Method, device and equipment for eliminating dynamic point cloud in noise map
Through the depth conversion and background difference method combined with adaptive nearest neighbor search, the problem of dynamic point cloud removal in laser SLAM is solved, the noise map is optimized, and the accuracy and robustness of positioning navigation are improved.
Patent Information
- Application Number
- CN202410998736.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-24
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-07-24
AI Technical Summary
When the existing laser SLAM method builds point cloud maps in static or low dynamic scenarios, it cannot effectively remove dynamic point clouds, resulting in dynamic objects remaining in the noise map, affecting the positioning navigation accuracy.
By obtaining the point cloud data frame and noise map at the current moment, using preset depth conversion rules and background difference principles to determine dynamic point clouds, combined with adaptive nearest neighbor search rules, delete dynamic point clouds from the noise map, and obtain an optimized point cloud map .
有效剔除噪声地图中的动态物体,提高点云地图的精度和鲁棒性,为后续定位导航任务提供了更准确的支持。
Smart Images

Figure CN118982480A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of robot platform mapping, positioning and navigation technology, and in particular to a method, device, electronic device, non-transitory computer-readable storage medium and computer program product for removing dynamic point clouds in noise maps. Background Art
[0002] With the development of autonomous driving technology and robotics technology, the simultaneous localization and mapping (SLAM) technology has received more and more attention. Among them, SLAM can be divided into two categories: visual SLAM and laser SLAM according to the difference in the type of sensors used. The visual SLAM has high requirements for lighting and texture and is not robust enough in complex environments; the laser SLAM has higher robustness and accuracy.
[0003] In the prior art, SLAM mapping algorithms based on three-dimensional lidar (for example, Lego-LOAM, LIO-SAM, and FAST-LIO2, etc.) are usually used to construct high-precision point cloud maps in static scenes.
[0004] However, the laser SLAM methods in the prior art are all based on the assumption of static scenes or low-dynamic scenes, while there are a large number of dynamic objects in actual mapping scenes, such as pedestrians and vehicles; therefore, the laser SLAM methods in the prior art cannot effectively remove dynamic point clouds (i.e., point clouds about dynamic objects) in the constructed point cloud maps. Summary of the invention
[0005] The present invention provides a method, an apparatus, an electronic device, a non-transitory computer-readable storage medium and a computer program product for removing dynamic point clouds in a noise map, so as to solve the defects in the prior art.
[0006] The present invention provides a method for removing dynamic point clouds in a noise map, comprising: Get the point cloud data frame, noise map and pose information at the current moment; The point cloud data frame is composed of point cloud data in the global coordinate system at the current moment, which is recorded as the current frame; the noise map is a three-dimensional point cloud map containing dynamic point clouds, which is formed by splicing point cloud data frames before the current moment in the noise map coordinate system; Based on the current frame, the noise map and the pose information, a first depth map and a second depth map are determined using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; and the first depth map and the second depth map have the same size and the same number of pixels; Based on the first depth map and the second depth map, determining at least one first dynamic point cloud in the current frame by using a background difference principle; Based on the at least one first dynamic point cloud, determining a second set of dynamic point clouds in the noise map using an adaptive nearest neighbor search rule; The second dynamic point cloud set is deleted from the noise map to obtain an optimized point cloud map.
[0007] The present invention also provides a device for removing dynamic point clouds in a noise map, comprising: The data acquisition module is configured to: acquire a point cloud data frame, a noise map, and position information at the current moment; wherein the point cloud data frame is composed of point cloud data in the global coordinate system at the current moment, recorded as the current frame; the noise map is a three-dimensional point cloud map containing a dynamic point cloud, which is formed by splicing the point cloud data frames before the current moment in the noise map coordinate system; A depth map conversion module is configured to: determine a first depth map and a second depth map based on the current frame, the noise map and the pose information using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; and the first depth map and the second depth map have the same size and the same number of pixels; The difference and dynamic calculation module is configured to: determine at least one first dynamic point cloud in the current frame based on the first depth map and the second depth map by using a background difference principle; An adaptive search module is configured to: determine a second set of dynamic point clouds in the noise map using an adaptive nearest neighbor search rule based on the at least one first dynamic point cloud; The noise removal module is configured to: delete the second dynamic point cloud set from the noise map to obtain an optimized point cloud map.
[0008] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for removing dynamic point clouds from a noise map as described in any one of the above is implemented.
[0009] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for removing a dynamic point cloud from a noise map as described in any one of the above is implemented.
[0010] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the computer program implements any of the above methods for removing dynamic point clouds from a noise map.
[0011] As described above, the method for eliminating dynamic point clouds in noise maps of the present invention performs ground segmentation and height segmentation of point clouds on the current frame and the noise map containing the dynamic point clouds; converts the depth information of the point clouds into visually usable image information, uses the background difference method in visual theory to perform depth image comparison on the current frame and the noise map, screens out the initial dynamic point cloud and calculates the dynamic score; performs an adaptive nearest neighbor search on the noise map according to the dynamic score to eliminate the dynamic point clouds therein, thereby ultimately effectively eliminating the dynamic objects in the noise map and obtaining an optimized point cloud map; and further provides effective support for subsequent downstream positioning and navigation tasks. BRIEF DESCRIPTION OF THE DRAWINGS
[0012] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0013] Figure 1 It is one of the flowcharts of the method for removing dynamic point clouds in a noise map provided by an embodiment of the present invention; Figure 2 is a schematic diagram of a device for performing laser SLAM provided by an embodiment of the present invention; Figure 3 This is a second flow chart of a method for removing dynamic point clouds from a noise map provided by an embodiment of the present invention; Figure 4 This is a third flow chart of a method for removing dynamic point clouds from a noise map provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of conversion between a rectangular coordinate system and a spherical coordinate system provided by an embodiment of the present invention; Figure 6a is the first depth map provided by an embodiment of the present invention; Figure 6b is the second depth map provided by an embodiment of the present invention; Figure 7a is a schematic diagram of a point cloud of the current frame provided by an embodiment of the present invention; Figure 7b is a schematic diagram of the ground point cloud provided by an embodiment of the present invention; Figure 8a is a schematic diagram of the noise map provided by an embodiment of the present invention; Figure 8b is a schematic diagram of the height point cloud provided by an embodiment of the present invention; Fig. 9This is a fourth flow chart of a method for removing dynamic point clouds from a noise map provided by an embodiment of the present invention; Fig.10 is a schematic diagram of the background difference principle provided by an embodiment of the present invention; Fig.11 is a schematic diagram of the structure of a device for removing dynamic point clouds in a noise map provided by an embodiment of the present invention; Fig.12 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION
[0014] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0015] In order to solve the above problems of the prior art, the inventors of the present invention have proposed a solution for removing dynamic point clouds in a noise map as described in the following embodiment after research.
[0016] Example: The following describes the implementation details of the solution for removing dynamic point clouds in noise maps according to the present invention in conjunction with the accompanying drawings.
[0017] Figure 1 1 is a flow chart of a method for removing dynamic point clouds from a noise map provided by an exemplary embodiment of the present invention. This embodiment can be applied to electronic devices (for example, a processor or a main control computer mounted on a mobile robot platform), such as Figure 1 As shown, the method for removing dynamic point clouds in a noise map includes the following steps: S110, obtaining the point cloud data frame, noise map and posture information at the current moment; Among them, the point cloud data frame is composed of point cloud data in the global coordinate system at the current moment, recorded as the current frame; the noise map is a three-dimensional point cloud map containing a dynamic point cloud, which is formed by splicing the point cloud data frames before the current moment in the noise map coordinate system.
[0018] S120: Based on the current frame, the noise map, and the pose information, a first depth map and a second depth map are determined using a preset depth conversion rule.
[0019] The first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; the first depth map and the second depth map have the same size and the same number of pixels.
[0020] S130. Based on the first depth map and the second depth map, determine at least one first dynamic point cloud in the current frame using a background difference principle.
[0021] S140. Based on the at least one first dynamic point cloud, determine a second dynamic point cloud set in the noise map using an adaptive nearest neighbor search rule.
[0022] S150: Delete the second dynamic point cloud set from the noise map to obtain an optimized point cloud map.
[0023] The posture information is the posture data or parameters generated by the device or apparatus performing laser SLAM at the current moment.
[0024] The present invention does not limit the acquisition method. For example, an electronic device as an execution subject (for example, a processor or a main control computer carried by a mobile robot platform) can directly perform device communication of laser SLAM to obtain the point cloud data frame, noise map and posture information at the current moment.
[0025] As an alternative example, refer to Figure 2 The device or apparatus for performing laser SLAM may be a mobile robot platform. The mobile robot platform may include a robot platform chassis, a main control machine, a laser radar sensor, and an IMU sensor. The robot platform chassis is loaded with the main control machine, the laser radar sensor, and the IMU sensor; the main control machine is connected to the robot platform chassis, the laser radar sensor, and the IMU sensor in sequence through wired means.
[0026] It should be noted that the present invention does not limit the models of the main control machine, laser radar sensor, and IMU sensor, which can be determined according to actual needs. For example, the main control machine can be Intel NUC8i7 HVK; the laser radar sensor can be Velodyne-16-line laser radar; and the IMU sensor can be MTi-300IMU.
[0027] The specific implementation of the above steps will be described below and will not be repeated here.
[0028] As described above, the method for eliminating dynamic point clouds in noise maps of the present invention performs ground segmentation and height segmentation of point clouds on the current frame and the noise map containing the dynamic point clouds; converts the depth information of the point clouds into visually usable image information, uses the background difference method in visual theory to perform depth image comparison on the current frame and the noise map, screens out the initial dynamic point cloud and calculates the dynamic score; performs an adaptive nearest neighbor search on the noise map according to the dynamic score to eliminate the dynamic point clouds therein, thereby ultimately effectively eliminating the dynamic objects in the noise map and obtaining an optimized point cloud map; and further provides effective support for subsequent downstream positioning and navigation tasks.
[0029] Based on the above embodiment, as an optional implementation mode, refer to Figure 3 and Figure 4 Step S120 "determining a first depth map and a second depth map based on the current frame, the noise map and the pose information using a preset depth conversion rule" includes the following implementation methods: S1210. Based on the posture information, convert the current frame to a noise map coordinate system to obtain a converted current frame.
[0030] As an optional example, the following calculation formula may be used to determine the conversion current frame;
[0031] in, Indicates the current frame of the conversion; represents the posture information; represents the current frame; represents the rotation matrix; Represents the translation vector.
[0032] S1220. Based on a random sampling consistency algorithm, the converted current frame is processed to obtain a ground point cloud.
[0033] Reference Figure 3 , through step S1220, the ground segmentation is performed on the current frame scan in the global coordinate system to obtain the ground point cloud.
[0034] As an optional example, the plane model fitting method in the Random Sample Consensus (RANSAC) algorithm can be used to extract the ground point cloud more accurately. RANSAC is an iterative method for fitting models and identifying outliers in data. It selects a subset from the point cloud data set of the current frame by random sampling to fit a model, and finds the inliers that conform to the model and the outliers that do not conform to the model in the entire data set based on the model. By iterating this process, a model that fits the data set is finally obtained.
[0035] Specifically, three points can determine a plane, and the plane formula is as follows:
[0036] The distance from each point (x, y, z) to the plane in the point cloud data set of the current frame can be obtained ;
[0037] Among them, (A, B, C) represents the normal vector of the plane, and D represents a constant term.
[0038] like ( represents the distance threshold), the point is considered to be on the plane. The threshold is set according to the standard deviation of the distance from all points to the plane at each iteration.
[0039] like , the point is determined to be a non-ground point. Specifically,
[0040]
[0041] in, Indicates the total number of point clouds in the point cloud dataset of the current frame; Represents the average distance from all point clouds to the plane in the point cloud dataset of the current frame; Represents the distance from the i-th point cloud to the plane.
[0042] Iterations The value of will affect the speed and effect of ground fitting, which is a more important indicator. The appropriate value can be selected through the following formula :
[0043] in, The number of points chosen for each iteration; is the ratio of abnormal points in the point cloud; is the probability of selecting at least one normal point.
[0044] The above method can be used to extract ground point cloud more accurately. Optionally, the point cloud of the current frame is as follows: Figure 7a As shown, the ground point cloud is Figure 7b shown.
[0045] S1230 : Process the noise map based on a straight-through filtering algorithm to obtain a height point cloud.
[0046] Reference Figure 3, perform height segmentation on the noisy map containing the dynamic target to obtain the height point cloud.
[0047] Specifically, the basic formula of the direct-pass filter can be expressed as:
[0048] in, A point representing the filtered point cloud data of the noise map; Represents the i-th point in the point cloud data of the noise map; Indicates the axis of filtering (here is the z-axis, i.e., height); Indicates the lower limit of the filtering range; Indicates the upper limit of the filtering range.
[0049] The filter axis and height range are set, and a filtering operation is performed on the point cloud within a specific range of the z-axis, so as to screen out the height point cloud.
[0050] Wherein, optionally, the noise map is as follows Figure 8a As shown, the height point cloud is Figure 8b shown.
[0051] S1240: Process the ground point cloud using the preset depth conversion rule to obtain the first depth map. S1250: Process the height point cloud using the preset depth conversion rule to obtain the second depth map.
[0052] Reference Figure 3 , converting the three-dimensional point cloud into a depth map, that is, converting the ground point cloud into a first depth map and the height point cloud into a second depth map respectively.
[0053] Specifically, the coordinates of each point cloud in the ground point cloud and the height point cloud can be calculated using the following formula: Convert from rectangular coordinate system to spherical coordinate system; get ground point cloud in spherical coordinate system , and the height point cloud in spherical coordinates .
[0054]
[0055] Among them, refer to Figure 5 , is the straight-line distance from the point cloud to the laser radar, is the elevation angle, is the azimuth.
[0056] Based on the ground point cloud in the spherical coordinate system , and the height point cloud in spherical coordinates , the first depth map (such as Figure 6a ) and the second depth map (as shown Figure 6b as shown).
[0057]
[0058] in, Represents the pixel point in the first depth map The set of distances contained in ; Represents the pixel point in the second depth map The set of distances contained in ; represents the pixel at the coordinate (i, j) in the first depth map; Represents the pixel at coordinate (i, j) in the second depth map.
[0059] Based on the above embodiment, as an optional implementation mode, refer to Fig. 9 Step S130 "determining at least one first dynamic point cloud in the current frame based on the first depth map and the second depth map using the background difference principle" may include the following implementations: S1310: Calculate the pixel difference between each group of pixel pairs with matching positions by using the first depth map and the second depth map.
[0060] As an alternative example, refer to Fig.10 Based on the principle of background difference method (i.e., comparing the pixel difference between the current frame image and the background frame image pixel by pixel to obtain the target), the first depth map and the second depth image are compared pixel by pixel at each moment and matrix subtraction is performed to obtain the pixel difference between the first depth map and the second depth map. , the calculation formula is as follows:
[0061] in, represents the pixel at the coordinate (i, j) in the first depth map; Represents the pixel at coordinate (i, j) in the second depth map.
[0062] S1320. Determine a corresponding preliminary dynamic point cloud in the current frame based on the pixel difference of each group of pixel pairs.
[0063] Optionally, refer to Figure 3 , determine whether the background difference value is greater than a threshold. Specifically, if the pixel difference is greater than a preset pixel difference threshold, it is considered that the point cloud at the corresponding pixel position in the current frame belongs to a dynamic point cloud; thus, it can be determined as the preliminary dynamic point cloud.
[0064] S1330. Calculate a dynamic score corresponding to each preliminary selected dynamic point cloud based on the pixel difference corresponding to each preliminary selected dynamic point cloud.
[0065] As an alternative example, refer to Figure 3 , perform dynamic score calculation. Specifically, for each dynamic point cloud detected in the background difference method, the mean square error (MSE) is used to calculate the dynamic score of its pixel difference in the depth image:
[0066] in, Represents the first depth map and the second depth map The depth value difference of pixels; Indicates the total number of pixels; represents the dynamic score. The larger the value, the greater the difference between the first depth map and the second depth map, and the higher the corresponding dynamic point cloud score. The obtained dynamic score is assigned to each dynamic point cloud to indicate its dynamic degree.
[0067] S1340: Select at least one preliminary dynamic point cloud whose dynamic score is greater than a preset score threshold as the at least one first dynamic point cloud.
[0068] Based on the above embodiment, as an optional implementation, step S140 "determining a second dynamic point cloud set in the noise map by using an adaptive nearest neighbor search rule based on the at least one first dynamic point cloud" may include the following implementation: Reference Figure 3 , using adaptive nearest neighbor search to determine the dynamic target in the noise map.
[0069] Specifically, 1) based on the adaptive nearest neighbor search rule and the dynamic score corresponding to the at least one first dynamic point cloud, a search root node, a leaf node, and nodes between the root node and the leaf node that match each of the first dynamic point clouds are determined in the noise map.
[0070] Here, determining the search root node, leaf node, and nodes between the root node and the leaf node that match each of the first dynamic point clouds can be understood as determining the search range and depth based on the dynamic score.
[0071] 2) Based on the search root node, leaf node and each node between the root node and the leaf node that matches each of the first dynamic point clouds, the Euclidean distance between each node and each of the first dynamic point clouds (recorded as target points) is calculated in sequence starting from the leaf node in the direction of backtracking from the leaf node to the root node.
[0072] As an optional example, the following calculation formula may be used to sequentially calculate the Euclidean distance between each node and each of the first dynamic point clouds;
[0073] in, Indicates the X-axis coordinate of the current node (referred to as the current point) participating in the calculation among the nodes in the noise map coordinate system; represents the X-axis coordinate of the first dynamic point cloud in the noise map coordinate system; Indicates the Y-axis coordinate of the current node participating in the calculation in the noise map coordinate system; represents the Y-axis coordinate of the first dynamic point cloud in the noise map coordinate system; Indicates the Z-axis coordinate of the current node participating in the calculation in the noise map coordinate system; Represents the Z-axis coordinate of the first dynamic point cloud in the noise map coordinate system.
[0074] 3) Using the Euclidean distance between each node and each of the first dynamic point clouds, determine the nearest neighbor point matching each of the first dynamic point clouds as the second dynamic point cloud.
[0075] 4) Putting the second dynamic point cloud matching each of the first dynamic point clouds into an ordered set to obtain the second dynamic point cloud set.
[0076] It should be noted that during the backtracking process, if the distance between the target point and the current point is less than the current closest distance, the nearest neighbor point and the nearest distance are updated; the next time the judgment is made, the updated nearest neighbor point and the nearest distance are used for judgment, and this iteration is repeated until the search ends.
[0077] As described above, the method for eliminating dynamic point clouds in noise maps of the present invention performs ground segmentation and height segmentation of point clouds on the current frame and the noise map containing the dynamic point clouds; converts the depth information of the point clouds into visually usable image information, uses the background difference method in visual theory to perform depth image comparison on the current frame and the noise map, screens out the initial dynamic point cloud and calculates the dynamic score; performs an adaptive nearest neighbor search on the noise map according to the dynamic score to eliminate the dynamic point clouds therein, thereby ultimately effectively eliminating the dynamic objects in the noise map and obtaining an optimized point cloud map; and further provides effective support for subsequent downstream positioning and navigation tasks.
[0078] The following is a description of the device for removing dynamic point clouds in a noise map provided by the present invention. The device for removing dynamic point clouds in a noise map described below and the method for removing dynamic point clouds in a noise map described above can be referenced to each other.
[0079] Fig.11 FIG. 1 is a schematic diagram of a structure of an apparatus for removing dynamic point clouds from a noise map provided by an exemplary embodiment of the present invention. Fig.11 As shown, the device for removing dynamic point clouds in a noise map includes: The data acquisition module 101 is configured to: acquire the point cloud data frame, noise map and posture information at the current moment; wherein the point cloud data frame is composed of point cloud data in the global coordinate system at the current moment, recorded as the current frame; the noise map is a three-dimensional point cloud map containing dynamic point clouds, which is formed by splicing the point cloud data frames before the current moment in the noise map coordinate system; the depth map conversion module 102 is configured to: determine the first depth map and the second depth map based on the current frame, noise map and posture information using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the current frame. Corresponding to the noise map; the first depth map and the second depth map have the same size and the same number of pixels; the difference and dynamic calculation module 103 is configured to: based on the first depth map and the second depth map, use the background difference principle to determine at least one first dynamic point cloud in the current frame; the adaptive search module 104 is configured to: based on the at least one first dynamic point cloud, use the adaptive nearest neighbor search rule to determine the second dynamic point cloud set in the noise map; the noise deletion module 105 is configured to: delete the second dynamic point cloud set from the noise map to obtain an optimized point cloud map.
[0080] Fig.12 An example of a physical structure diagram of an electronic device is shown in FIG. Fig.12As shown, the electronic device may include: a processor (processor) 1210, a communication interface (Communications Interface) 1220, a memory (memory) 1230 and a communication bus 1240, wherein the processor 1210, the communication interface 1220, and the memory 1230 communicate with each other through the communication bus 1240. The processor 1210 can call the logic instructions in the memory 1230 to execute a method for removing dynamic point clouds from a noise map, the method comprising: obtaining a point cloud data frame, a noise map and pose information at a current moment; wherein the point cloud data frame is composed of point cloud data in a global coordinate system at a current moment, recorded as a current frame; the noise map is a three-dimensional point cloud map containing dynamic point clouds, which is formed by splicing point cloud data frames before the current moment in a noise map coordinate system; based on the current frame, the noise map and the pose information, a first depth map and a second depth map are determined using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; the first depth map and the second depth map have the same size and the same number of pixels; based on the first depth map and the second depth map, at least one first dynamic point cloud in the current frame is determined using a background difference principle; based on the at least one first dynamic point cloud, a second dynamic point cloud set is determined in the noise map using an adaptive nearest neighbor search rule; the second dynamic point cloud set is deleted from the noise map to obtain an optimized point cloud map.
[0081] In addition, the logic instructions in the above-mentioned memory 1230 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on such an understanding, the technical solution of the present invention can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.
[0082] On the other hand, the present invention also provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the method provided by the above methods for removing dynamic point clouds from noise maps, the method including: obtaining a point cloud data frame, a noise map and posture information at the current moment; wherein the point cloud data frame is composed of point cloud data in a global coordinate system at the current moment, recorded as a current frame; the noise map is a three-dimensional point cloud map containing dynamic point clouds, which is formed by splicing point cloud data frames before the current moment in the noise map coordinate system; based on The current frame, noise map and posture information are used to determine a first depth map and a second depth map using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; the first depth map and the second depth map have the same size and the same number of pixels; based on the first depth map and the second depth map, at least one first dynamic point cloud in the current frame is determined using a background difference principle; based on the at least one first dynamic point cloud, a second dynamic point cloud set is determined in the noise map using an adaptive nearest neighbor search rule; the second dynamic point cloud set is deleted from the noise map to obtain an optimized point cloud map.
[0083] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the method for removing dynamic point clouds from noise maps provided by the above methods, the method comprising: obtaining a point cloud data frame, a noise map and posture information at a current moment; wherein the point cloud data frame is composed of point cloud data in a global coordinate system at a current moment, recorded as a current frame; the noise map is a three-dimensional point cloud map containing dynamic point clouds, which is formed by splicing point cloud data frames before the current moment in a noise map coordinate system; based on the current frame, the noise map and posture information information, using a preset depth conversion rule to determine a first depth map and a second depth map; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; the first depth map and the second depth map have the same size and the same number of pixels; based on the first depth map and the second depth map, using the background difference principle, determining at least one first dynamic point cloud in the current frame; based on the at least one first dynamic point cloud, using an adaptive nearest neighbor search rule, determining a second dynamic point cloud set in the noise map; deleting the second dynamic point cloud set from the noise map to obtain an optimized point cloud map.
[0084] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0085] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0086] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for removing dynamic point clouds from a noisy map, characterized in that: include: Get the point cloud data frame, noise map and pose information at the current moment; The point cloud data frame is composed of point cloud data in the global coordinate system at the current moment, which is recorded as the current frame; the noise map is a three-dimensional point cloud map containing dynamic point clouds, which is formed by splicing point cloud data frames before the current moment in the noise map coordinate system; Based on the current frame, the noise map and the pose information, a first depth map and a second depth map are determined using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; and the first depth map and the second depth map have the same size and the same number of pixels; Based on the first depth map and the second depth map, determining at least one first dynamic point cloud in the current frame by using a background difference principle; Based on the at least one first dynamic point cloud, determining a second set of dynamic point clouds in the noise map using an adaptive nearest neighbor search rule; The second dynamic point cloud set is deleted from the noise map to obtain an optimized point cloud map.
2. The method according to claim 1, characterized in that Based on the current frame, the noise map, and the pose information, a first depth map and a second depth map are determined using a preset depth conversion rule, including: Based on the posture information, the current frame is converted into a noise map coordinate system to obtain a converted current frame; Based on a random sampling consensus algorithm, the converted current frame is processed to obtain a ground point cloud; Based on a straight-through filtering algorithm, the noise map is processed to obtain a height point cloud; Processing the ground point cloud using the preset depth conversion rule to obtain the first depth map; The height point cloud is processed using the preset depth conversion rule to obtain the second depth map.
3. The method according to claim 1, characterized in that Based on the first depth map and the second depth map, at least one first dynamic point cloud in the current frame is determined by using a background difference principle, including: Calculate the pixel difference between each group of pixel pairs with matching positions using the first depth map and the second depth map; Based on the pixel difference of each group of pixel pairs, determining a corresponding preliminary selected dynamic point cloud in the current frame; Calculate a dynamic score corresponding to each preliminary selected dynamic point cloud based on the pixel difference corresponding to each preliminary selected dynamic point cloud; At least one preliminarily selected dynamic point cloud having a dynamic score greater than a preset score threshold is selected as the at least one first dynamic point cloud.
4. The method according to claim 1, characterized in that Based on the at least one first dynamic point cloud, determining a second dynamic point cloud set in the noise map using an adaptive nearest neighbor search rule comprises: Determine, in the noise map, a search root node, a leaf node, and nodes between the root node and the leaf node that match each of the first dynamic point clouds based on the adaptive nearest neighbor search rule and the dynamic score corresponding to the at least one first dynamic point cloud; Based on the search root node, leaf node and nodes between the root node and the leaf node that match each of the first dynamic point clouds, in the direction of backtracking from the leaf node to the root node, starting from the leaf node, sequentially calculate the Euclidean distance between each node and each of the first dynamic point clouds; Using the Euclidean distance between each node and each of the first dynamic point clouds, determine the nearest neighbor point matching each of the first dynamic point clouds as the second dynamic point cloud; The second dynamic point clouds matching each of the first dynamic point clouds are put into an ordered set to obtain a second dynamic point cloud set.
5. A device for removing dynamic point clouds from a noise map, characterized in that: include: The data acquisition module is configured to: acquire a point cloud data frame, a noise map, and position information at the current moment; wherein the point cloud data frame is composed of point cloud data in the global coordinate system at the current moment, recorded as the current frame; the noise map is a three-dimensional point cloud map containing a dynamic point cloud, which is formed by splicing the point cloud data frames before the current moment in the noise map coordinate system; A depth map conversion module is configured to: determine a first depth map and a second depth map based on the current frame, the noise map and the pose information using a preset depth conversion rule; wherein the first depth map corresponds to the current frame, and the second depth map corresponds to the noise map; and the first depth map and the second depth map have the same size and the same number of pixels; The difference and dynamic calculation module is configured to: determine at least one first dynamic point cloud in the current frame based on the first depth map and the second depth map by using a background difference principle; An adaptive search module is configured to: determine a second set of dynamic point clouds in the noise map using an adaptive nearest neighbor search rule based on the at least one first dynamic point cloud; The noise removal module is configured to: delete the second dynamic point cloud set from the noise map to obtain an optimized point cloud map.
6. An electronic 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 program, the method for removing dynamic point clouds in a noise map as described in any one of claims 1 to 4 is implemented.
7. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for removing dynamic point clouds from a noise map as claimed in any one of claims 1 to 4 is implemented.
8. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for removing dynamic point clouds from a noise map as claimed in any one of claims 1 to 4 is implemented.
Citation Information
Patent Citations
Semantic point cloud map construction method and system and intelligent robot
CN115063550A
Dynamic point removing method for autonomous mobile platform and system and equipment thereof
CN115272730A
Dynamic obstacle removal method suitable for low-wire-harness 3D laser radar
CN116879870A
Semantic dynamic modeling method based on point cloud data processing
CN117115263A
Three-dimensional point cloud map dynamic object filtering method based on Livox laser radar
CN117409166A
