Real-time point cloud smoothing method and device, terminal equipment and storage medium
By performing real-time preprocessing and cache queue updates on radar point cloud data, combined with centroid and normal vector calculations, real-time smoothing of radar point cloud data is achieved, solving the problem of the inability to update point cloud maps in real time in existing technologies, and improving the smoothing effect and efficiency.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SHENZHEN XGRIDS-INNOVATION CO LTD
- Filing Date
- 2026-05-21
- Publication Date
- 2026-06-26
AI Technical Summary
Existing technologies cannot achieve real-time smoothing of radar point cloud data, especially they cannot support continuous updates of the point cloud map as sensor input flows.
By acquiring and preprocessing real-time input radar point cloud data, updating the point cloud map using a cache queue, and calculating the centroid and normal vector of the head data in the cache queue, real-time smoothing of the point cloud data is achieved.
It enables real-time processing of point cloud data, solves the technical problem of not being able to smooth in real time, improves smoothing accuracy and robustness, and reduces latency.
Smart Images
Figure CN122289062A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of laser point cloud processing, and in particular to a real-time point cloud smoothing method, apparatus, terminal equipment, and storage medium. Background Technology
[0002] Point cloud smoothing is a crucial preprocessing step in systems such as 3D reconstruction, SLAM, and autonomous driving perception. Its goal is to suppress measurement noise, motion distortion, and multipath reflection artifacts introduced by LiDAR or depth camera acquisition while preserving geometric features (such as edges and corners). Currently, the mainstream method is still represented by offline Moving Least Squares (MLS). However, this method cannot be incrementally updated; it can only construct the entire point cloud at once and cannot support continuous updates of the point cloud map as sensor input flows. Summary of the Invention
[0003] In view of this, embodiments of this application provide a real-time point cloud smoothing method, which can effectively solve the problem of not being able to smooth radar point cloud data in real time.
[0004] In a first aspect, embodiments of this application provide a real-time point cloud smoothing method, including: Acquire real-time input radar point cloud data and preprocess the radar point cloud data to obtain standard point cloud data; The standard point cloud data is stored from the tail of the queue into a cache queue, and the standard point cloud data is input into the current point cloud map to update the point cloud map. Obtain the header standard point cloud data in the cache queue, process each point in the header standard point cloud data in turn, and obtain the centroid and normal vector of each point in the header standard point cloud data. Based on the normal vector and centroid of each point in the head standard point cloud data, a smoothing operation is performed on each point in the head standard point cloud data to obtain multiple smooth points. The multiple smooth points are used to form smooth point cloud data. Return to the step of acquiring real-time input radar point cloud data, save all the smoothed point cloud data and output the smoothed point cloud map.
[0005] In some embodiments, after inputting the standard point cloud data into the current point cloud map to update the point cloud map, the method further includes: In the point cloud map, points whose locations are outside a preset range are deleted; Delete points in the point cloud map whose update time exceeds the preset time.
[0006] In some embodiments, the radar point cloud data includes radar pose, displacement, and point cloud coordinates; The preprocessing of the radar point cloud data to obtain standard point cloud data includes: The radar pose and displacement corresponding to the radar point cloud data are obtained, and the coordinates of each point in the radar point cloud data are transformed to obtain radar point cloud data in the world coordinate system. The radar point cloud data in the world coordinate system is used as standard point cloud data.
[0007] In some embodiments, the step of sequentially processing each point in the head standard point cloud data to obtain the centroid and normal vector of each point in the head standard point cloud data includes: Each point in the head standard point cloud data is obtained sequentially; In the point cloud map, a plane search with a preset radius is performed on the current point in the point cloud to obtain all neighboring points of the current point; Based on all the neighboring points and the current point, the average constructed centroid of the current point is calculated; The covariance matrix of the current point and all its neighboring points is constructed and eigenvalues are decomposed to obtain multiple eigenvalues. The eigenvector corresponding to the smallest eigenvalue is used as the normal vector of the current point.
[0008] In some embodiments, after performing a planar search with a preset radius on the current point in the point cloud to obtain all neighboring points of the current point, the method further includes: Calculate the number of all neighboring points of the current point. If the total number of neighboring points is less than a preset value, discard the current point.
[0009] In some embodiments, after using the eigenvector corresponding to the minimum eigenvalue as the normal vector of the current point, the method further includes: Obtain the largest eigenvalue among the plurality of eigenvalues; Calculate the flatness based on the maximum and minimum eigenvalues; If the flatness is greater than the preset flatness, then the preset radius is modified, and the step of obtaining all neighboring points of the current point is re-executed until the flatness is less than the preset flatness, or the number of iterations is greater than the preset number of repetitions, then the iteration stops.
[0010] In some embodiments, the smoothing operation is performed on each point in the head standard point cloud data based on the normal vector and centroid of each point to obtain a plurality of smoothed points, which are used to form smoothed point cloud data, including: Each point in the head standard point cloud data is obtained sequentially; Determine the current centroid and current normal vector corresponding to the current point; Determine the current projection plane based on the current centroid and the current normal vector; Project the current point onto the current projection plane to obtain the corresponding smooth point; Repeat the steps of determining the current centroid and current normal vector corresponding to the current point until the smoothed points corresponding to all points in the head standard point cloud data are obtained. Secondly, this application also provides a real-time point cloud smoothing device, comprising: The point cloud acquisition module is used to acquire real-time input radar point cloud data and preprocess the radar point cloud data to obtain standard point cloud data. The caching module is used to store the standard point cloud data from the tail of the queue into the cache queue, and input the standard point cloud data into the current point cloud map to update the point cloud map; The data processing module is used to acquire the header standard point cloud data in the cache queue, process each point in the header standard point cloud data in sequence, and obtain the centroid and normal vector of each point in the header standard point cloud data. The smoothing module is used to perform a smoothing operation on each point in the head standard point cloud data based on the normal vector and centroid of each point in the head standard point cloud data to obtain multiple smoothed points, which are used to form smoothed point cloud data. The loop output module is used to return to the step of acquiring the radar point cloud data in real time, save all the smoothed point cloud data and output the smoothed point cloud map.
[0011] Thirdly, this application also provides a terminal device, the terminal device including a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the real-time point cloud smoothing method.
[0012] Fourthly, this application also provides a readable storage medium storing a computer program that, when executed on a processor, implements the real-time point cloud smoothing method described above.
[0013] The embodiments of this application have the following beneficial effects: This application achieves real-time processing of the head standard point cloud data in the cache queue by updating the point cloud map in real time and caching standard point cloud data. It processes each point in the head standard point cloud data sequentially to obtain the centroid and normal vector of each point. Then, based on the normal vector and centroid of each point in the head standard point cloud data, a smoothing operation is performed on each point to obtain multiple smoothed points. These multiple smoothed points are used to form smoothed point cloud data. This achieves the technical effect of real-time processing of point cloud data and solves the technical problem of not being able to smooth data in real time. Attached Figure Description
[0014] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0015] Figure 1 This paper illustrates a flowchart of a real-time point cloud smoothing method according to an embodiment of this application. Figure 2 This paper illustrates a schematic diagram of the process for obtaining the centroid and normal vector of a point in a point cloud according to an embodiment of this application. Figure 3 This illustration shows a schematic diagram of an original point cloud according to an embodiment of this application; Figure 4 A schematic diagram of a smoothed point cloud according to an embodiment of this application is shown; Figure 5 A schematic diagram of a real-time point cloud smoothing device according to an embodiment of this application is shown. Detailed Implementation
[0016] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.
[0017] The components of the embodiments of this application described and illustrated in the accompanying drawings can be arranged and designed in a variety of different configurations. Therefore, the following detailed description of the embodiments of this application provided in the drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.
[0018] In the following text, the terms "comprising," "having," and their cognates, which may be used in various embodiments of this application, are intended only to indicate a particular feature, number, step, operation, element, component, or combination thereof, and should not be construed as primarily excluding the presence of one or more other features, numbers, steps, operations, elements, components, or combinations thereof, or adding the possibility of one or more combinations thereof. Furthermore, the terms "first," "second," "third," etc., are used only for distinguishing descriptions and should not be construed as indicating or implying relative importance.
[0019] Unless otherwise specified, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of this application pertain. Terms (such as those defined in commonly used dictionaries) shall be interpreted as having the same meaning as in their contextual meaning in the relevant technical field and shall not be construed as having an idealized or overly formal meaning, unless clearly defined in the various embodiments of this application.
[0020] The following detailed description of some embodiments of this application is provided in conjunction with the accompanying drawings. Unless otherwise specified, the following embodiments and features can be combined with each other.
[0021] This application addresses the problems of existing technologies by proposing a real-time point cloud smoothing method. The method involves acquiring real-time input radar point cloud data, preprocessing the radar point cloud data to obtain standard point cloud data, caching the radar data using a cache queue, updating the point cloud map with newly acquired radar point cloud data, and then performing a smoothing operation on the oldest radar point cloud data in the cache queue to finally obtain a smoothed point cloud image. This achieves real-time point cloud smoothing.
[0022] The real-time point cloud smoothing method will be explained below with reference to some specific embodiments.
[0023] Figure 1 A flowchart of a real-time point cloud smoothing method according to an embodiment of this application is shown. Exemplarily, the real-time point cloud smoothing method includes the following steps: Step S100: Obtain real-time input radar point cloud data and preprocess the radar point cloud data to obtain standard point cloud data.
[0024] The technical solution of this application is applied to the real-time acquisition stage of point clouds. Therefore, it acquires the currently scanned radar point cloud data through radar. This radar point cloud data is real-time point cloud data and needs to be preprocessed. The radar can be a laser point cloud radar, etc.
[0025] The radar point cloud data includes the radar's angle R and displacement t at the time of scanning, as well as the spatial coordinates of each point in the scanned point cloud P.
[0026] The point cloud contains multiple points, so P here refers to the set of coordinates of each point in the currently scanned point cloud, and the coordinates obtained at this time are determined with the radar coordinate system as the origin.
[0027] The coordinates of each point in the point cloud P can be transformed based on the angle R and displacement t mentioned above to obtain the standard point cloud data P' in the world coordinate system.
[0028] Step S200: Store the standard point cloud data from the tail of the queue into the cache queue, and input the standard point cloud data into the current point cloud map to update the point cloud map.
[0029] To facilitate the processing of these standard point cloud data, the newly acquired standard point cloud data will be stored in a cache queue. This cache queue follows the first-in, first-out principle. Newly added data is stored in the queue from the tail, and data is retrieved from the head of the queue when data is retrieved.
[0030] It is understandable that in this queue, the first data added is at the head of the queue, and the latest data added is at the tail of the queue.
[0031] For newly added standard point cloud data, it is also necessary to update it in the current point cloud map. During the point cloud map update, in addition to adding points, points whose locations are outside a preset range will be deleted; points whose update time exceeds a preset time will also be deleted.
[0032] A point cloud map is a map that records the locations of all point clouds scanned by a point cloud radar.
[0033] The point cloud map used in this embodiment is an incrementally generated map with an incremental hash-octree data structure. Under this data structure, the point cloud map is divided into voxels with an interval of S. Each voxel maintains an octree. When a new point cloud needs to be updated to the map, the voxel to which each point in the newly added point cloud belongs is first determined, and then the octree within that voxel is incrementally updated, thereby achieving dynamic augmentation of the point cloud map.
[0034] It is understood that the aforementioned cache queue has a certain size. For example, if the queue length is 20, it can store 20 points. That is, before the queue is full, 20 point cloud data points can be continuously acquired and updated in the point cloud map. Therefore, subsequent smoothing processing can be performed after updating a maximum of 20 point cloud data points. Thus, the cache queue in this embodiment can ensure that enough point cloud data can be updated in real time in the point cloud map to ensure sufficient information for subsequent smoothing operations.
[0035] The queue length can be adjusted according to actual needs. For example, if the queue length is set to 30, it means that 30 point cloud data can be updated first, and then the obtained point cloud data can be smoothed in the point cloud map.
[0036] Step S300: Obtain the header standard point cloud data in the cache queue, and process each point in the header standard point cloud data in sequence to obtain the centroid and normal vector of each point in the header standard point cloud data.
[0037] In this embodiment, point cloud smoothing is performed by processing point cloud data in the cache queue in a certain order. Specifically, standard point cloud data that has been stored is retrieved from the head of the cache queue and smoothed. After smoothing is completed, the next standard point cloud data is smoothed.
[0038] When the standard point cloud data at the head of the cache queue is retrieved, the point cloud data is split into individual points, and then processing operations are performed on each point.
[0039] It is understandable that in the aforementioned steps, these point cloud data are updated into the point cloud map. Before smoothing, it is necessary to obtain the centroid and normal vector of each point in the point cloud data in the point cloud map.
[0040] The specific steps for calculating the centroid and normal vector are as follows: Figure 2 As shown, it includes: Step S310: Sequentially acquire each point in the head standard point cloud data.
[0041] There are multiple points in the standard point cloud data in the header. The points in the point cloud data can be obtained in the order of storage, and the obtained point is recorded as the current point.
[0042] Step S320: In the point cloud map, perform a planar search with a preset radius on the current point in the point cloud to obtain all neighboring points of the current point.
[0043] The preset radius is a fixed value. Using the current point as the center, the search returns to all neighboring points within this preset radius. Searching for neighboring points is to calculate the normal vector and centroid of the plane containing the current point. Some points may be noise points. Such points have few neighboring points. Therefore, if the number of neighboring points is less than the preset number after a search, the current point can be discarded and step S310 can be executed again.
[0044] Step S330: Calculate the average construction centroid of the current point based on all the neighboring points and the current point.
[0045] After obtaining all neighboring points, the average construction is calculated by combining all neighboring points and the current point to obtain the average construction centroid, which is the centroid of the current point.
[0046] Step S340: Construct the covariance matrix and decompose the eigenvalues of the current point and all the neighboring points to obtain multiple eigenvalues, and take the eigenvector corresponding to the smallest eigenvalue as the normal vector of the current point.
[0047] In this embodiment, a covariance matrix is constructed for the current point and all its neighboring points to perform eigenvalue decomposition, thereby obtaining multiple eigenvalues. The eigenvector corresponding to the smallest eigenvalue is the normal vector of the current point. Here, the normal vector refers to the vector perpendicular to the plane containing the current point. Therefore, it is also necessary to determine whether the calculated normal vector is a plane, i.e., whether the search result has found a plane.
[0048] Here, the minimum eigenvalue is defined as lambda_min, and the maximum eigenvalue is defined as lambda_max. Based on these minimum and maximum eigenvalues, the planarity can be calculated.
[0049] The expression for calculating this flatness is: planarity = lambda_min / lambda_max.
[0050] The smaller the flatness, the flatter the plane. A preset flatness can be set. When the flatness is greater than the preset flatness, it means that no suitable plane has been found, and it may be a non-plane falling on the structural boundary. Therefore, the search radius can be reduced, and then step S310 can be re-executed for iterative calculation, recalculating the centroid and normal vector. Iteration stops only when the calculated flatness is less than the preset flatness or the number of iterations is greater than the preset number of iterations.
[0051] The search radius can be reduced by a certain percentage each time, such as by 20%, 30%, or 50%, with the same percentage reduction each time.
[0052] Step S400: Sequentially acquire each point in the head standard point cloud data; determine the current centroid and current normal vector corresponding to the current point; determine the current projection plane based on the current centroid and the current normal vector; project the current point onto the current projection plane to obtain the corresponding smooth point; repeat the steps of determining the current centroid and current normal vector corresponding to the current point until the smooth points corresponding to all points in the head standard point cloud data are obtained.
[0053] After obtaining the normal vector and centroid of each point in the standard point cloud data, a smoothing operation can be performed on each point.
[0054] Specifically, it still involves acquiring points from the standard point cloud data in the head at once, with the currently acquired point being the current point, and simultaneously acquiring the current centroid and current normal vector corresponding to that current point.
[0055] Then, based on the current centroid and the current normal vector, the current projection plane is determined, and the current point is projected onto the current projection plane to obtain a smoothed point; the steps of determining the current centroid and the current normal vector corresponding to the current point are repeated until all points in the head standard point cloud data have undergone smoothing operation.
[0056] It is understandable that each projection is an operation that converges the points toward the center. For points in the center of the point cloud, the projection may not cause much displacement, and the projection planes of each point are likely to converge. For points on the periphery and noisy points, the projection plane will be closer to the center, thereby achieving the technical effects of noise reduction and smoothing.
[0057] In this way, a smoothing operation can be performed on each point in the standard point cloud data. After performing a smoothing operation on each point in a standard point cloud data, smoothed points can be obtained after multiple smoothing operations.
[0058] Step S500: Return to the step of acquiring real-time input radar point cloud data, save all the smoothed point cloud data and output the smoothed point cloud map.
[0059] In this embodiment, the smoothing method first performs a smoothing operation on the standard point cloud data at the head of the cache queue, and then performs smoothing on the next set of standard point cloud data. At the same time, after processing each set of point cloud data that has been taken out for smoothing, another set of standard point cloud data can be input into the cache queue.
[0060] Therefore, by repeating the steps between S100 and S400, in each iteration, the smoothed points of each standard point cloud data can be obtained. By saving and outputting these smoothed points, the smoothed point cloud image can be obtained.
[0061] Exemplary point cloud images before and after smoothing, for example Figure 3 and Figure 4 As shown. Figure 3 This is the original point cloud before smoothing. Figure 4 This is the smoothed point cloud. As you can see in the smoothed image, after removing... Figure 3 Various scattered points and noise are present, and the overall image plane thickness is significantly reduced.
[0062] The real-time point cloud smoothing method in this embodiment decouples spatial index maintenance from point cloud inflow through an octree map structure. Newly added point cloud data only requires incremental updates to the octree in the corresponding voxel, while old point cloud data is automatically forgotten according to set spatial and temporal thresholds. This ensures stable and controllable map memory usage, supporting end-to-end low-latency smoothing under continuous input and fundamentally solving the technical problem of not being able to smooth point clouds in real time. Furthermore, an adaptive plane search mechanism is introduced. When the flatness of the inner point cloud at the initial radius is greater than a preset value, the radius is automatically reduced and the search is iteratively repeated until the flatness is satisfied or the maximum number of iterations is reached. This mechanism makes the estimation of normal vectors and centroids more closely resemble the local geometric essence, avoiding oversmoothing or offset caused by traditional fixed-radius MLS algorithms at corners and beam / column edges, significantly improving smoothing accuracy and robustness.
[0063] Figure 5 A schematic diagram of a real-time point cloud smoothing device according to an embodiment of this application is shown. Exemplarily, the real-time point cloud smoothing device includes: The point cloud acquisition module 10 is used to acquire real-time input radar point cloud data and preprocess the radar point cloud data to obtain standard point cloud data. Cache module 20 is used to store the standard point cloud data from the tail of the queue into the cache queue, and input the standard point cloud data into the current point cloud map to update the point cloud map; Data processing module 30 is used to acquire the header standard point cloud data in the cache queue, process each point in the header standard point cloud data in sequence, and obtain the centroid and normal vector of each point in the header standard point cloud data. The smoothing module 40 is used to perform a smoothing operation on each point in the head standard point cloud data according to the normal vector and centroid of each point in the head standard point cloud data to obtain multiple smoothed points, and the multiple smoothed points are used to form smoothed point cloud data. The loop output module 50 is used to return to the step of acquiring the radar point cloud data in real time, save all the smoothed point cloud data and output the smoothed point cloud map.
[0064] It is understood that the apparatus in this embodiment corresponds to the real-time point cloud smoothing method in the above embodiments, and the options in the above embodiments are also applicable to this embodiment, so they will not be described again here.
[0065] This application also provides a terminal device, exemplary of which includes a processor and a memory, wherein the memory stores a computer program, and the processor executes the computer program to enable the terminal device to perform the functions of the various modules in the above-described real-time point cloud smoothing method or the above-described real-time point cloud smoothing device.
[0066] The processor can be an integrated circuit chip with signal processing capabilities. The processor can be a general-purpose processor, including at least one of a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Network Processor (NP), Digital Signal Processor (DSP), Application-Specific Integrated Circuit (ASIC), Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The general-purpose processor can be a microprocessor or any conventional processor, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
[0067] The memory can be, but is not limited to, Random Access Memory (RAM), Read Only Memory (ROM), Programmable Read-Only Memory (PROM), Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), etc. The memory is used to store computer programs, and the processor can execute the computer programs accordingly after receiving execution instructions.
[0068] This application also provides a readable storage medium for storing the computer program used in the aforementioned terminal device.
[0069] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings show the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that, in alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0070] In addition, the functional modules or units in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0071] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a smartphone, personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0072] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.
Claims
1. A real-time point cloud smoothing method, characterized in that, include: Acquire real-time input radar point cloud data and preprocess the radar point cloud data to obtain standard point cloud data; The standard point cloud data is stored from the tail of the queue into a cache queue, and the standard point cloud data is input into the current point cloud map to update the point cloud map. Obtain the header standard point cloud data in the cache queue, process each point in the header standard point cloud data in turn, and obtain the centroid and normal vector of each point in the header standard point cloud data. Sequentially acquire each point in the head standard point cloud data; determine the current centroid and current normal vector corresponding to the current point; Determine the current projection plane based on the current centroid and the current normal vector; Project the current point onto the current projection plane to obtain the corresponding smooth point; repeat the steps of determining the current centroid and current normal vector corresponding to the current point until the smooth points corresponding to all points in the head standard point cloud data are obtained; Return to the step of acquiring real-time input radar point cloud data, save all the smoothed point cloud data and output the smoothed point cloud map.
2. The real-time point cloud smoothing method according to claim 1, characterized in that, After inputting the standard point cloud data into the current point cloud map to update the point cloud map, the method further includes: In the point cloud map, points whose locations are outside a preset range are deleted; In the point cloud map, points whose update time exceeds a preset time will be deleted.
3. The real-time point cloud smoothing method according to claim 1, characterized in that, The radar point cloud data includes radar pose, displacement, and point cloud coordinates; The preprocessing of the radar point cloud data to obtain standard point cloud data includes: The radar pose and displacement corresponding to the radar point cloud data are obtained, and the coordinates of each point in the radar point cloud data are transformed to obtain radar point cloud data in the world coordinate system. The radar point cloud data in the world coordinate system is used as standard point cloud data.
4. The real-time point cloud smoothing method according to claim 1, characterized in that, The step of sequentially processing each point in the head standard point cloud data to obtain the centroid and normal vector of each point in the head standard point cloud data includes: Each point in the head standard point cloud data is obtained sequentially; In the point cloud map, a plane search with a preset radius is performed on the current point in the point cloud to obtain all neighboring points of the current point; Based on all the neighboring points and the current point, the average constructed centroid of the current point is calculated; The covariance matrix of the current point and all its neighboring points is constructed and eigenvalues are decomposed to obtain multiple eigenvalues. The eigenvector corresponding to the smallest eigenvalue is used as the normal vector of the current point.
5. The real-time point cloud smoothing method according to claim 4, characterized in that, After performing a planar search with a preset radius on the current point in the point cloud to obtain all neighboring points of the current point, the method further includes: Calculate the number of all neighboring points of the current point. If the total number of neighboring points is less than a preset value, discard the current point.
6. The real-time point cloud smoothing method according to claim 4, characterized in that, After using the eigenvector corresponding to the minimum eigenvalue as the normal vector of the current point, the method further includes: Obtain the largest eigenvalue among the plurality of eigenvalues; Calculate the flatness based on the maximum and minimum eigenvalues; If the flatness is greater than the preset flatness, then the preset radius is modified, and the step of obtaining all neighboring points of the current point is re-executed until the flatness is less than the preset flatness, or the number of iterations is greater than the preset number of repetitions, then the iteration stops.
7. A real-time point cloud smoothing device, characterized in that, include: The point cloud acquisition module is used to acquire real-time input radar point cloud data and preprocess the radar point cloud data to obtain standard point cloud data. The caching module is used to store the standard point cloud data from the tail of the queue into the cache queue, and input the standard point cloud data into the current point cloud map to update the point cloud map; The data processing module is used to acquire the header standard point cloud data in the cache queue, process each point in the header standard point cloud data in sequence, and obtain the centroid and normal vector of each point in the header standard point cloud data. The smoothing module is used to sequentially acquire each point in the head standard point cloud data; Determine the current centroid and current normal vector corresponding to the current point; Determine the current projection plane based on the current centroid and the current normal vector; Project the current point onto the current projection plane to obtain the corresponding smooth point; repeat the steps of determining the current centroid and current normal vector corresponding to the current point until the smooth points corresponding to all points in the head standard point cloud data are obtained; The loop output module is used to return to the step of acquiring the radar point cloud data in real time, save all the smoothed point cloud data and output the smoothed point cloud map.
8. A terminal device, characterized in that, The terminal device includes a processor and a memory, the memory storing a computer program, and the processor executing the computer program to implement the real-time point cloud smoothing method according to any one of claims 1-6.
9. A readable storage medium, characterized in that, It stores a computer program that, when executed on a processor, implements the real-time point cloud smoothing method according to any one of claims 1-6.