Automatic updating method of laser point cloud navigation map

Abstract

The invention discloses an automatic updating method for a laser point cloud navigation map, and the method is realized through an automatic updating system for the laser point cloud navigation map. The automatic updating system comprises a data acquisition module, a data processing module, a data fusion module, a data verification module, a comparative analysis module, an incremental learning module, a map updating module, a map storage and application module and a cloud platform. The automatic updating method comprises the following steps: step 1, constructing a dynamic environment characteristic real-time monitoring system; 2, establishing a distributed computing and cloud platform architecture; 3, designing a map updating algorithm based on incremental learning; according to the automatic updating method for the laser point cloud navigation map, the navigation map is automatically updated in real time according to the updated map data. In the whole data acquisition and updating process, through strict quality control, it is ensured that the system can operate efficiently and stably, and accurate and real-time navigation map service is provided for intelligent equipment.

Description

Technical Field

[0001] This invention belongs to the field of intelligent navigation technology and relates to an automatic updating method for laser point cloud navigation maps. Background Technology

[0002] As time progresses and the environment dynamically changes, navigation maps face numerous challenges. In cities, new buildings rise, roads are altered due to construction or renovation, and traffic signs and markings may be updated or worn out. In industrial plants, equipment relocation, the addition of new shelving, or adjustments to aisle layouts all cause existing laser point cloud navigation maps to gradually deviate from the actual environmental conditions. Traditional map updates often rely on manual operation, requiring significant investment of manpower, resources, and time, and have long update cycles, making it difficult to meet the urgent needs of smart devices for real-time, accurate navigation. This leads to problems such as positioning deviations and path planning errors in smart devices during operation due to relying on outdated map information, seriously affecting their working efficiency and safety. Summary of the Invention

[0003] In order to overcome at least one deficiency of the prior art, the present invention provides an automatic updating method for laser point cloud navigation maps.

[0004] To achieve the above objectives, the present invention adopts the following technical solution: an automatic update method for laser point cloud navigation maps. This method is implemented through an automatic update system for laser point cloud navigation maps. The automatic update system includes a data acquisition module, a data processing module, a data fusion module, a data verification module, a comparative analysis module, an incremental learning module, a map update module, a map storage and application module, and a cloud platform. The automatic update method includes the following steps: Step 1: Construct a real-time monitoring system for dynamic environmental characteristics: Collect and preprocess multi-source data of the surrounding environment using different data sources; Step 2: Establish a distributed computing and cloud platform architecture: for the storage, management, and execution of large-scale computing tasks of map data; Step 3: Design an incremental learning-based map update algorithm; The incremental learning-based map update algorithm compares and analyzes the validated new data with the existing map data, identifies areas of environmental change, and updates the map accordingly; Step 4: Store the verified updated map data back to the cloud platform and synchronize it to relevant smart devices in a timely manner.

[0005] Furthermore, the method for constructing a real-time monitoring system for dynamic environmental characteristics includes: Step 11: Equip the smart device with a data source and use high-precision clock synchronization technology to collect surrounding environmental data, generating real-time laser point cloud data, visual image data, sensor data, and other multi-source data. Step 12: Preprocess the raw multi-source data to remove noise, perform motion compensation, and calibration; Step 13: Analyze the processed laser point cloud data using machine learning algorithms, extract dynamic environmental features, and store and cache the data according to a specific data format.

[0006] Furthermore, the method for establishing a distributed computing and cloud platform support architecture involves local smart devices connecting to the cloud platform via a wireless network. The multi-core processor and GPU on the device perform preliminary processing on the laser point cloud data, and the processed dynamic environmental features are uploaded to the cloud platform. The cloud platform's computing cluster decomposes the map update task into multiple sub-tasks through a distributed computing framework, which are then distributed to different computing nodes for parallel execution. The cloud platform stores and manages the map data, and data download, update, and synchronization with the smart devices are achieved through data transmission lines.

[0007] Furthermore, the method for establishing a map update algorithm based on incremental learning to process dynamic environmental change information includes: Step 30: Establish a multi-source data fusion and verification mechanism to fuse and verify newly acquired laser point cloud data with existing navigation map data; Step 31: Compare and analyze the verified data with existing navigation map data, use feature matching and difference detection algorithms to locate areas in the environment that have changed, and extract their geometric and semantic features; Step 32: Update the map using an incremental learning algorithm based on the detected changed areas.

[0008] Furthermore, the multi-source data fusion and verification mechanism includes methods for fusing and verifying data. Step 301: Fuse the laser point cloud data, visual image data, and sensor data; Step 302: Use cross-validation to verify the fused data. Compare the information of the same object or area obtained from different data sources to determine the consistency and reliability of the data. If the data verification is successful, the reliable data is transmitted to the map update module for map updating. If there are problems with the data, it is fed back to the data source for data correction or re-collection.

[0009] Furthermore, the data processing module receives multi-source data collected by the data acquisition module, uses machine learning algorithms to extract features from the data, and the processing steps include data preprocessing, feature classification, target recognition, and the extraction of dynamic environmental feature information.

[0010] Furthermore, the data fusion module receives LiDAR data, visual image data, and sensor data processed by the data processing module. It uses a multi-source data fusion algorithm to fuse different types of data, employing data alignment, feature fusion, and information complementarity techniques. The fused data is then input into the data verification module. Through a cross-validation mechanism, information about the same object or region obtained from different data sources is compared to determine the consistency and reliability of the data. If the data verification passes, the reliable data is transmitted to the map update module for map updating. If there are problems with the data, it is returned to the corresponding data source for data correction or re-collection.

[0011] Furthermore, the comparison analysis module receives the laser point cloud data verified by the data verification module, and uses a feature matching algorithm to compare the new point cloud data with the features in the existing map data. If there are differences in the comparison, a difference detection algorithm is used to accurately identify the environmental change area, and the environmental feature information of the environmental change area is transmitted to the incremental learning module.

[0012] Furthermore, the incremental learning module receives environmental change area data analyzed by the comparative analysis module. Based on the new environmental feature information, the module dynamically adjusts the topology and geometric information of the map. The map update module performs local map updates based on the topology and geometric information. The updated map data is output to the map storage and application module. If problems are found during the update process, the parameters are readjusted and the update is performed.

[0013] Furthermore, the cloud platform is equipped with a distributed computing framework, which decomposes the map update task into multiple sub-tasks and distributes them to different computing nodes for simultaneous computation.

[0014] In summary, the advantages of this invention are: The automatic update method for laser point cloud navigation maps of this invention preprocesses the collected raw data, removing noise, performing motion compensation and calibration, extracting dynamic environmental features, and uploading the processed data to a cloud platform. Upon receiving the data, the cloud platform first verifies and fuses the data to ensure its accuracy and reliability. Using an incremental learning-based map update algorithm, the new data is compared and analyzed with existing map data to identify areas of environmental change and update the map accordingly. During the update process, a multi-source data fusion and verification mechanism is used to repeatedly verify the update results, ensuring the accuracy and stability of the map update. The updated map data is stored back on the cloud platform and promptly synchronized to relevant smart devices, achieving automatic and real-time updates of the navigation map. Throughout the entire data acquisition and update process, strict quality control ensures the system operates efficiently and stably, providing accurate and real-time navigation map services to smart devices. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the automatic update method for the laser point cloud navigation map of the present invention.

[0016] Figure 2 This is a diagram illustrating the architecture of the automatic update system for the laser point cloud navigation map of the present invention. Detailed Implementation

[0017] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that, unless otherwise specified, the following embodiments and features described therein can be combined with each other.

[0018] Example: like Figures 1-2 As shown, an automatic update method for a laser point cloud navigation map is provided. This method is implemented through an automatic update system for the laser point cloud navigation map. The automatic update system includes a data acquisition module, a data processing module, a data fusion module, a data verification module, a comparative analysis module, an incremental learning module, a map update module, a map storage and application module, and a cloud platform. The automatic update method includes the following steps: Step 1: Construct a real-time monitoring system for dynamic environmental characteristics: Collect and preprocess multi-source data of the surrounding environment using different data sources; Step 2: Establish a distributed computing and cloud platform architecture: for the storage, management, and execution of large-scale computing tasks of map data; Step 3: Design an incremental learning-based map update algorithm; The incremental learning-based map update algorithm compares and analyzes the validated new data with the existing map data, identifies areas of environmental change, and updates the map accordingly; Step 4: Store the verified updated map data back to the cloud platform and synchronize it to relevant smart devices in a timely manner.

[0019] Methods for constructing a real-time monitoring system for dynamic environmental characteristics include: Step 11: Equip the smart device with a data source and use high-precision clock synchronization technology to collect surrounding environmental data, generating real-time laser point cloud data, visual image data, sensor data, and other multi-source data. Data sources include high-definition cameras and LiDAR, acquiring visual image data and laser point cloud data of the surrounding environment. Image recognition technologies, such as deep learning-based object classification and recognition algorithms, are used to semantically verify and supplement objects in the laser point cloud data. For example, if the laser point cloud data detects an object that might be a traffic sign but cannot accurately determine its specific type, combining it with visual image data and using image recognition algorithms can clearly identify the traffic sign as a "No Left Turn" sign, thus providing more comprehensive and accurate information for map updates. Laser point cloud data can reflect the three-dimensional spatial position and shape information of objects in the environment in great detail, providing a rich data foundation for subsequent feature extraction; Step 12: Preprocess the raw multi-source data to remove noise, perform motion compensation, and calibration; When denoising the raw laser point cloud data, a deep learning-based denoising algorithm is employed. This algorithm, fully trained on a large amount of real point cloud data, can effectively identify and remove noise points. A bilateral filtering algorithm is then used to further filter the data, removing noise while preserving the edge and detail features of the point cloud data to the greatest extent possible. Combining IMU and odometry data, motion compensation and calibration are performed on the laser point cloud data using a Kalman filter algorithm to accurately correct point cloud position deviations caused by equipment movement. Step 13: Analyze the processed laser point cloud data using machine learning algorithms, extract dynamic environmental features, and store and cache the data according to a specific data format; For road scenarios, the focus is on identifying dynamic targets such as vehicles, pedestrians, traffic signs, and construction areas. In industrial scenarios, the focus is on key information such as equipment movement, changes in cargo stacking, and temporary occupancy of passageways. Taking vehicle detection as an example, a deep learning-based target detection model can quickly and accurately identify the position and outline of vehicles from complex laser point cloud data. Real-time monitoring and precise extraction of these dynamic environmental features provide a timely and accurate information source for the automatic updating of navigation maps.

[0020] Methods for establishing distributed computing and cloud platform support architectures include Local smart devices connect to the cloud platform via a wireless network. The device's multi-core processor and GPU perform initial processing of the laser point cloud data. The processed dynamic environmental features are then uploaded to the cloud platform. The cloud platform's computing cluster uses a distributed computing framework to decompose the map update task into multiple sub-tasks, which are then distributed across different computing nodes for parallel execution. The cloud platform stores and manages the map data, and enables data download, updates, and synchronization with the smart devices via data transmission lines.

[0021] Methods for developing map update algorithms based on incremental learning to handle dynamic environmental changes include: Step 30: Establish a multi-source data fusion and verification mechanism to fuse and verify newly acquired laser point cloud data with existing navigation map data; Step 31: Compare and analyze the verified data with existing navigation map data, use feature matching and difference detection algorithms to locate areas in the environment that have changed, and extract their geometric and semantic features; For example, in a road scenario, by comparing laser point cloud data at adjacent times, a new construction fence is detected on a certain road section. The algorithm can accurately identify the area where the construction fence is located and extract its geometric and semantic features. Step 32: Update the map using an incremental learning algorithm based on the detected changed areas; Incremental learning algorithms allow models to gradually update their parameters using new data without retraining on all the data, thus enabling them to quickly adapt to changes in the environment.

[0022] During map updates, new environmental features are incorporated into the existing map model, and the map's topology and geometric information are dynamically adjusted.

[0023] Taking road map updates as an example, if a change in lane markings is detected on a road, the algorithm updates the lane information of that road on the map based on the new lane marking features. This includes the number of lanes, lane width, and lane curvature, ensuring that the map can accurately reflect the actual environmental conditions in real time. Simultaneously, by learning from and analyzing historical update data, the algorithm can continuously optimize its update strategy, improving update efficiency and accuracy.

[0024] Multi-source data fusion and verification mechanisms include methods for data fusion and verification, such as: Step 301: Fuse the laser point cloud data, visual image data, and sensor data; Fusion processing includes data alignment, feature fusion, and information complementarity; Step 302: Use cross-validation to verify the fused data. Compare the information of the same object or area obtained from different data sources to determine the consistency and reliability of the data. If the data verification is successful, the reliable data is transmitted to the map update module for map updating. If there are problems with the data, it is fed back to the data source for data correction or re-collection.

[0025] The data acquisition module accurately scans the surrounding environment to obtain various multi-source data, including laser point cloud data, visual image data, and sensor data. The data acquisition module includes multiple acquisition devices such as lidar, camera, IMU, and odometer to acquire laser point cloud data, visual image data, and sensor data in real time.

[0026] The data processing module receives multi-source data collected by the data acquisition module, uses machine learning algorithms to extract features from the data, and processes the data through steps such as data preprocessing, feature classification, and target recognition. It also extracts dynamic environmental feature information, such as vehicles, pedestrians, and traffic signs.

[0027] The data fusion module receives LiDAR data, visual image data, and sensor data processed by the data processing module, and uses multi-source data fusion algorithms to fuse different types of data. It employs technologies such as data alignment, feature fusion, and information complementarity to perform the fusion operation.

[0028] The merged data enters the data verification module, which uses a cross-validation mechanism to compare information about the same object or area obtained from different data sources to determine the consistency and reliability of the data. If the data verification passes, the reliable data is transmitted to the map update module for map updates; if there are problems with the data, it is returned to the corresponding data source for data correction or re-collection to ensure the quality of the map update data.

[0029] The comparison and analysis module receives the laser point cloud data verified by the data verification module, and uses a feature matching algorithm to compare the new point cloud data with the features in the existing map data. If there are differences, the difference detection algorithm is used to accurately identify the environmental change area, and the environmental feature information of the environmental change area is transmitted to the incremental learning module. The incremental learning module receives environmental change data analyzed by the comparative analysis module. Based on the new environmental features, this module dynamically adjusts the topology and geometric information of the map.

[0030] The map update module performs local map updates based on topological and geometric information, such as node additions, edge updates, and attribute adjustments. The updated map data is then output to the map storage and application module. If any issues are detected during the update process, parameters are readjusted before resuming the update to ensure the accuracy and reliability of the map update.

[0031] The cloud platform employs a distributed computing framework, breaking down map update tasks into multiple sub-tasks and distributing them across different computing nodes for simultaneous computation. When handling city-scale map update tasks, distributed computing can reduce computation time from several hours in traditional single-machine computing to tens of minutes, significantly improving computational efficiency. The cloud platform stores and manages map data, providing download, update, and synchronization services for map data to smart devices. Through high-speed network connections, smart devices can promptly obtain the latest map data and upload locally updated map data to the cloud platform for backup and sharing. The cloud platform utilizes big data analytics to statistically analyze the map update frequency and content across different regions and time periods, providing data support for optimizing map update strategies. Based on the analysis results, it rationally adjusts the allocation of computing resources to improve the overall performance and efficiency of the map update service.

[0032] Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort should fall within the scope of protection of the present invention.

Claims

1. An automatic updating method for laser point cloud navigation maps, characterized in that: This method is implemented through an automatic update system for laser point cloud navigation maps. The automatic update system includes a data acquisition module, a data processing module, a data fusion module, a data verification module, a comparative analysis module, an incremental learning module, a map update module, a map storage and application module, and a cloud platform. The automatic update method includes the following steps: Step 1: Construct a real-time monitoring system for dynamic environmental characteristics: Collect and preprocess multi-source data of the surrounding environment using different data sources; Step 2: Establish a distributed computing and cloud platform architecture: for the storage, management, and execution of large-scale computing tasks of map data; Step 3: Design an incremental learning-based map update algorithm; The incremental learning-based map update algorithm compares and analyzes the validated new data with the existing map data, identifies areas of environmental change, and updates the map accordingly; Step 4: Store the verified updated map data back to the cloud platform and synchronize it to relevant smart devices in a timely manner.

2. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The method for constructing a real-time monitoring system for dynamic environmental characteristics includes: Step 11: Equip the smart device with a data source and use high-precision clock synchronization technology to collect surrounding environmental data, generating real-time laser point cloud data, visual image data, sensor data, and other multi-source data. Step 12: Preprocess the raw multi-source data to remove noise, perform motion compensation, and calibration; Step 13: Analyze the processed laser point cloud data using machine learning algorithms, extract dynamic environmental features, and store and cache the data according to a specific data format.

3. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The method for establishing a distributed computing and cloud platform support architecture involves local smart devices connecting to the cloud platform via a wireless network. The multi-core processor and GPU on the device perform preliminary processing on the laser point cloud data. The processed dynamic environmental features are then uploaded to the cloud platform. The cloud platform's computing cluster decomposes the map update task into multiple sub-tasks through a distributed computing framework, which are then distributed to different computing nodes for parallel execution. The cloud platform stores and manages the map data and enables data download, update, and synchronization with the smart devices through data transmission lines.

4. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The method for establishing a map update algorithm based on incremental learning to process dynamic environmental change information includes: Step 30: Establish a multi-source data fusion and verification mechanism to fuse and verify the newly acquired laser point cloud data with existing navigation map data; Step 31: Compare and analyze the verified data with existing navigation map data, use feature matching and difference detection algorithms to locate areas in the environment that have changed, and extract their geometric and semantic features; Step 32: Update the map using an incremental learning algorithm based on the detected changed areas.

5. The automatic updating method for a laser point cloud navigation map according to claim 4, characterized in that: The multi-source data fusion and verification mechanism includes methods for data fusion and verification. Step 301: Fuse the laser point cloud data, visual image data, and sensor data; Step 302: Use cross-validation to verify the fused data. Compare the information of the same object or area obtained from different data sources to determine the consistency and reliability of the data. If the data verification is successful, the reliable data is transmitted to the map update module for map updating. If there are problems with the data, it is fed back to the data source for data correction or re-collection.

6. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The data processing module receives multi-source data collected by the data acquisition module, uses machine learning algorithms to extract features from the data, and the processing steps include data preprocessing, feature classification, target recognition, and the extraction of dynamic environmental feature information.

7. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The data fusion module receives LiDAR data, visual image data, and sensor data processed by the data processing module. It uses a multi-source data fusion algorithm to fuse different types of data, employing data alignment, feature fusion, and information complementarity techniques. The fused data is then input into the data verification module. Through a cross-validation mechanism, information about the same object or region obtained from different data sources is compared to determine the consistency and reliability of the data. If the data verification passes, the reliable data is transmitted to the map update module for map updating. If there are problems with the data, it is returned to the corresponding data source for data correction or re-collection.

8. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The comparison analysis module receives the laser point cloud data verified by the data verification module, and uses a feature matching algorithm to compare the new point cloud data with the features in the existing map data. If there are differences, the difference detection algorithm is used to accurately identify the environmental change area, and the environmental feature information of the environmental change area is transmitted to the incremental learning module.

9. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The incremental learning module receives environmental change area data analyzed by the comparative analysis module. Based on the new environmental feature information, the module dynamically adjusts the topology and geometric information of the map. The map update module performs local map updates based on the topology and geometric information. The updated map data is output to the map storage and application module. If a problem is found during the update process, the parameters are readjusted and the update is performed.

10. The automatic updating method for a laser point cloud navigation map according to claim 1, characterized in that: The cloud platform is equipped with a distributed computing framework, which decomposes the map update task into multiple sub-tasks and distributes them to different computing nodes for simultaneous computation.

Images