A point cloud map incremental dynamic updating method and device for environment three-dimensional modeling

Abstract

This invention discloses a method and device for incremental dynamic updating of point cloud maps in 3D environmental modeling, comprising the following steps: acquiring front-end data and constructing an environmental map; extracting keyframe features from the environmental map, performing loop closure detection, and acquiring a sensor pose map; optimizing the sensor pose map and correcting the keyframe poses; classifying the updated keyframes into historical keyframes and newly created keyframes, and sequentially performing local point cloud transient update, global point cloud transient update, and redundant historical keyframe deletion processing, finally outputting historical keyframe data with outdated point clouds and redundant keyframes removed. The beneficial effects of this invention are: improved system efficiency, reduced storage space, ensuring the map always reflects the latest state, and wide applicability.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, specifically to a method and device for incremental dynamic updating of point cloud maps for 3D environmental modeling. Background Technology

[0002] In the field of robotics, Simultaneous Localization and Mapping (SLAM) is a core method for achieving autonomous navigation and environmental perception. Traditional 3D Lidar SLAM typically assumes a static environment and lacks the ability to adapt to dynamic environments in real time, failing to meet the needs of long-term robot operation. To address the map freshness issue, map updating technology is introduced. This involves continuously acquiring new 3D point cloud data from the environment to update the original map data, maintaining consistency between the map and the real physical environment.

[0003] Existing map updates are divided into full map updates and incremental map updates. Full map update algorithms rely on periodically collecting and reconstructing the entire map area, which is inefficient, resource-intensive, and wasteful of computational resources. Existing incremental update methods typically perform pose convex optimization based on newly collected point cloud data combined with historical maps, and then directly overlay the newly collected and updated point cloud onto the historical map. This process fails to process the historical point cloud map, resulting in a large accumulation of outdated points, redundant keyframes in the pose graph, decreased positioning accuracy, poor robustness of point cloud matching, wasted map data storage resources, and reduced efficiency of online positioning computation. Therefore, there is an urgent need for an efficient and intelligent long-term incremental map update method. Summary of the Invention

[0004] The purpose of this invention is to provide a method for incremental dynamic updating of point cloud maps in 3D environmental modeling, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] In a first aspect, embodiments of the present invention provide a method for incremental dynamic updating of point cloud maps in 3D environmental modeling, comprising the following steps:

[0007] Obtain front-end data and build an environment map;

[0008] Extract keyframe features from the environmental map, perform loop closure detection, and obtain the sensor pose map;

[0009] Optimize the sensor pose graph and correct the pose of key frames;

[0010] After the sensor pose map is updated, the keyframes are divided into historical keyframes and newly created keyframes. After passing through local point cloud transient update, global point cloud transient update, and redundant historical keyframe deletion processing, the final output is historical keyframe data with outdated point cloud and redundant keyframes removed.

[0011] Preferably, the front-end data is acquired through IMU, 3D laser data, or GNSS sensors.

[0012] Preferably, the specific process of the local point cloud transient update processing is as follows:

[0013] Each laser point in a laser frame is treated as a scanning ray.

[0014] The relative distances of the dynamic attribute probabilities around each laser point are distributed according to a one-dimensional Gaussian distribution.

[0015] The dynamic attribute probability of the laser point is updated using Bayes' theorem.

[0016] Preferably, the global point cloud transient update process is as follows:

[0017] Acquire highly dynamic point cloud data with transient updates of local point clouds;

[0018] Remove outdated point cloud data spanning long periods of time, including global transient updates and recall of occluded points that were mistakenly deleted;

[0019] The nearest neighbor search difference is used to determine whether there is a newly created local point cloud near the historical key frame point, and outdated points are removed by probability update.

[0020] Simultaneously, by constructing a virtual ray backprojection onto the newly created keyframe, points that were mistakenly deleted due to occlusion can be detected and recalled.

[0021] Preferably, the specific process for deleting redundant historical keyframes is as follows:

[0022] Calculate the overlap rate between historical keyframes and newly created keyframes to initially screen out redundant keyframe nodes to be deleted.

[0023] Based on the different situations of nodes and edges, there are four types of deletion cases. By deleting old nodes and simultaneously finding new replacement nodes, the complexity of the overall pose graph is reduced.

[0024] Preferably, the four deletion scenarios are as follows:

[0025] Delete adjacent edges of historical nodes;

[0026] Delete the loop edge of the historical node;

[0027] Delete edges related to both historical and newly created nodes;

[0028] Delete isolated nodes from the past.

[0029] Secondly, the present invention provides an incremental dynamic update device for point cloud maps in 3D environmental modeling, used to implement the method described in any of the above embodiments, including:

[0030] Get the module, get the front-end data, and build the environment map;

[0031] The extraction module extracts key frame features from the environmental map, performs loop closure detection, and obtains the sensor pose map.

[0032] The optimization module optimizes the sensor pose map and corrects the pose of key frames.

[0033] The loading module divides the keyframes after the sensor pose map is updated into historical keyframes and newly created keyframes. After passing through local point cloud transient update, global point cloud transient update, and redundant historical keyframe deletion processing, the final output is historical keyframe data with outdated point cloud and redundant keyframes removed.

[0034] Thirdly, the present invention provides an electronic device including at least one processor, the processor being communicatively connected to at least one memory, wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the method described in any of the above embodiments.

[0035] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions that are used to cause a processor to execute the method described in any of the above embodiments.

[0036] The present invention has the following significant advantages over the prior art:

[0037] 1. Improve update efficiency: By using incremental updates to avoid global reconstruction, combined with the backend graph optimization module's accurate correction of keyframe poses and the historical map loading module's improvement of matching robustness, map update time is shortened, significantly improving system operating efficiency and reducing redundant calculations and resource waste caused by full updates.

[0038] 2. Optimize resource utilization: By deleting redundant historical keyframes, the complexity of the pose graph is effectively reduced, and the consumption of computing resources is reduced; at the same time, outdated point clouds are removed, saving storage space and achieving efficient utilization of storage resources.

[0039] 3. Enhanced dynamic adaptability and matching robustness: The loop closure detection of the localization module, combined with backend graph optimization, ensures the accuracy of keyframe poses, enabling the map to respond to environmental changes in real time; the historical map loading module further improves the robustness of point cloud matching by loading historical tile maps and fusing them with current data, ensuring that the map always reflects the real environmental state and providing reliable support for long-term robot localization and mapping.

[0040] 4. Wide range of applicable scenarios: The above technical solutions are easy to integrate and can be adapted to various devices such as indoor service robots and outdoor unmanned vehicles, and have strong versatility. Attached Figure Description

[0041] Figure 1 is an overall flowchart of the method according to an embodiment of the present invention;

[0042] Figure 2 is a functional block diagram of the device according to an embodiment of the present invention;

[0043] Figure 3 shows a technical solution diagram for updating historical map data;

[0044] Figure 4 is a schematic diagram of the principle of the local transient loading module;

[0045] Figure 5 shows the technical solution for the global point cloud transient loading module;

[0046] Figure 6 shows the recall scheme for accidental deletion of occluded points;

[0047] Figure 7 shows four types of deletion scenarios for the redundant keyframe deletion module;

[0048] Figure 8 This is a block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] Please see Figure 1 The present invention provides a technical solution:

[0051] An incremental dynamic update method for point cloud maps in 3D environmental modeling includes the following steps:

[0052] S100. Obtain front-end data and build an environment map;

[0053] S200. Extract key frame features from the environmental map, perform loop closure detection, and obtain the sensor pose map;

[0054] S300. Optimize the sensor pose graph and correct the keyframe pose;

[0055] S400 divides the keyframes after the sensor pose map is updated into historical keyframes and newly created keyframes. After passing through local point cloud transient update, global point cloud transient update, and redundant historical keyframe deletion processing in sequence, it finally outputs historical keyframe data with outdated point cloud and redundant keyframes removed.

[0056] In embodiments of the present invention, such as Figure 2 As shown, in order to synchronize with the implementation of the above steps, a specific embodiment of the present invention also provides an incremental dynamic update device for point cloud maps of 3D environmental modeling. The device includes an acquisition module 100, an extraction module 200, an optimization module 300, and a loading module 400.

[0057] The method in this embodiment can be used for robots to update robot maps in real time, without relying on periodic map reconstruction as in traditional technologies. This avoids problems such as the accumulation of a large number of "outdated points" in the point cloud map (such as historical points that do not exist in new data, such as temporary construction areas and seasonally changing leaves), redundant keyframes in the pose graph, which lead to decreased positioning accuracy, poor robustness of point cloud matching, waste of storage resources, and reduced efficiency of online positioning calculations. This method enables robots to achieve efficient and intelligent long-term incremental map updates.

[0058] It should be noted that the term "robot" in this application is used in a broad sense, encompassing various automated devices with autonomous navigation, environmental perception, and map building / updating capabilities.

[0059] In an embodiment of the present invention, the front-end data in step S100 receives newly acquired IMU, 3D laser data, and GNSS sensor data through the acquisition module 100, estimates the current robot pose in real time through the state estimator, and constructs a map of the surrounding environment using an online voxel map.

[0060] In an embodiment of the present invention, step S200 extracts key frame features through extraction module 200 and performs loop closure detection with historical map.

[0061] In an embodiment of the present invention, the process of extracting keyframe features is as follows:

[0062] The system receives real-time data from 3D LiDAR (10 frames / second), IMU (100Hz), and GNSS (10Hz), and estimates the robot's real-time pose through point cloud registration and multi-sensor fusion. When the robot moves a distance of 0.5 meters or its pose changes significantly, a keyframe is generated, and features such as edge points and planar points are extracted from the keyframe (for subsequent matching).

[0063] In embodiments of the present invention, keyframe features include geometric features of the point cloud (such as edges and planes) and global descriptors (for matching with historical maps).

[0064] In an embodiment of the present invention, the loop closure detection process includes:

[0065] Extract global descriptors (used to characterize the global features of the point cloud in the current keyframe) from the current keyframe.

[0066] The descriptor is matched with the features of keyframes in the historical map, and the matching threshold is set to 0.8.

[0067] If a loop is detected, meaning the current keyframe matches a historical keyframe, then the loop edge is recorded.

[0068] In an embodiment of the present invention, step S300 specifically includes:

[0069] The process involves acquiring "nodes" from keyframes as the pose graph, and constructing a global pose graph using two types of "edges" as constraints. These two types of "edges" include odometry edges and loop closure edges. Specifically, odometry edges are formed by the relative poses between adjacent keyframes calculated by the front-end odometry module, creating constraint relationships between adjacent nodes. Loop closure edges are obtained by the extraction module 200 through loop closure detection, representing the pose constraint relationships between the current keyframe and historical keyframes.

[0070] Based on this, the optimization module 300 uses the nonlinear least squares method to perform global optimization on the constructed pose graph. By minimizing the error of each constraint edge in the pose graph, it eliminates the cumulative error caused by sensor drift and other factors, and finally outputs the corrected keyframe pose, thereby achieving accurate correction of the global pose and thus enabling map calibration, providing a reliable pose association basis for subsequent historical map data updates.

[0071] In an embodiment of the present invention, the loading module 400 comprises a local point cloud transient loading module, a global point cloud transient loading module, and a redundant historical keyframe deletion module. In a specific embodiment, the local point cloud transient loading module, the global point cloud transient loading module, and the redundant historical keyframe deletion module can be integrated into a dynamic tile map loader, which corrects the map calibrated by the optimization module 300 during the point cloud data loading process before loading it to the acquisition module 100.

[0072] In an embodiment of the present invention, the specific processing procedure of step S400 is as follows:

[0073] like Figure 3As shown, the corrected keyframe pose is divided into historical keyframes and newly created keyframes, which are then sequentially loaded into the local point cloud transient loading module, the global point cloud transient loading module, and the redundant historical keyframe deletion module. Finally, the historical keyframe data with outdated point clouds and redundant keyframes is output.

[0074] Specifically, in this application, "transient" refers to the dynamic attribute probability of a point cloud, "local transient" refers to the dynamic probability of a point cloud within a short period (during a single data acquisition), and "global transient" refers to the dynamic probability of a point cloud over a longer time span. The principle of local transient update is as follows: Figure 4 As shown.

[0075] Specifically, each laser point scan in a frame is considered as a scanning ray, where the black point is the origin of the laser scan, the white point is the free point f sampled in the free space along the ray's path, and the gray point is the occupied point o where the ray finally hits the object's surface. After aggregating keyframes into a local map, the ɛl around each occupied point oi or free point fi follows a one-dimensional Gaussian distribution with respect to distance (x): Where σ is the standard deviation, ɛl is initially set to 0.5, k nearest neighbors are retrieved for each point in Oi or Fi, α is the upper limit of the dynamic attribute probability of the point cloud, β is the offset adjustment coefficient of the dynamic attribute probability, and ɛl of these nearest neighbors is updated using Bayes' theorem: The technical solution for the global point cloud transient loading module is as follows: Figure 5 As shown, global point cloud transient update refers to the process of first filtering highly dynamic point clouds through local transients, and then further removing outdated point clouds with long time spans.

[0076] In an embodiment of the present invention, in the global transient update, local point cloud maps are constructed for historical keyframes and newly created keyframes respectively, and the differences in nearest neighbor search are used to determine which historical keyframe points do not have newly created local point clouds nearby, thereby gradually updating through a probabilistic update method, and finally, points greater than a set threshold are initially regarded as outdated points.

[0077] An anomaly exists here: when a new keyframe map is created, some historical observation data may be in the blind zone of the current scan due to some obstructions. These historical observation points will also be regarded as outdated points, but in fact these points are not outdated.

[0078] In embodiments of the present invention, methods for handling abnormal phenomena, such as... Figure 6 As shown, a 3D voxel map is constructed based on the point clouds of historical keyframes and newly created keyframes. When an outdated point is detected, a virtual ray is constructed and back-projected into the newly created keyframe. When a voxel in the newly created keyframe is detected to be occupying the virtual ray, the outdated point can be identified as a mistakenly deleted point due to occlusion.

[0079] The dynamic update at the point cloud level is now complete. The final step is to remove redundant keyframes using the redundant keyframe removal module. In actual map updates, some areas may be updated frequently. Over time, the pose optimization map will accumulate a large number of redundant nodes, leading to increased optimization time and the accumulation of duplicate point clouds.

[0080] In embodiments of the present invention, the scheme for removing redundant keyframes includes:

[0081] Calculate the overlap rate between historical keyframes and newly created keyframes to initially screen out redundant keyframe nodes to be deleted. The overlap rate between historical and newly created keyframes is calculated using the overlap ratio of the voxel mesh, as shown in the following formula: Redundant keyframe nodes to be deleted are initially identified. Based on the differences between the nodes and edges to be deleted, four deletion scenarios can be identified, as detailed below. Figure 7 As shown, it includes:

[0082] Delete adjacent edges of historical nodes;

[0083] Delete the loop edge of the historical node;

[0084] Delete edges related to both historical and newly created nodes;

[0085] Delete isolated nodes from the past.

[0086] By deleting old nodes and synchronously searching for new replacement nodes, redundant historical keyframes are removed after each update, reducing the complexity of the overall pose graph. Loading module 400 loads the point cloud data after deleting redundant historical keyframes into acquisition module 100, completing the incremental update of the map.

[0087] In a specific embodiment of the present invention, the robot is equipped with an incremental dynamic update device for point cloud maps of three-dimensional environmental modeling provided in the embodiments of the present invention. During the operation of the robot, the acquisition module receives real-time data from 3D LiDAR (10 frames / second), IMU (100Hz) and GNSS (10Hz), and generates a key frame every 0.5 meters or a significant angle change by locating the robot's walking distance.

[0088] In a specific embodiment of the present invention, the extraction module 200 extracts the global descriptor of the key frame and matches it with the features of the key frame in the historical map, with the matching threshold set to 0.8.

[0089] In a specific embodiment of the present invention, the optimization module 300 loads a local point cloud slice with a radius of 20 meters based on the current pose and converts it into a voxel map with a voxel size of 0.1 meters.

[0090] In a specific embodiment of the present invention, in the local point cloud transient loading module, each point retrieves 10 nearest neighbor points with a search radius of 0.2 meters, an initial dynamic probability of 0.5, a standard deviation of 0.1 meters, and points with a dynamic probability greater than 0.9 are marked as high dynamic points.

[0091] In a specific embodiment of the present invention, in the global point cloud transient loading module, the voxel size of the voxel map constructed from historical and newly created keyframe point clouds is 0.2 meters. Five nearest neighbor points are retrieved for historical point cloud points, with a search half distance of 0.5 meters. The initial value of the global dynamic probability is 0.5, the standard deviation is 0.2 meters, and points with a probability greater than 0.95 are initially marked as outdated points.

[0092] In a specific embodiment of the present invention, the overlap rate is calculated using the formula overlap_ratio in the redundant historical keyframe deletion module, and the overlap rate threshold is 0.85.

[0093] In a specific embodiment of the present invention, steps S100 to S300 are the incremental mapping stage, and step S400 is the historical map data update stage. The incremental mapping stage and the historical map data update stage begin when the system starts by loading historical 3D point cloud maps and pose maps from local storage or a cloud server. If it is the first run, an empty map is initialized. The vehicle is equipped with a 3D LiDAR, IMU, and GNSS to collect environmental data in real time. The LiDAR generates 10 frames of point cloud per second, the IMU provides a high-frequency attitude of 100Hz, and the GNSS provides a global position reference of 10Hz.

[0094] In a specific embodiment of the present invention, in actual testing, the method was run for 6 months in an office park of 200m×250m, completing a total of six rounds of updates, processing 500GB of data, reducing storage space from 1.2TB to 780GB, saving about 35%, maintaining positioning accuracy within 0.1 meters, and significantly improving robustness in dynamic environments such as construction areas and vegetation changes.

[0095] Please see Figure 8 , Figure 8 A schematic diagram of the mechanism of an electronic device 20 that can implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of control devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the invention described and / or claimed herein.

[0096] Electronic device 20 includes at least one processor 21 and a memory, such as read-only memory (ROM) 22 and random access memory (RAM) 23, communicatively connected to at least one processor 21. The memory stores computer programs executable by at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 22 or loaded from storage unit 28 into the RAM 13. The RAM 23 may also store various programs and data required for the operation of electronic device 20. The processor 21, ROM 22, and RAM 23 are interconnected via bus 24. Input / output (I / O) interface 25 is also connected to bus 24.

[0097] Multiple components in electronic device 20 are connected to I / O interface 25, including: input unit 26, such as keyboard, mouse, etc.; output unit 27, such as various types of monitors, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 29, such as network card, modem, wireless transceiver, etc. Communication unit 29 allows electronic device 20 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0098] Processor 21 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 21 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 21 performs the various methods and processes described above.

[0099] In some embodiments, the methods described above can be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 28. In some embodiments, part or all of the computer program can be loaded and / or installed on electronic device 20 via ROM 22 and / or communication unit 29. When the computer program is loaded into RAM 23 and executed by processor 21, one or more steps of the methods described above can be performed. Alternatively, in other embodiments, processor 21 can be configured to perform the methods described above by any other suitable means (e.g., by means of firmware).

[0100] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0101] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0102] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0103] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0104] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0105] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0106] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for incremental dynamic updating of point cloud maps in 3D environmental modeling, characterized in that, Includes the following steps: Obtain front-end data and build an environment map; Extract keyframe features from the environmental map, perform loop closure detection, and obtain the sensor pose map; Optimize the sensor pose graph and correct the pose of key frames; After the sensor pose map is updated, the key frames are divided into historical key frames and newly created key frames. After passing through local point cloud transient update, global point cloud transient update and redundant historical key frame deletion processes, the final output is historical key frame data with outdated point cloud and redundant key frames removed. The specific process of the local point cloud transient update processing is as follows: Each laser point in a laser frame is treated as a scanning ray. The relative distances of the dynamic attribute probabilities around each laser point are distributed according to a one-dimensional Gaussian distribution. The dynamic attribute probability of the nearest point of the laser point is updated using Bayes' theorem. The global point cloud transient update process is as follows: Acquire and filter highly dynamic point cloud data with transient local point cloud updates; Remove outdated point cloud data spanning long periods of time, including global transient updates and recall of occluded points that were mistakenly deleted; The nearest neighbor search difference is used to determine whether there is a newly created local point cloud near the historical key frame point, and outdated points are removed by probability update. Simultaneously, by constructing a virtual ray backprojection onto the newly created keyframe, points that were mistakenly deleted due to occlusion can be detected and recalled.

2. The method for incremental dynamic updating of point cloud maps for 3D environmental modeling according to claim 1, characterized in that: The front-end data is acquired through IMU, 3D laser data, and GNSS sensors.

3. The method for incremental dynamic updating of point cloud maps for 3D environmental modeling according to claim 2, characterized in that: The specific process for deleting redundant historical keyframes is as follows: Calculate the overlap rate between historical keyframes and newly created keyframes to initially screen out redundant keyframe nodes to be deleted. Based on the different situations of nodes and edges, four types of deletion scenarios are identified. By deleting old nodes and simultaneously searching for new replacement nodes, the complexity of the overall pose graph is reduced. The four types of deletion scenarios are as follows: Delete adjacent edges of historical nodes, delete loop edges of historical nodes, delete edges related to historical and newly created nodes, and delete isolated edges of historical nodes.

4. A point cloud map incremental dynamic update device for environmental 3D modeling, used to implement the method described in any one of claims 1 to 3, characterized in that, include: Get the module, get the front-end data, and build the environment map; The extraction module extracts key frame features from the environmental map, performs loop closure detection, and obtains the sensor pose map. The optimization module optimizes the sensor pose map and corrects the pose of key frames. The loading module divides the keyframes after the sensor pose map is updated into historical keyframes and newly created keyframes. After passing through local point cloud transient update, global point cloud transient update, and redundant historical keyframe deletion processing, the final output is historical keyframe data with outdated point cloud and redundant keyframes removed. The specific process of the local point cloud transient update processing is as follows: Each laser point in a laser frame is treated as a scanning ray. The relative distances of the dynamic attribute probabilities around each laser point are distributed according to a one-dimensional Gaussian distribution. The dynamic attribute probability of the nearest point of the laser point is updated using Bayes' theorem. The global point cloud transient update process is as follows: Acquire and filter highly dynamic point cloud data with transient local point cloud updates; Remove outdated point cloud data spanning long periods of time, including global transient updates and recall of occluded points that were mistakenly deleted; The nearest neighbor search difference is used to determine whether there is a newly created local point cloud near the historical key frame point, and outdated points are removed by probability update. Simultaneously, by constructing a virtual ray backprojection onto the newly created keyframe, points that were mistakenly deleted due to occlusion can be detected and recalled.

5. An electronic device, characterized in that, The method includes at least one processor, which is communicatively connected to at least one memory, wherein the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the method of any one of claims 1 to 3.

Images