[0085] S501, acquire the original point cloud map and the original point cloud trajectory, which are obtained by the acquisition device through the real-time positioning and map building module.

[0086] S502, perform principal component analysis on the original point cloud map to obtain the pose matrix of the original point cloud map, and the pose matrix is ​​used to represent the pose of the original point cloud map relative to the ground.

[0087] The original point cloud map may have a certain degree of tilt. For example, if the ground in the preset area is level, objects at the same height in the actual scene should also have the same height in the original point cloud map. However, due to errors in the acquisition or data processing, objects at the same height in the actual scene may have different heights in the original point cloud map, indicating that the original point cloud map has a certain degree of tilt. Because the real-time positioning and construction module of the acquisition device may have certain errors, or the real-time positioning and construction solution provided by the supplier may have some flaws, the original point cloud map output by the real-time positioning and construction module may have a certain degree of tilt. The tilt of the original point cloud map may cause a large posture error of objects in the point cloud map, leading to incorrect or inaccurate calculation of obstacle positions, thus resulting in pose errors during positioning and affecting positioning accuracy and path tracking performance.

[0088] For example, if the ground in the area where the navigation map is to be generated is level, but the ground in the point cloud map is sloping, this will cause the buildings in the point cloud map to be tilted. Directly using this point cloud map for navigation will result in a deviation between the road areas in the map and the actual road areas due to the tilt angle, leading to navigation errors. Generating a 2D raster map based on a tilted point cloud map will also inevitably be inaccurate. Therefore, mobile robots can correct the tilt of the original point cloud map before generating the actual map.

[0089] When performing tilt correction, it is first necessary to determine whether the original point cloud map is tilted. The mobile robot can determine whether the original point cloud map is tilted based on its pose.

[0090] For example, a mobile robot can perform principal component analysis based on the original point cloud map to obtain the pose matrix of the original point cloud map, which is used to characterize the pose of the original point cloud map relative to the ground.

[0091] S503, Based on the pose matrix, determine whether the original point cloud map is tilted.

[0092] The pose matrix can include tilt angle information, which represents the angle of the original point cloud map relative to the horizontal plane of the corresponding region. The mobile robot can use the tilt angle information to determine whether the original point cloud map is tilted relative to the ground.

[0093] S504, if the original point cloud map is tilted, then the tilt correction is performed on the original point cloud map and the original point cloud trajectory based on the pose matrix.

[0094] If the original point cloud map is tilted relative to the ground, the mobile robot can correct the tilt of the original point cloud map.

[0095] The original point cloud trajectory is matched with the original point cloud map. The original point cloud trajectory can be inside the original point cloud map. Therefore, when the original point cloud map is tilted, the mobile robot can also tilt the original point cloud trajectory, so that the tilted original point cloud map and the original point cloud trajectory still match after tilt correction.

[0096] For example, after the mobile robot determines the tilt angle of the original point cloud map relative to the opposite side based on the pose matrix, it can rotate the original point cloud map and the original point cloud trajectory by the corresponding angle using the coordinate system corresponding to the pose matrix as a reference, so that the ground corresponding to the original point cloud map and the original point cloud trajectory is parallel to the ground in the actual scene.

[0097] To better illustrate the technical effects of S502-S504, the embodiments of this application are described below. Figure 6-9 The map is displayed before and after tilt correction. Figure 6 This is a schematic diagram of a tilted point cloud map provided in an embodiment of this application; Figure 7 This is a schematic diagram of a tilt-corrected point cloud map provided in an embodiment of this application; Figure 8 This application provides an embodiment of a tilted point cloud map; Figure 9 This is a schematic diagram of a tilt-corrected point cloud map provided in an embodiment of this application. Figure 6-9 In the diagram, the black dot at the center of the circle can be used to represent the origin of the coordinate system. Figure 6-9 There are straight lines, a first dashed line, and a second dashed line. The first dashed line is composed of uniformly spaced small line segments, and the second dashed line is composed of two connected small line segments with two points in between. That is, in... Figure 6 and Figure 8 In the middle, the horizontal line is the first dashed line, and the vertical line is the second dashed line; in Figure 7 and Figure 9 In the diagram, the circular curve is the first dashed line, and the vertical line is the second dashed line. Within the coordinate system, there can be planes represented by the X, Y, and Z axes, with the straight line, the first dashed line, and the second dashed line each representing one of these planes. In one possible implementation, the plane formed by the straight line, the second dashed line, and the origin can be used to represent the plane containing the ground, namely the XOY plane.

[0098] Figure 6 and Figure 8 The point cloud in the image is equivalent to the point cloud seen in a longitudinal section. From Figure 6 As can be seen, the point cloud is tilted relative to the XOY plane. Figure 6 The point cloud map corresponding to the point cloud in the image is... Figure 7 , Figure 7 It's like looking down at a map on the ground. Figure 7 The point cloud map in the image has a certain deviation, so the point cloud is tilted. When the point cloud is mapped to the XOY plane, it will inevitably lead to the deviation of the point cloud map. Figure 8 This is the point cloud after tilt correction. Figure 8 Compared to Figure 6 The point cloud becomes longer because Figure 6 The point clouds in the image originally had some tilt. After tilt correction, the point clouds parallel to the horizontal plane are naturally longer than the tilted point clouds. From Figure 8 As can be seen, after tilt correction, the point cloud is basically parallel to the XOY plane. Based on Figure 8 The map obtained from the point cloud is Figure 9 Compared to Figure 9 and Figure 7 As can be seen, the tilt of the point cloud will cause map deviation: Figure 7 The map in the middle is relative to Figure 9 The map in the image has a significant deviation from the origin of the coordinate system. Therefore, in this embodiment of the application, before determining the target map, the point cloud can be tilted and corrected according to the steps in S502-S504, so that the generated map is more accurate.

[0099] S505, perform parameter processing on the tilt-corrected original point cloud map and the tilt-corrected original point cloud trajectory to obtain an intermediate point cloud map and an intermediate point cloud trajectory, wherein the intermediate point cloud map and the intermediate point cloud trajectory contain feature parameters for filtering out afterimages.

[0100] It should be noted that if the original point cloud map and the original point cloud trajectory are not tilted, this step can be: performing parameter processing on the original point cloud map and the original point cloud trajectory to obtain an intermediate point cloud map and an intermediate point cloud trajectory, wherein the intermediate point cloud map and the intermediate point cloud trajectory contain feature parameters for filtering out afterimages.

[0101] S506, according to the intermediate point cloud trajectory and the preset afterimage filtering rule, the intermediate point cloud map is subjected to afterimage filtering processing to obtain the target point cloud map, wherein the preset afterimage filtering rule is associated with the feature parameters.

[0102] In this embodiment, S504-S506 are similar to S101-S103 in the previous embodiment and can be referenced interchangeably, so they will not be described in detail here.

[0103] S507, Extract the target point cloud within a preset height range from the target point cloud map.

[0104] Because the point cloud map is tilted and corrected when filtering out afterimages in the original point cloud map, the mobile robot can directly extract the target point cloud within a preset height range from the target point cloud map when generating a raster map based on the point cloud map after filtering out afterimages, and thus generate a raster map based on the target point cloud.

[0105] S508, Based on the target point cloud, generate a raster map of the area corresponding to the original point cloud map.

[0106] In this embodiment, the mobile robot can correct the tilt of the original point cloud map and the original point cloud trajectory before filtering out the afterimages in the original point cloud map. This allows the original point cloud map and the original point cloud trajectory to match the actual ground, making the generated target point cloud and target grid map match the actual ground more accurately, thereby improving the positioning accuracy of the map.

[0107] It should be noted that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0108] The method for generating a navigation map for a mobile robot in this embodiment can filter out afterimages in a point cloud map based on the point cloud trajectory, thereby obtaining a point cloud map after removing afterimages. A raster map can then be obtained based on the point cloud map after removing afterimages. To better illustrate the solution in this application, the method is divided into several steps: filtering out personnel afterimages from the point cloud map and generating a raster map chain based on the point cloud map after removing afterimages.

[0109] Figure 10 This is a schematic diagram of a process for filtering out personnel ghosting in a point cloud map, as provided in an embodiment of this application. Figure 10 As shown, filtering out human ghosting in a point cloud map can include the following steps:

[0110] S1001 calculates the overall PCA of the point cloud map to obtain the pose matrix of the point cloud map.

[0111] Point cloud maps and point cloud trajectories can be collected by personnel-controlled vehicles within a preset area and obtained through a real-time localization and construction module. Each point cloud point and trajectory point in the collected point cloud map can include position coordinates (x, y, z). Ideally, the collected point cloud should be horizontal, but in reality, the point cloud may be tilted relative to the horizontal plane. For the original point cloud map obtained through the real-time localization and construction module, it can be determined whether the point cloud map is tilted. The mobile robot can first calculate the overall PCA of the point cloud map to obtain the point cloud map. Figure 4 x4 pose matrix M.

[0112] Since point cloud trajectories can be included within point cloud maps and their positional relationships match each other, it is only necessary to determine whether the point cloud map is tilted to determine whether the point cloud trajectory is tilted.

[0113] S1002, use the relative position attitude matrix to correct the point cloud map and point cloud trajectory.

[0114] If the tilt of the point cloud map is determined based on the pose matrix, the mobile robot can use the pose matrix M to correct the tilt of the point cloud map and point cloud trajectory, thus obtaining the tilt-free point cloud map and point cloud trajectory.

[0115] Of course, if the point cloud map is determined to be untilted based on the pose matrix, then steps S1001 and S1002 are not required, and subsequent steps can be executed directly.

[0116] S1003, calculate the volume density of the point cloud map and add a local volume density value to each point in the point cloud map.

[0117] The local volume density value of each point cloud point can be used to characterize the number of point cloud points within a preset range from that point cloud point. For solid obstacles, since the object is fixed, the mapping scan will acquire a large number of measurement points on the object's surface, resulting in a higher point cloud density for solid obstacles. However, for pseudo-obstacles such as moving human figures, their positions are not fixed, and fewer sampling points are taken in their temporary areas during the mapping scan, resulting in a lower point cloud density for pseudo-obstacles such as moving human figures. Therefore, in this embodiment, the volume density of the point cloud map can be calculated, and a local volume density value can be added to each point cloud point in the point cloud map, thereby enabling the identification of afterimage points based on volume density.

[0118] The formula for calculating the local volume density of point cloud points has been described in the previous embodiments and will not be repeated here.

[0119] After adding a local volumetric density value to each point cloud point, the point cloud point can have four parameter values: (x, y, z, ρ). Based on the local point cloud density value, subsequent ghosting filtering can be performed. For example, the volumetric density of a wall point cloud is generally greater than the point cloud density of a ghost image because people are dynamic, while walls are static. Static walls are scanned multiple times, resulting in a higher point cloud density.

[0120] S1006 calculates the attitude value of each trajectory point by using adjacent trajectory points before and after it, and adds the attitude value to the stored data structure.

[0121] Point cloud trajectories reflect the driving trajectory of the data acquisition device. By analyzing adjacent trajectory points within the point cloud trajectory, the 3D pose value of each trajectory point can be calculated and added to its storage data structure, giving each trajectory point 10 attribute values ​​(x, y, z, roll, pitch, yaw). The 3D pose value of a trajectory point can include coordinate position information and angle information, where roll, pitch, and yaw represent the angle information. Based on the completed 3D pose, the vehicle body space of the corresponding data acquisition device in the point cloud map can be calculated.

[0122] S1005, interpolate the trajectory points with large spacing so that the distance between two adjacent trajectory points is less than the length of the vehicle body.

[0123] This embodiment performs ghosting filtering based on vehicle body space. Therefore, it is necessary to determine all vehicle body spaces in the point cloud map, and these spaces should be continuous. However, when the distance between trajectory points is too long, the determined vehicle body spaces cannot be continuous, resulting in gaps in the vehicle body space. Consequently, it becomes impossible to identify whether the point cloud in these gaps is a ghosting effect. Therefore, interpolation processing can be performed on the trajectory points to ensure that the distance between adjacent trajectory points in the point cloud trajectory is less than or equal to the length of the acquisition device.

[0124] Mobile robots can interpolate trajectory points with large gaps, so that the vehicle body cuboid can be connected between every two trajectory points.

[0125] S1006, traverse the corrected trajectory points and remove the point cloud points within the vehicle body cuboid corresponding to the trajectory points.

[0126] The location of the vehicle body space indicates that it is a passable location, meaning that this area should not be identified as an obstacle, and therefore all afterimage points within the vehicle body space can be eliminated.

[0127] For the point cloud trajectory after interpolation, each trajectory point can be traversed. Based on the 3D pose corresponding to the trajectory point and combined with the model of the acquisition device, the rectangular bounding box occupied by the acquisition device at that trajectory point can be calculated. For each trajectory point, the corresponding rectangular bounding box can be determined, and the afterimages within the rectangular bounding box can be removed. It should be noted that the vehicle space can include the vehicle surface.

[0128] S1007, traverse the corrected trajectory points and propose points whose volume density in the vicinity of the vehicle body is less than the threshold corresponding to the trajectory point.

[0129] Furthermore, there may be afterimages of people within the vicinity of the cuboid frame. In this case, afterimage points can be removed based on local point cloud density values. The steps for determining afterimage points have been described in the previous embodiments and will not be repeated here.

[0130] After removing the afterimages of people inside and near the rectangular frame corresponding to the data acquisition device, the processed point cloud map and point cloud trajectory can be output. The positioning module can then perform localization based on the processed point cloud map and point cloud trajectory.

[0131] In this embodiment, the spatial location of the acquisition device in the point cloud map is determined by the point cloud specification, thereby filtering out personnel ghosting in the 3D point cloud map, reducing the probability of mismatch of personnel areas in the original point cloud map, and improving the positioning accuracy of the positioning module that uses the 3D point cloud map as input.

[0132] Figure 11 This is a schematic diagram illustrating a process for generating a raster map from a point cloud map after filtering out point afterimages, as provided in an embodiment of this application. (Refer to...) Figure 11 The process of generating a raster map from a point cloud map after filtering out point ghosting can include:

[0133] S1101, using the corrected trajectory point cloud to calculate the average z-value of the trajectory point cloud.

[0134] The trajectory points can be the center position of the data acquisition device. The average z-value of the corrected trajectory point cloud is equivalent to the height of the center position of the data acquisition device. The average z-value reflects the average height of the data acquisition device when it travels within the area. The acquired point cloud map may be very high, but in reality, the generated map does not need a point cloud that is much higher; only a point cloud within the passable height range is required. Therefore, the average z-value of the trajectory point cloud can be calculated based on the trajectory points.

[0135] S1102, using a point cloud map with filtered personnel afterimages, calculates the maximum and minimum Z values ​​for cropping based on the average Z value and radar calibration parameters, and cuts out the point cloud within the range of these Z values.

[0136] After determining the average z-value, the point cloud can be extracted from the point cloud map after removing afterimages based on the average z-value.

[0137] If the point cloud map has already been tilted when filtering out ghost images, so that the tilt-corrected point cloud map has been corrected by the transformation matrix to be basically aligned with the XoY plane or the point cloud map itself is not tilted, then the corresponding cut-off height Z value can be determined directly based on the average z value, and then the point cloud below the height Z value can be cut from the point cloud map after ghost image filtering.

[0138] If the point cloud map is tilted but no tilt correction is performed, the maximum and minimum Z values ​​for cropping can be calculated based on the average Z value and radar calibration parameters, and the point cloud map within the range of these Z values ​​can be cropped to generate a driving height sub-point cloud.

[0139] S1103, Project the extracted point cloud onto the XOY plane, and mark each grid point containing point cloud points as an obstacle according to the grid map resolution.

[0140] The extracted point cloud can be projected onto a two-dimensional plane, namely the XOY plane. Then, it can be divided into grids, and each grid point containing a point cloud point can be marked as an obstacle according to the grid map resolution. For example, an obstacle could be a wall.

[0141] S1104 performs a closing operation on the initially defined raster map and outputs the final raster map.

[0142] Closing operations can be performed on the calibrated raster map to output the final raster map.

[0143] In this embodiment, a two-dimensional raster map can be generated from the point cloud map after filtering out afterimages, directly filtering out false obstacles caused by afterimages of people, eliminating the need for manual editing of the two-dimensional raster map to remove obstacles, and reducing the user's learning cost and operation difficulty.

[0144] This application also provides a mobile robot that can perform ghosting filtering on a point cloud map using the steps described in the above method embodiments, and generate a target map or a target grid map based on the point cloud map after ghosting filtering. The mobile robot may include a positioning module, and based on the positioning module, the mobile robot can use the target map or the target grid map for navigation.

[0145] In one possible implementation, the mobile robot may include a real-time localization and mapping (RTL) module. The mobile robot can receive control signals and respond to these signals to move within a preset area. During movement, the mobile robot can collect data through radar, cameras, sensors, and other data acquisition devices, and input this data into the RTL module. The RTL module can output a point cloud map and a point cloud trajectory based on the input data and stored algorithms. Based on the point cloud trajectory and point cloud map, the mobile robot can obtain a point cloud map and / or a raster map after removing afterimages, following the steps in the various method embodiments described above.

[0146] Reference Figure 12 The diagram illustrates a mobile robot navigation map generation device according to an embodiment of this application, which may specifically include an acquisition module 1201, a processing module 1202, and a filtering module 1203, wherein:

[0147] The acquisition module 1201 is used to acquire the original point cloud map and the original point cloud trajectory, which are obtained by the acquisition device through the real-time positioning and map building module.

[0148] Processing module 1202 is used to perform parameter processing on the original point cloud map and the original point cloud trajectory to obtain an intermediate point cloud map and an intermediate point cloud trajectory, wherein the intermediate point cloud map and the intermediate point cloud trajectory contain feature parameters for filtering out afterimages.

[0149] The filtering module 1203 is used to perform afterimage filtering processing on the intermediate point cloud map according to the intermediate point cloud trajectory and the preset afterimage filtering rules to obtain the target point cloud map and / or the target raster map. The preset afterimage filtering rules are associated with the feature parameters.

[0150] In one possible implementation, the processing module 1202 includes:

[0151] The feature parameter calculation submodule is used to calculate the local volume density value of each point in the original point cloud map, add the local volume density value as the feature parameter to the original point cloud map to obtain the intermediate point cloud map; and calculate the three-dimensional pose of each trajectory point in the original point cloud trajectory, add the three-dimensional pose as the label parameter to the original point cloud trajectory to obtain the intermediate point cloud trajectory.

[0152] In one possible implementation, the aforementioned feature parameter calculation submodule includes:

[0153] The local volume density calculation unit is used to calculate the local volume density value corresponding to each point in the original point cloud map using the following formula:

[0154] ρ=N / (4 / 3*π*R 3 )

[0155] R = (W + L) / 2

[0156] Wherein, ρ is the local volume density value of the point cloud, R is the radius of the sphere centered on the point cloud, W is the length of the acquisition device, L is the width of the acquisition device, and N is the number of point cloud points in the sphere centered on the point cloud.

[0157] In one possible implementation, the above processing module further includes:

[0158] The distance determination submodule is used to determine the distance between two adjacent trajectory points of the intermediate point cloud trajectory;

[0159] The difference processing submodule is used to perform interpolation processing on the intermediate point cloud trajectory if the distance is greater than the length of the acquisition device, so that the distance between two adjacent trajectory points is less than the length of the acquisition device.

[0160] In one possible implementation, the filtering module 1203 includes:

[0161] The stereo frame determination submodule is used to determine the stereo frame corresponding to the acquisition device based on the three-dimensional pose of each trajectory point in the intermediate point cloud trajectory.

[0162] The filtering submodule is used to filter out afterimages from the intermediate point cloud map based on the stereo frame and the local volume density value to obtain the target point cloud map.

[0163] In one possible implementation, the above-mentioned filtering submodule includes:

[0164] The filtering unit is used to filter out all point cloud points in the area corresponding to the 3D frame in the intermediate point cloud map; and to determine the afterimage points from the point cloud points within a preset range from the outer surface of the 3D frame according to the local volume density value, and to filter out the afterimage points from the intermediate point cloud map to obtain the target point cloud map.

[0165] In one possible implementation, the filtering unit further includes:

[0166] The afterimage point determination subunit is used to determine the point cloud points in the intermediate point cloud map that meet the following conditions as the afterimage points:

[0167]

[0168] Where, ρ nρ is the local volume density of the point cloud points. lmin R is the volume density threshold. n R is the distance between the center of the circumscribed sphere corresponding to the 3D frame and the point cloud points. sbox R is a preset radius threshold. scan Let be the radius of the circumscribed sphere corresponding to the solid frame.

[0169] In one possible implementation, the filtering module 1203 includes:

[0170] The stereo frame determination submodule is used to determine the stereo frame corresponding to the acquisition device based on the three-dimensional pose of each trajectory point in the intermediate point cloud trajectory.

[0171] The target point cloud map determination submodule is used to perform afterimage filtering on the intermediate point cloud map based on the stereo frame to obtain the target point cloud map;

[0172] The target raster map generation submodule is used to generate a target raster map based on the target point cloud map.

[0173] In one possible implementation, the target raster map generation submodule mentioned above includes:

[0174] The target point cloud cropping unit is used to crop the target point cloud within a preset height range from the target point cloud map;

[0175] The target raster map generation unit is used to generate a raster map of the area corresponding to the original point cloud map based on the target point cloud.

[0176] In one possible implementation, the target raster map generation submodule mentioned above includes:

[0177] The interception distance determination unit is used to determine the interception distance range of the point cloud map based on the target point cloud map and the intermediate point cloud trajectory, wherein the distance range includes a maximum value and a minimum value;

[0178] The interception unit is used to intercept the target point cloud from the target point cloud map based on the interception distance range;

[0179] The target raster map generation unit is used to generate a raster map of the area corresponding to the original point cloud map based on the target point cloud.

[0180] In one possible implementation, the target raster map generation unit described above is configured to include:

[0181] The projection subunit is used to project the target point cloud onto a two-dimensional plane parallel to the ground of the region to obtain a projection plane point cloud.

[0182] Divide into grid sub-units to divide the projection plane point cloud into multiple square grids according to a preset resolution;

[0183] A marking subunit is used to determine whether to mark the grid as an obstacle based on whether there are point cloud points in each grid.

[0184] A sub-unit is generated to perform image closing operations on each of the marked grid cells to generate the grid map.

[0185] In one possible implementation, the above-mentioned device further includes:

[0186] The pose matrix calculation module is used to perform principal component analysis on the point cloud map to obtain the pose matrix of the point cloud map, and the pose matrix is ​​used to represent the pose of the point cloud map relative to the ground.

[0187] The tilt determination module is used to determine whether the point cloud map is tilted based on the pose matrix.

[0188] The correction module is used to correct the tilt of the point cloud map and the point cloud estimate based on the pose matrix if the point cloud map is tilted.

[0189] As the apparatus embodiments are basically similar to the method embodiments, they are described in a relatively simple manner. For relevant details, please refer to the description in the method embodiment section.

[0190] Figure 13 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Figure 13 As shown, the terminal device 130 of this embodiment includes: at least one processor 1300 ( Figure 13 The diagram shows only one processor, a memory 1301, and a computer program 1302 stored in the memory 1301 and executable on the at least one processor 1300, which, when executing the computer program 1302, implements the steps in any of the above-described method embodiments.

[0191] The terminal device 130 may be a computing device such as a desktop computer, laptop, handheld computer, or cloud terminal device. This terminal device may include, but is not limited to, a processor 1300 and a memory 1301. Those skilled in the art will understand that... Figure 13 This is merely an example of terminal device 130 and does not constitute a limitation on terminal device 130. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, etc.

[0192] The processor 1300 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0193] In some embodiments, the memory 1301 may be an internal storage unit of the terminal device 130, such as a hard disk or memory of the terminal device 130. In other embodiments, the memory 1301 may be an external storage device of the terminal device 130, such as a plug-in hard disk, SmartMediaCard (SMC), SecureDigital (SD) card, or FlashCard equipped on the terminal device 130. Furthermore, the memory 1301 may include both internal and external storage units of the terminal device 130. The memory 1301 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 1301 can also be used to temporarily store data that has been output or will be output.

[0194] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0195] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.

[0196] The embodiments described above are only used to illustrate the technical solutions of this application, and are not intended to limit it. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method for generating navigation maps for a mobile robot, characterized in that, include: The original point cloud map and the original point cloud trajectory are obtained by the acquisition device through the real-time positioning and map building module. The original point cloud map and the original point cloud trajectory are processed to obtain an intermediate point cloud map and an intermediate point cloud trajectory, which contain feature parameters for filtering out afterimages. Based on the intermediate point cloud trajectory and the preset afterimage filtering rules, the intermediate point cloud map is subjected to afterimage filtering processing to obtain the target point cloud map and / or the target raster map. The preset afterimage filtering rules are associated with the feature parameters. The step of parameter processing of the original point cloud map and the original point cloud trajectory to obtain the intermediate point cloud map and the intermediate point cloud trajectory includes: Calculate the local volume density value of each point in the original point cloud map, and add the local volume density value as the feature parameter to the original point cloud map to obtain the intermediate point cloud map; and calculate the three-dimensional pose of each trajectory point in the original point cloud trajectory, and add the three-dimensional pose as the feature parameter to the original point cloud trajectory to obtain the intermediate point cloud trajectory. The step of performing afterimage filtering processing on the intermediate point cloud map according to the intermediate point cloud trajectory and preset afterimage filtering rules to obtain the target point cloud map includes: Based on the three-dimensional pose of each trajectory point in the intermediate point cloud trajectory, the corresponding 3D bounding box of the acquisition device is determined; Based on the 3D frame and the local volume density value, the intermediate point cloud map is subjected to afterimage filtering to obtain the target point cloud map.

2. The method as described in claim 1, characterized in that, The calculation of the local volume density value of each point in the original point cloud map includes: The local volume density value corresponding to each point in the original point cloud map is calculated using the following formula: in, The local volume density value of the point cloud points. Let the radius of the sphere centered on the point cloud points be denoted as . The length of the acquisition device, Where N is the width of the acquisition device, and N is the number of point cloud points in the sphere centered on the point cloud points.

3. The method as described in claim 1, characterized in that, After adding the three-dimensional pose as the feature parameter to the original point cloud trajectory to obtain the intermediate point cloud trajectory, the method further includes: Determine the distance between two adjacent trajectory points of the intermediate point cloud trajectory; If the distance is greater than the length of the acquisition device, then the intermediate point cloud trajectory is interpolated so that the distance between two adjacent trajectory points is less than the length of the acquisition device.

4. The method according to any one of claims 1-3, characterized in that, The step of filtering out afterimages from the intermediate point cloud map based on the 3D bounding box and the local volume density value to obtain the target point cloud map includes: All point cloud points in the intermediate point cloud map corresponding to the 3D frame are filtered out; and based on the local volume density value, afterimage points are determined from the point cloud points within a preset range from the outer surface of the 3D frame, and the afterimage points are filtered out from the intermediate point cloud map to obtain the target point cloud map.

5. The method as described in claim 4, characterized in that, The step of determining the afterimage points from the point cloud points within a preset range from the outer surface of the stereo frame based on the local volume density value includes: Points in the intermediate point cloud map that meet the following conditions are identified as the afterimage points: in, The local volume density of the point cloud points. The volume density threshold, The distance between the center of the circumscribed sphere corresponding to the 3D frame and the point cloud points is given. The preset radius threshold, Let be the radius of the circumscribed sphere corresponding to the solid frame.

6. The method according to any one of claims 1-3, characterized in that, The step of performing afterimage filtering processing on the intermediate point cloud map according to the intermediate point cloud trajectory and preset afterimage filtering rules to obtain the target raster map includes: Based on the three-dimensional pose of each trajectory point in the intermediate point cloud trajectory, determine the stereo frame corresponding to the acquisition device; Based on the stereo frame, the intermediate point cloud map is subjected to afterimage filtering to obtain the target point cloud map; Based on the target point cloud map, a target raster map is generated.

7. The method as described in claim 6, characterized in that, The step of generating a target raster map based on the target point cloud map includes: Extract the target point cloud within a preset height range from the target point cloud map; Based on the target point cloud, a raster map of the area corresponding to the original point cloud map is generated.

8. The method as described in claim 7, characterized in that, The step of generating a target raster map based on the target point cloud map includes: Based on the target point cloud map and the intermediate point cloud trajectory, the interception distance range of the point cloud map is determined, and the distance range includes the highest value and the lowest value; Based on the stated interception distance range, the target point cloud is intercepted from the target point cloud map; Based on the target point cloud, a raster map of the area corresponding to the original point cloud map is generated.

9. The method as described in claim 7 or 8, characterized in that, The step of generating a raster map of the region corresponding to the original point cloud map based on the target point cloud includes: The target point cloud is projected onto a two-dimensional plane parallel to the ground of the region to obtain a projection plane point cloud; According to the preset resolution, the projection plane point cloud is divided into multiple square grids; Based on whether point cloud points exist in each grid, determine whether to mark the grid as an obstacle; An image closing operation is performed on each of the marked grid cells to generate the grid map.

10. The method as described in claim 1, characterized in that, After obtaining the original point cloud map and point cloud trajectory, the method further includes: Principal component analysis is performed on the original point cloud map to obtain the pose matrix of the original point cloud map, which is used to represent the pose of the original point cloud map relative to the ground. Based on the pose matrix, determine whether the original point cloud map is tilted; If the original point cloud map is tilted, the tilt of the original point cloud map and the original point cloud trajectory is corrected based on the pose matrix.

11. A mobile robot, characterized in that, The mobile robot generates a target point cloud map and / or a target grid map by means of the method described in any one of claims 1-10.

12. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-10.

13. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-10.

Images