A laser point cloud map construction and retrieval system and method

Abstract

This invention proposes a laser point cloud map construction and retrieval system and method, including a client and a server. The server includes a point cloud map construction module, a point cloud map retrieval module, an algorithm service module, a database, and a runtime environment. To address potential regional overlap during multiple point cloud map constructions, a unified standard map fusion method is designed. In the algorithm-level structural design, an adapter design pattern is adopted to re-encapsulate the basic functions of ROS. Simultaneously, an extensible algorithm executor module is designed and written to implement a registration mechanism for algorithm processing logic and dynamic matching during algorithm calls. This invention improves the efficiency of laser point cloud map construction and point cloud data processing. During point cloud map construction, it solves the problem of point cloud map elevation errors due to overlapping mapping areas. The convenient point cloud map data construction and retrieval service provides fundamental support for subsequent point cloud map data analysis and applications.

Description

Technical Field

[0001] This invention relates to the field of point cloud map construction and retrieval technology, and in particular to a laser point cloud map construction and retrieval system and method. Background Technology

[0002] With the rapid development of technologies such as autonomous vehicles, robots, drones, and virtual reality, point cloud maps, as a crucial infrastructure related to navigation, will continue to be widely used and promoted, especially in industrial, military, and medical fields. In recent years, relatively mature research has been conducted on point cloud map construction algorithms, with corresponding research directions mainly including vision-based scene map construction and LiDAR-based point cloud map construction. LiDAR point cloud maps primarily obtain point cloud data containing information such as distance and orientation relative to the environment directly from sensors using LiDAR, unaffected by many environmental factors (such as lighting). Furthermore, with the emergence of long-range, high-resolution 3D LiDARs such as the Velodyne VLS-128 and Ouster OS1-128, directly capturing precise details of the environment in 3D space using LiDAR has become more suitable. In addition, LiDAR can provide high-frequency range measurements and keep errors relatively stable within a small range, without needing to consider the measurement distance. Therefore, research on LiDAR-based state estimation and real-time map construction methods has become increasingly popular.

[0003] In point cloud map construction algorithms, the LIO-SAM mapping algorithm, proposed by Shan T in 2020, is a highly representative LiDAR mapping method based on a tightly coupled LiDAR-inertial odometry approach. This algorithm obtains the globally unified pose of the LiDAR scan by optimizing modules including the LiDAR odometry module, IMU pre-integration module, GPS factor, and loop closure detection factor. This achieves highly accurate and real-time trajectory estimation and map construction for mobile robots, making it a fundamental algorithm for default LiDAR point cloud map construction.

[0004] However, the aforementioned research on point cloud map construction remains at the algorithmic level. Software for point cloud map construction and retrieval using such algorithms is relatively scarce, and existing research results are managed through closed-source and commercial models, making them inaccessible. Commercially available point cloud processing software, such as CloudCompare, QGIS, and FugroViewer, only provide basic point cloud processing operations (e.g., creation, viewing, polishing, and auxiliary data analysis) for one or more point cloud files, and do not possess laser point cloud map construction capabilities. Among open-source software related to point cloud map construction, Autoware is highly representative. It is an open-source autonomous driving software stack that includes essential functions from localization and target detection to path planning and control. However, this also results in high coupling between modules, complex configuration operations, and many algorithmic logics, including point cloud map construction, are embedded, making it difficult to extract individual modules for independent research, such as mapping or positioning / navigation. Furthermore, it cannot effectively and reliably store and retrieve already constructed point cloud map data. In view of this, this invention combines mature point cloud map construction algorithms to develop efficient methods and software systems for construction, storage, and retrieval. Specifically, it features pluggable adaptation to point cloud map construction algorithms and provides efficient storage and retrieval functions for point cloud map construction data, offering a convenient application software platform solution for downstream research tasks such as localization, matching, and navigation. Summary of the Invention

[0005] To address the above problems, this invention provides a laser point cloud map construction and retrieval system, comprising:

[0006] Client and server;

[0007] The server-side includes: business layer, computing layer, database, and runtime environment;

[0008] The business layer includes: point cloud map construction module and point cloud map retrieval module;

[0009] The computation layer includes: an algorithm service module;

[0010] The client is configured to initiate operation commands and display information;

[0011] The server is configured to receive operation commands from the client and respond to and process those commands.

[0012] The business layer is connected to the client and is used to receive and respond to the client's operation instructions; the computing layer is connected to the business layer and the database. The business layer calls the algorithm service module of the computing layer, and the algorithm service module performs point cloud map construction and map data processing.

[0013] The point cloud map building module is configured to call the database and algorithm service modules to perform point cloud map building operations when the server receives a point cloud map building instruction.

[0014] The point cloud map retrieval module is configured to call the database and algorithm service modules to perform point cloud map retrieval operations when the server receives a point cloud map retrieval command.

[0015] The algorithm service module is configured to provide the algorithms called by the service client;

[0016] The database is configured to store point cloud map data and point cloud map construction information;

[0017] The runtime environment provides the necessary runtime environment for both the client and server.

[0018] Furthermore, the system is implemented using a client / server architecture; the client includes a front-end user interface and a presentation layer; the front-end user interface is developed using graphical user interface applications and icon processing technology, and the presentation layer uses asynchronous requests to render the front-end user interface.

[0019] Furthermore,

[0020] MinIO, an open-source object storage suite based on GoLang, stores point cloud map data and introduces a distributed cluster extension HOST to it.

[0021] Furthermore,

[0022] The databases include MySQL, MongoDB, Redis, and OSS.

[0023] The MySQL database stores relational data of the point cloud map data, the MongoDB database stores the trajectory point information of the point cloud map and configures the spatial index of MongoDB, the Redis database is used to cache the point cloud map data that has been successfully retrieved, and the OSS database implements object cloud storage for the constructed point cloud map data.

[0024] The trajectory point information includes: the three-dimensional coordinate information of the trajectory point, the latitude and longitude information corresponding to the trajectory point, and the mapping record ID information corresponding to each trajectory point.

[0025] Furthermore,

[0026] The algorithm service module includes algorithm invocation, algorithm operation interface, point cloud map construction algorithm, point cloud map retrieval algorithm, and global standardization algorithm for point cloud maps;

[0027] Furthermore, the algorithm invocation process is as follows:

[0028] When designing the algorithm's structure, the adapter design pattern is adopted to encapsulate the basic functions of ROS. At the same time, an extensible algorithm executor module is designed and written to implement the registration mechanism for algorithm processing logic and dynamic matching during algorithm invocation. This achieves the standard of non-intrusive and low coupling for algorithm modification or upgrade. When an algorithm request is initiated, matching information is obtained from the pre-registered algorithms based on the selected algorithm. It is then determined whether the algorithm exists. If it exists, the algorithm is started; otherwise, a map building failure response is returned. If the algorithm starts successfully, the algorithm call is executed. During the execution process, the algorithm operation interface is matched according to the request parameters. The algorithm operation interface starts running the algorithm and returns the result.

[0029] Furthermore, the process of the point cloud map construction algorithm is as follows:

[0030] Laser point cloud data is acquired through LiDAR. Motion distortion correction is performed on the laser point cloud data based on IMU data. After downsampling and feature extraction, feature point clouds are generated. Inter-frame matching of the feature point clouds is performed to optimize pose accuracy and initially construct a point cloud map. Global navigation satellite system data is extracted, and loop closure information generated by continuous detection of arriving scenes is used to construct a factor map and perform factor map optimization. The pre-integration method is used to reduce IMU error and improve the accuracy of subsequent point cloud motion distortion correction. A new factor map is constructed based on the pre-integration results to optimize and generate the final point cloud map.

[0031] Furthermore, the process of the point cloud map retrieval algorithm is as follows:

[0032] The client initiates a search request by inputting GPS latitude and longitude information. After receiving the request, the server performs a trajectory point search, using the spatial index range query feature of the database to retrieve trajectory points that meet the conditions. The trajectory points are sorted by distance, and the nearest trajectory point is selected as the neighboring point. The neighboring point search results are verified. If there is a trajectory point close to the current GPS point, the specific mapping record ID is located based on the additional information of the neighboring point. Then, the mapping record information corresponding to the neighboring point is retrieved from the database. Based on the mapping record information, the corresponding point cloud map information is retrieved from the database, and the point cloud information is loaded from the stored target location for use.

[0033] Furthermore, the global standardized execution process for point cloud maps is as follows:

[0034] The client sends a map building request, and the server calls the map building logic operation. When it finds that the current map building area overlaps with the existing map building area, it extracts multi-angle feature descriptors from the point cloud map data built by the map building algorithm, and combines them with the descriptors corresponding to the point cloud map information retrieved from the database with the current location GPS information to estimate the correlation between them, obtain the transformation matrix, and transform the input point cloud map according to the transformation matrix to complete the global standardization of the point cloud map. The output point cloud is then stored in the database.

[0035] This invention also proposes a method for constructing and retrieving laser point cloud maps, based on the above-described system implementation.

[0036] Use the client to initiate operation commands and display information;

[0037] The server receives operation instructions from the client and responds to and processes these instructions.

[0038] The server-side business layer receives and responds to the client's operation commands, and calls the algorithm service module of the computing layer, which in turn calls the data from the database.

[0039] When the server receives a point cloud map construction instruction, it uses the server's point cloud map construction module to call the server's database and the server's algorithm service module to perform point cloud map construction operations.

[0040] When the server receives a point cloud map retrieval command, it uses the server's point cloud map retrieval module to call the server's database and the server's algorithm service module to perform point cloud map retrieval operations.

[0041] The algorithm service module provides the algorithms called by the service client.

[0042] Use a database to store point cloud map data and point cloud map construction information;

[0043] It also provides the runtime environment required for both the client and server.

[0044] The beneficial effects of the technical solution provided by this invention are:

[0045] This invention proposes a laser point cloud map construction and retrieval system and method. The system includes a client and a server. The server includes a point cloud map construction module, a point cloud map retrieval module, an algorithm service module, a database, and a runtime environment. To address the potential for overlapping areas during multiple point cloud map constructions, a unified standard map fusion operation is designed. At the algorithm level, the adapter design pattern is used to encapsulate the basic functions of ROS. An extensible algorithm executor module is designed and written to implement a registration mechanism for algorithm processing logic and dynamic matching of algorithm calls. This invention improves the efficiency of laser point cloud map construction and laser point cloud data processing. During point cloud map construction, global point cloud map standardization solves the problem of point cloud map elevation errors when mapping areas overlap. Furthermore, convenient point cloud map information and data storage and retrieval services provide fundamental support for subsequent point cloud map data analysis and applications. Attached Figure Description

[0046] Figure 1 This is a block diagram of the laser point cloud map construction and retrieval system according to an embodiment of the present invention;

[0047] Figure 2 This is an architecture diagram of an OSS system for laser point cloud data storage according to an embodiment of the present invention;

[0048] Figure 3 This is a flowchart illustrating the algorithm matching call process in this embodiment;

[0049] Figure 4 This is a block diagram illustrating the execution of the laser point cloud map construction process according to an embodiment of the present invention;

[0050] Figure 5 This is a flowchart illustrating the global standardized process execution of a laser point cloud map according to an embodiment of the present invention.

[0051] Figure 6 This is a block diagram illustrating the execution of the laser point cloud map retrieval process according to an embodiment of the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0053] The block diagram of the laser point cloud map construction and retrieval system in this embodiment is as follows: Figure 1 The system is implemented using a client / server architecture, specifically including:

[0054] Client and server;

[0055] The client includes a front-end user interface and a presentation layer. The front-end user interface is developed using graphical user interface applications and icon processing technology. In this embodiment, it is developed using the cross-platform C++ graphical user interface application development framework Qt combined with image icon processing technology. The presentation layer uses asynchronous requests to render the front-end user interface. In this embodiment, the presentation layer is based on ROS (Robot Operating System).

[0056] The server-side includes: business layer, computing layer, database, and runtime environment;

[0057] The business layer includes: point cloud map construction module and point cloud map retrieval module;

[0058] The computation layer includes: an algorithm service module;

[0059] The databases include: MySQL, MongoDB, Redis, and OSS.

[0060] The algorithm service module includes algorithm invocation, algorithm operation interface, point cloud map construction algorithm, point cloud map retrieval algorithm, and global standardized execution process for point cloud maps;

[0061] The system utilizes a MySQL database to store relational data of the point cloud map, a MongoDB database to store trajectory point information, and a MongoDB spatial index. This MongoDB spatial index allows the GPS latitude and longitude information of the trajectory points to be converted into virtual geographic location information, enabling GPS-based queries of nearby point cloud map information. Furthermore, leveraging the characteristics of its spatial index, a GPS latitude and longitude-based spatial index related to the trajectory points is constructed. This step aims to improve the query efficiency of GPS-based nearby point cloud map information. A Redis database is used to cache successfully retrieved point cloud map data, and an OSS database provides object cloud storage for the constructed point cloud map data.

[0062] The trajectory point information includes: the three-dimensional coordinates of the trajectory points, the latitude and longitude information corresponding to the trajectory points, and the ID information corresponding to each trajectory point. Before loading the point cloud map, the trajectory point information must first be loaded. The point cloud map information is searched by maintaining an association index between each trajectory point and the mapping information on the trajectory point information, and then the corresponding point cloud map data is loaded from the physical storage medium.

[0063] The client is configured to initiate operation commands and display information;

[0064] The server is configured to receive operation commands from the client and respond to and process those commands.

[0065] The business layer is connected to the client and is used to receive and respond to the client's operation instructions; the computing layer is connected to the business layer and the database. The business layer calls the algorithm service module of the computing layer, and the algorithm service module performs point cloud map construction and map data processing.

[0066] The point cloud map building module is configured to call the database and algorithm service modules to perform point cloud map building operations when the server receives a point cloud map building instruction.

[0067] The point cloud map retrieval module is configured to call the database and algorithm service modules to perform point cloud map retrieval operations when the server receives a point cloud map retrieval command.

[0068] The algorithm service module is configured to provide the algorithms called by the service client;

[0069] The database is configured to store point cloud map data and point cloud map construction information;

[0070] The runtime environment provides the necessary runtime environment for both the client and server.

[0071] The architecture diagram of the laser point cloud data object storage (OSS) system in this embodiment is referenced. Figure 2 :

[0072] This invention employs a distributed data storage scheme, treating all data as objects. Each object contains data, metadata, and a unique identifier. To ensure the reliability and stability of point cloud map data storage and to accommodate potential data migration or sharing, this invention selects the open-source object storage suite MinIO based on GoLang to store the actual point cloud map data files, and introduces a distributed cluster extension HOST. The cluster can consist of multiple MinIO instances and their corresponding disks. When a client requests access to point cloud map data, the object storage service first routes the request to a load balancing unit. This unit is responsible for load balancing based on user activity before reaching the specific MinIO instance, thus balancing the request pressure on each MinIO instance.

[0073] The flowchart of the algorithm matching call process in this embodiment is shown in the reference diagram. Figure 3 :

[0074] The diagram illustrates the algorithm access and invocation process design. Considering the possibility of future changes to the mapping algorithm, the software code structure employs adapter design patterns and references the eight principles of object-oriented programming and RESTful interface development specifications. The basic functions of the Robot Operating System (ROS) are re-encapsulated, and an extensible algorithm executor module is designed and written to implement a registration mechanism for algorithm processing logic and dynamic matching of algorithm calls. This achieves a non-intrusive and low-coupling standard for algorithm modification or upgrades, significantly reducing the overhead of incremental development and subsequent modifications and expansions. The point cloud map retrieval logic is also encapsulated into dynamic and static libraries to reduce call coupling. When a mapping request is initiated, matching information is retrieved from pre-registered algorithms based on the selected algorithm. The system then checks if the algorithm exists. If it does, the algorithm service node is started; otherwise, a mapping failure response is returned. If the algorithm starts successfully, the algorithm call is executed. During execution, the service matches the algorithm operation object based on the request parameters. The algorithm operation object is the concrete implementation of the algorithm operation interface. Once everything is ready, the algorithm operation object starts running and returns the result, indicating a successful request response.

[0075] refer to Figure 4 , Figure 4This is a flowchart illustrating the laser point cloud map construction process of this invention. Laser point cloud data is acquired via LiDAR. When constructing the point cloud map, the algorithm first corrects motion distortion of the acquired laser point cloud data based on information from the IMU sensor, and then performs downsampling and feature extraction operations as needed. The extracted point cloud data serves as input for backend processing. Next, the algorithm stores the current pose and point cloud data as keyframes in a sliding window in memory, and finds historical keyframes with the same feature structure through inter-frame matching. Then, it calculates the pose of the current frame relative to historical frames using LidarOdometry to optimize the pose accuracy of the current frame. Subsequently, the algorithm uses the ICP algorithm to calculate the odometry increment based on the globally estimated pose, thereby further optimizing the pose estimation. Following this, the algorithm uses the previously processed pose and laser point cloud data to perform a forward and backward search to obtain the optimal solution. Then, a nonlinear optimization method is used to optimize the pose and feature points to improve accuracy. The algorithm includes a loop closure detection module at the end. By comparing odometry information with point cloud features, it detects whether there are loops between the current frame and all previous frames. It extracts global navigation satellite system data, continuously detects whether loop information generated by repeated arrival scenes is generated, constructs a factor map, and performs factor map optimization. It uses a pre-integration method to reduce IMU error, improves the accuracy of subsequent point cloud motion distortion correction, and constructs a new factor map based on the pre-integration results to correct the map, eliminate cumulative errors, obtain the final optimized path trajectory and map data, and complete the entire algorithm execution process.

[0076] The flowchart for the global standardization process of the laser point cloud map in this embodiment is shown below. Figure 5 :

[0077] Multiple mappings in overlapping areas may result in inconsistent elevation information due to errors in LiDAR data acquisition, especially elevation errors. The global standardization execution process for point cloud maps is designed to address the issue of overlapping areas that may occur when building multiple point cloud maps.

[0078] The client sends a map building request, and the server invokes the mapping logic to extract multi-angle feature (planar point cloud features and spherical projection point cloud features) descriptors from the point cloud map data (current temporary point cloud information and existing point cloud information at the current location) constructed by the mapping algorithm. These descriptors are then combined with the descriptors corresponding to the point cloud map information retrieved from the database using the current location's GPS information to estimate the correlation between them, deriving a transformation matrix. The input point cloud map is then transformed according to this transformation matrix, thus completing the global standardization of the point cloud map. During the global standardization process, two additional logic modules are introduced for viewpoint-invariant scene recognition and adaptive loop closure detection. After point cloud map fusion, a global optimization module further optimizes the point cloud data. Finally, the output point cloud is persisted to the database.

[0079] The flowchart of the laser point cloud map retrieval process in this embodiment is shown below. Figure 6 :

[0080] In this embodiment of the invention, MongoDB, a non-relational database, is used to store point cloud map trajectory information, and its spatial index is used to retrieve nearby map data. A MySQL relational database is used to index point cloud map information, including construction records and necessary point cloud map data. The client initiates a search request by inputting GPS latitude and longitude information. Upon receiving the request, the server performs a nearby (trajectory point) search in MongoDB. The system first compares this information with the index established during the mapping and storage of trajectory points. Simultaneously, it utilizes the range query feature of MongoDB's spatial index to retrieve one or more trajectory points that meet the criteria. The system sorts the trajectory points by distance and extracts the closest trajectory point, i.e., the nearest point. The nearest point search results are verified. If a trajectory point close to the current GPS point exists, the system locates the specific mapping record ID based on the additional information of that nearest point. It then retrieves the mapping record information corresponding to the nearest (trajectory) point from the MySQL database. Based on the mapping record information, the system retrieves the corresponding point cloud map information from the MySQL database. After obtaining the point cloud data storage information, the system loads the point cloud information from the storage target location for use.

[0081] A further embodiment proposes a laser point cloud map construction and retrieval method, based on the above system implementation, including:

[0082] Use the client to initiate operation commands and display information;

[0083] The server receives operation instructions from the client and responds to and processes these instructions.

[0084] The server-side business layer receives and responds to the client's operation commands and calls the algorithm service module of the computing layer. The server-side computing layer then calls data from the database.

[0085] When the server receives a point cloud map construction instruction, it uses the server's point cloud map construction module to call the server's database and the server's algorithm service module to perform point cloud map construction operations.

[0086] When the server receives a point cloud map retrieval command, it uses the server's point cloud map retrieval module to call the server's database and the server's algorithm service module to perform point cloud map retrieval operations.

[0087] The algorithm service module provides the algorithms called by the service client.

[0088] Use a database to store point cloud map data and point cloud map construction information;

[0089] It also provides the runtime environment required for both the client and server.

[0090] This invention provides a convenient application software platform solution for point cloud map construction and downstream research tasks such as localization, matching, and navigation. It improves the efficiency of laser point cloud map construction and laser point cloud data processing. Furthermore, by standardizing the global point cloud map, it solves the problem of point cloud map elevation errors when mapping areas overlap. Simultaneously, the convenient storage and retrieval services for point cloud map information and data provide fundamental support for subsequent point cloud map data analysis and applications.

[0091] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A laser point cloud map construction and retrieval system, characterized in that, Comprise: Client and server; Server includes: business layer, operation layer, database, running environment; Business layer includes: point cloud map construction module, point cloud map retrieval module; Operation layer includes: algorithm service module; Client is configured for the initiation of operating instructions and information display; Server is configured for receiving operating instructions from the client, and responding to and processing the operating instructions; Business layer is connected with the client for receiving and responding to the operating instructions of the client; operation layer is connected with business layer and database, business layer calls algorithm service module of operation layer, and algorithm service module executes point cloud map construction and map data processing; Point cloud map construction module is configured to call database and algorithm service module for point cloud map construction operation when server receives point cloud map construction instruction; Point cloud map retrieval module is configured to call database and algorithm service module for point cloud map retrieval operation when server receives point cloud map retrieval instruction; Algorithm service module is configured to provide algorithm called by server; Database is configured to store point cloud map data and point cloud map construction information; Running environment provides the running environment required by client and server; Algorithm service module includes algorithm call, algorithm operation interface, point cloud map construction algorithm, point cloud map retrieval algorithm, and global standardization algorithm of point cloud map; The laser point cloud map construction process is as follows: According to the information in the IMU sensor, the acquired laser point cloud data is corrected for motion distortion, and the extracted point cloud is used as the input of the back-end processing. The pose of the current time and the point cloud data are stored as key frames in the sliding window in the memory, and the historical key frames are found through inter-frame matching. The pose of the current frame relative to the historical frame is calculated, the pose accuracy of the current frame is optimized, the ICP algorithm is used to calculate the odometer increment based on the pose, the pose and point cloud are optimized based on the nonlinear optimization method, the odometer information and the point cloud are compared to detect whether there is a loop between the current frame and all previous frames, a factor graph is constructed based on the global navigation satellite system data and the loop information generated by continuously detecting whether the scene is repeated, and factor graph optimization is performed. The pre-integration method is used to reduce the IMU error, and a new factor graph is constructed according to the pre-integration result to correct the map and obtain the final optimized path trajectory and map data. 2.The laser point cloud map construction and retrieval system of claim 1, wherein, The system is realized by using C / S architecture.

3. The laser point cloud map construction and retrieval system according to claim 1, wherein The point cloud map data is stored by using the open source object storage kit MinIO based on GoLang, and the HOST is introduced for distributed cluster expansion.

4. The laser point cloud map construction and retrieval system according to claim 1, wherein The database includes MySQL database, MongoDB database, Redis database and OSS database. The MySQL database stores point cloud map data in relational data, the MongoDB database stores track point information of the point cloud map, a spatial index of the MongoDB is configured, the Redis database is used for caching successfully retrieved point cloud map data, and the OSS database implements object cloud storage on the constructed point cloud map data. The track point information includes three-dimensional coordinate information of the track point, longitude and latitude information corresponding to the track point, and mapping record id information corresponding to each track point.

5. The laser point cloud map construction and retrieval system of claim 1, wherein When an algorithm call of a point cloud map construction algorithm, a point cloud map retrieval algorithm, and a global standardization algorithm of the point cloud map is performed, the algorithm is run through an algorithm operation interface and a result is returned.

6. The system of claim 5, wherein, The algorithm call process is as follows: When the algorithm is designed in structure, an adapter design pattern is used to perform secondary encapsulation on basic functions of ROS, and an extensible algorithm executor module is designed and written to realize a registration mechanism of algorithm processing logic and dynamic matching when the algorithm is called, so that a non-invasive and low-coupling standard of algorithm modification or upgrade is achieved. When an algorithm request is initiated, matching information is obtained from the pre-registered algorithm according to the selected algorithm, it is judged whether the algorithm exists, if the algorithm exists, the algorithm is started, otherwise, response information of a mapping failure is returned. If the algorithm is successfully started, the algorithm call is executed. In the execution process, the algorithm operation interface is matched according to the request parameters, the algorithm operation interface starts to run the algorithm and returns the result.

7. The system of claim 5, wherein, The process of the point cloud map retrieval algorithm is as follows: The client initiates a retrieval request by inputting GPS longitude and latitude information. After the server receives the request, track point retrieval is performed. The track point that meets the condition is retrieved by using the spatial index range query feature in the database. The track points are sorted according to the distance and the nearest track point is selected as the adjacent point. The adjacent point retrieval result is verified. If there is a track point close to the current GPS point, the specific mapping record id is located according to the additional information of the adjacent point. Then the mapping record information corresponding to the adjacent point is obtained from the database. The point cloud map information corresponding to the mapping record information is retrieved in the database. The point cloud information is loaded from the storage target position for use. 8.The system of claim 5, wherein, The execution process of the global standardization of the point cloud map is as follows: The client sends a map construction request. The server calls the logic operation of mapping. When it is found that the current mapping area overlaps with the existing mapping area, the point cloud map data constructed by the mapping algorithm is subjected to multi-angle feature descriptor extraction. The descriptors corresponding to the point cloud map information retrieved from the database according to the current position GPS information are combined to estimate the correlation between each other to obtain a transformation matrix. The input point cloud map is converted according to the transformation matrix to complete the global standardization of the point cloud map. The output point cloud is stored in the database.

9. A laser point cloud map construction and retrieval method, implemented based on the system of any one of claims 1-8, wherein An operation instruction is initiated and information is displayed using the client. The service end receives operation instructions from the client end and responds to and processes the operation instructions; The business layer of the service end receives and responds to operation instructions from the client end and calls the algorithm service module of the operation layer, which calls the data of the database; When the service end receives a point cloud map construction instruction, the point cloud map construction module of the service end calls the database of the service end and the algorithm service module of the service end to perform a point cloud map construction operation; When the service end receives a point cloud map retrieval instruction, the point cloud map retrieval module of the service end calls the database of the service end and the algorithm service module of the service end to perform a point cloud map retrieval operation; The algorithm service module provides algorithms called by the service end; The database stores point cloud map data and point cloud map construction information; And provides the required operating environment for the client end and the service end.

Images