A method, system, electronic device, and storage medium for managing multi-source spatial data of high-precision maps.

Abstract

This application discloses a method, system, electronic device, and storage medium for managing multi-source spatial data of high-precision maps. The method includes converting spatial data from different data platforms into metadata; converting the metadata into a world coordinate system; removing metadata that does not meet preset conditions for coverage, positioning signal quality, and data continuity; converting the remaining metadata into a metadata file with a fixed directory structure; and generating a corresponding unique identifier for the entire task within the metadata file. This application is used for unified management of multi-source spatial data.

Description

Technical Field

[0001] This application belongs to the technical field of spatial data processing, specifically relating to a method, system, electronic device and storage medium for managing multi-source spatial data of high-precision maps. Background Technology

[0002] High-precision maps, real-scene 3D models, digital twin base maps, and other spatial digitization products are the core infrastructure for autonomous driving, eVTOL drones, vehicle-road-cloud collaboration, high-speed digitization, and intelligent surveying and mapping. The corresponding data production chain usually includes the following steps: field data collection, point cloud preprocessing, feature extraction and manual editing, quality inspection, and format compilation and output.

[0003] In existing technologies, different data production scenarios often rely on different platforms, and these platforms in the data production chain generally suffer from the following shortcomings: First, the use of heterogeneous multi-platform data at the laser point cloud acquisition end makes it impossible to process data uniformly at the back end. Spatial digital production involves multiple acquisition scenarios, such as urban roads (vehicle-mounted LiDAR), low-altitude mapping (UAV-mounted LiDAR), indoor parks (handheld devices), and underground pipelines (robot-mounted sensors). The data formats, coordinate systems, and sensor extrinsic parameters output by different platforms are all different. The existing data production chain requires the separate development of data access pipelines for different platforms. However, due to the lack of standardized metadata declaration mechanisms and automatic extrinsic parameter correction methods, it is impossible to uniformly schedule data in heterogeneous multi-platform mixed scenarios, such as integrated vehicle-mounted and UAV projects. Second, in the data preprocessing stage, single-frame point cloud (passed in one acquisition) is often used as the basic data segment for processing, ignoring the fact that the same area is usually scanned multiple times in actual acquisition. When using single-frame point cloud to project onto the map, it is easy to miss detection and noise, which makes the single-frame method unable to meet the usable standards for high-precision map production in actual production. Third, the cold start cost of algorithms in the element extraction end is high; although deep learning methods have high accuracy and can meet the needs of high-progress maps, they often require hundreds of kilometers of labeled data to achieve practical accuracy, making them completely unusable when new businesses are cold-started; rule-based algorithms can run with zero labels, but their accuracy is limited and difficult to improve. Fourth, the quality control relies on manual review, but the standards for manual review are subjective and the allocation of resources is not differentiated. This can easily lead to different confidence levels of results consuming the same amount of manual time, different people having completely different conclusions, and quality problems being discovered only before delivery, which increases rework costs. Fifth, there is no unified task tracking method in the entire data chain process, and it does not support data lineage tracing and map version optimization iteration. This results in the inability to trace data, the inability to objectively quantify the accuracy improvement of the new version map relative to the old version map after algorithm upgrade, the inability to recycle manual correction results into training data, and the difficulty of rolling back high-precision maps without geometric version control. Summary of the Invention

[0004] To address one or more of the aforementioned problems, the purpose of this application is to provide a method, system, electronic device, and storage medium for managing multi-source spatial data of high-precision maps, so as to uniformly manage multi-source spatial data.

[0005] The technical solution of this application is as follows.

[0006] The first aspect is a method for managing multi-source spatial data of high-precision maps, including: Spatial data from different data platforms are converted into metadata containing terminal type identifiers and sensor extrinsic parameters; Based on the terminal type identifier and sensor extrinsic parameters, the metadata is converted to the world coordinate system; Remove metadata that does not meet one or more of the preset conditions, such as coverage rate, positioning signal quality, and data continuity. The remaining metadata after removal is decoupled from different data platforms through sensor extrinsic calibration and converted into metadata files with a fixed directory structure; Generate a unique identifier for the corresponding end-to-end task in the metadata file.

[0007] As one implementation of the first aspect, it also includes: The metadata file, which contains a unique identifier for the entire task, is input into the high-precision map platform to perform the data feature extraction task, thereby forming a map data file.

[0008] Preferably, the metadata file includes a genus-based subfile in JSON format for genus tracing. After the high-precision map platform executes each sub-step of the data feature extraction task, it synchronously adds the processing records of each sub-step of the task to the lineage sub-file.

[0009] Furthermore, the completion status of the sub-steps of the data element extraction task is recorded in the lineage sub-file; If a task is interrupted or terminated abnormally, the data feature extraction task will restart from the unfinished sub-steps based on their completion status.

[0010] Furthermore, using the unique task identifier across the entire chain as the root node, after each sub-step of the data element extraction task is completed, one or more of the following information is added to the root node within the lineage sub-file: the corresponding processing tool identifier, algorithm version, execution parameters, output file path, and quality indicators for each sub-step.

[0011] Preferably, when the processing result of the data feature extraction task needs to be modified, the geometric difference between the corresponding data features in the map data file and the old version of the high-precision map is used as the processing record, and the version submission identifier is written into the lineage sub-file. The Hausdorff algorithm was used to calculate the geometric difference.

[0012] Furthermore, when modifying the processing results of the data feature extraction task, the geometric difference of the corresponding data features is used as a weighted training sample. When the amount of training sample data is greater than or equal to the first preset threshold, the training sample is used to train the deep learning segmentation model. When the amount of training sample data is greater than or equal to the second preset threshold, the training sample is used to train the end-to-end vectorized model. The deep learning segmentation model adopts a multimodal fusion model, and the end-to-end vectorization model adopts the CLRNet model or the U-Net model.

[0013] Secondly, a multi-source spatial data management system for high-precision maps includes: The spatial data receiving module is used to convert spatial data from different data platforms into metadata that includes terminal type identifiers and sensor extrinsic parameters. The coordinate transformation module is used to transform metadata into the world coordinate system based on the terminal type identifier and sensor extrinsic parameters. The data quality control module is used to remove metadata that does not meet one or more of the preset conditions, such as coverage, positioning signal quality, and data continuity. The metadata file generation module is used to decouple the remaining metadata after removal from different data platforms through sensor extrinsic calibration and convert it into a metadata file with a fixed directory structure. The task identifier generation module is used to generate a unique identifier for the corresponding end-to-end task in the metadata file.

[0014] Thirdly, an electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the multi-source spatial data management method for the high-precision map of any of the preceding claims.

[0015] Fourthly, a storage medium is a computer-readable storage medium storing computer instructions for enabling a computer to implement the multi-source spatial data management method for high-precision maps as described above.

[0016] Compared with the prior art, the advantages of this application are as follows: It adapts to various spatial digital production scenarios involving multiple data acquisition scenarios, avoiding the problem of inconsistent spatial data scheduling in heterogeneous multi-platform mixed scenarios; Data quality control reduces reliance on manual review, avoiding the drawbacks of different people spending the same amount of time, having drastically different conclusions, and often discovering quality problems only before delivery, thus increasing rework costs. A unified task tracking method is set up for the entire process of using spatial data, supporting data lineage tracing and map version optimization and iteration; After tracing data and upgrading algorithms, the accuracy improvement of the new version of the map relative to the old version can be objectively quantified, the modification results can be fed back as training data, and high-precision maps can achieve geometric-level data version updates and rollback control. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating a multi-source spatial data management method for high-precision maps according to this application.

[0018] Figure 2 This is a schematic diagram of the structural framework of a multi-source spatial data management system for high-precision maps according to this application.

[0019] Figure 3 This is a schematic diagram of the structural framework of an electronic device according to this application. Detailed Implementation

[0020] Referring to the illustrations, the principles of this application are illustrated by way of example implementation in a suitable operating environment. The following description is based on the specific embodiments of this application as illustrated, and should not be construed as limiting other specific embodiments not detailed herein, such as corresponding adjustments that may be made to the execution order of the steps of the method of this application while remaining consistent with the inventive concept.

[0021] like Figure 1 As shown in the figure, the multi-source spatial data management method for high-precision maps in this application includes the following steps.

[0022] S1. Convert all heterogeneous spatial data from different data platforms into metadata containing terminal type identifiers and sensor extrinsic parameters.

[0023] In step S1, after the data collection task is completed, different data platforms package the corresponding spatial data into a manifest.json file. The manifest.json file sets a task ID field for the data collection task, which is used to generate a unique identifier for the entire task in the future. The manifest.json file sets a terminal type field for the terminal type used in the data collection task, and the sensor extrinsic parameters collected are represented in the form of an SE(3) matrix. Different data platforms may include handheld terminals, drones, robots, unmanned vehicles, etc. The spatial data is collected by the sensors mounted on each data platform. The types of spatial data include laser point clouds, motion parameters, etc.

[0024] S2. Based on the terminal type identifier and sensor extrinsic parameters, convert the metadata to the world coordinate system.

[0025] In step S2, the terminal type identifier and sensor extrinsic parameters are extracted from the manifest.json file of the spatial data from step S1.

[0026] In other embodiments, metadata can also be converted to the world coordinate system for pose alignment.

[0027] S3. Remove metadata that does not meet one or more of the preset conditions, such as coverage rate, positioning signal quality, and data continuity.

[0028] In this embodiment, after removing the relevant metadata, the corresponding data platform can be notified to re-collect the corresponding spatial data.

[0029] In step S3, the preset coverage rate can be the ground coverage rate of the laser point cloud data, the positioning signal quality can be based on the number of available GNSS satellites, the data continuity can be an indicator of whether there are jump data frames in the collected IMU sensor data, and the data continuity can also be an indicator of whether the data volume interval between multiple frames of laser point cloud data is too dense or too sparse.

[0030] S4. The remaining metadata after removal is decoupled from different data platforms through sensor extrinsic calibration and converted into metadata files with a fixed directory structure.

[0031] In step S4, the fixed directory structure of the metadata file includes the collected spatial data sub-file, the processing data sub-file corresponding to each sub-step required by the subsequent high-precision map platform processing, and the lineage sub-file, etc.

[0032] Fixed directory structure refers to a type of file structure where the folder hierarchy, file storage paths, and naming rules are forcibly hardcoded within a file / compressed package / image.

[0033] S5. Generate the corresponding unique identifier for the end-to-end task in the metadata file and write the unique identifier for the end-to-end task into the lineage sub-file.

[0034] In step S5, the generation of the unique identifier for the end-to-end task mainly relies on the task ID field in the manifest.json file. The unique identifier for the end-to-end task can be set as a string containing task ID information. In other implementations, it can also be generated by combining other types of data based on the task ID field.

[0035] The unique identifier for the entire task can be generated in the lineage sub-file and used as the root node for lineage tracing, allowing data from different platforms to be traced throughout the entire task processing process on the high-precision map platform. In other implementations, all fields and data content of the manifest.json file can be converted into the sub-files of the metadata file for use.

[0036] S6. Input the metadata file with the unique identifier of the entire task into the high-precision map platform to perform the data feature extraction task and form a map data file.

[0037] In step S6, in addition to performing data feature extraction tasks, the high-precision map platform can also perform data feature learning, terrain and building model generation, geographic information association and matching, etc. The high-precision map platform is usually a cloud platform. For example, highway laser point cloud data collected from different data platforms are processed in steps S1-S5 to form a metadata file with a unique identifier for the entire task and sent to the high-precision map platform on the cloud platform. Then, the metadata file is input into the high-precision map platform, and ground segmentation, BEV projection, BEV intensity map generation, and ground marker line extraction are performed in sequence to form a map data file containing ground marker lines. The content of the processing in each sub-step is written into the metadata file. In particular, the processing process has a lineage sub-file and a unique identifier for the entire task, which enables data lineage tracing.

[0038] In step S6, the lineage sub-file is made in accordance with the ISO19115-1:2014 standard, which can support the metadata file to trace any and complete data usage history in the steps of being used by the high-precision map platform to process various tasks.

[0039] In this embodiment, the metadata file includes a JSON-formatted lineage sub-file for data lineage tracing; after the high-precision map platform executes each sub-step of the data feature extraction task, it synchronously adds the processing records of each sub-step of the task to the lineage sub-file, thereby tracing the relevant data and corresponding processing content.

[0040] In this preferred embodiment, the completion status of the sub-steps of the data element extraction task is recorded in the lineage sub-file; if the task is interrupted or ends abnormally, the data element extraction task is restarted from the unfinished sub-steps based on the completion status of the sub-steps.

[0041] In this preferred embodiment, the unique task identifier across the entire chain is used as the root node. After each sub-step of the data element extraction task is completed, one or more of the following are added to the root node of the lineage sub-file: the corresponding processing tool identifier, algorithm version, execution parameters, output file path, and quality indicators for each sub-step.

[0042] In this embodiment, when the processing result of the data feature extraction task needs to be modified, the geometric difference between the corresponding data features in the map data file and the old version of the high-precision map is used as the processing record, and the version submission identifier is written into the lineage sub-file.

[0043] In this preferred embodiment, when modifying the processing results of the data element extraction task, the geometric difference of the corresponding data elements is used as a weighted training sample. When the amount of training sample data is greater than or equal to a first preset threshold, the training sample is used to train a deep learning segmentation model to identify indicators such as the completeness or accuracy of data element extraction. When the amount of training sample data is greater than or equal to a second preset threshold, the training sample is used to train an end-to-end vectorization model to identify indicators such as the completeness or accuracy of data element extraction. In this further preferred embodiment, the training samples can be manually corrected before using the deep learning segmentation model and / or the end-to-end vectorization model. In this further preferred embodiment, the deep learning segmentation model adopts a multimodal fusion model architecture, the end-to-end vectorization model adopts a CLRNet model or a U-Net model, and the weight of the training sample can be set as the modification amount corresponding to the geometric difference or the confidence level of data element extraction.

[0044] In this embodiment, the Hausdorff algorithm can be used to calculate the geometric difference. The first preset threshold and the second preset threshold can be based on the scene of the collected data and the relevant parameters of the target object. For example, in the process of using vehicle-mounted laser to collect laser point cloud data and utilizing road surface marking lines, the length difference extracted before and after three ground marking lines is set as the segmentation point, and the corresponding segmentation point value is the preset threshold. The principle of the Hausdorff algorithm is: for each point in set P to the nearest point in set Q, and for each point in set Q to the nearest point in set P, the maximum value of the two is taken. Here, set P and set Q can be laser point cloud data from different frames.

[0045] Compared with the prior art, the beneficial effects of this embodiment are as follows: The core of this approach is to decouple metadata from different data platforms through sensor extrinsic calibration, converting it into metadata files with a fixed directory structure. Utilizing lineage sub-files and unique identifiers for end-to-end tasks, it achieves a traceable and iterative high-precision map data management system. It adapts to various spatial digitization production scenarios involving multiple data acquisition modes, avoiding the problem of existing technologies' inability to uniformly schedule spatial data in heterogeneous multi-platform mixed scenarios. Compared to existing manual review methods, its data quality control measures control data usage by calculating preset coverage, positioning signal quality, and data continuity, avoiding the drawbacks of different personnel consuming the same amount of time, potentially leading to drastically different conclusions, and quality issues often being discovered only before delivery, increasing rework costs. Throughout the entire spatial data usage process, it sets unique identifiers for end-to-end tasks and lineage sub-files as a unified task tracking method, supporting data lineage tracing and map version optimization iterations. After tracing data and upgrading algorithms, it can objectively quantify the accuracy improvement of the new map version relative to the old version, allowing modification results to flow back as training data, and enabling high-precision maps to achieve geometric-level data version updates and rollback control. Figure 2 As shown in the figure, this application discloses a multi-source spatial data management system for high-precision maps, which includes the following structure.

[0046] The spatial data receiving module is used to convert spatial data from different data platforms into metadata that includes terminal type identifiers and sensor extrinsic parameters.

[0047] The coordinate transformation module is used to transform metadata into the world coordinate system based on the terminal type identifier and sensor extrinsic parameters.

[0048] The data quality control module is used to remove metadata that does not meet one or more of the preset conditions, such as coverage, positioning signal quality, and data continuity.

[0049] The metadata file generation module is used to convert the remaining metadata after removal into metadata files with a fixed directory structure by decoupling it from different data platforms through sensor extrinsic calibration.

[0050] The task identifier generation module is used to generate a unique identifier for the corresponding end-to-end task in the metadata file and write the unique identifier for the end-to-end task into the lineage sub-file.

[0051] Optionally, this embodiment also includes a high-precision map platform, used to perform data feature extraction tasks on metadata files with unique identifiers for the entire task chain, to form map data files.

[0052] The system described above is used to implement the multi-source spatial data management method for the corresponding high-precision map in the foregoing embodiments, and has the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0053] like Figure 3 As shown, based on the same inventive concept, corresponding to any of the above embodiments, this application also discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the above-described method for managing multi-source spatial data of high-precision maps.

[0054] Specifically, the device includes: a processor 1010, a memory 1020, an input / output interface 1030, a communication interface 1040, and a bus 1050. The processor 1010, memory 1020, input / output interface 1030, and communication interface 1040 are interconnected within the device via the bus 1050.

[0055] The processor 1010 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), GPU (Graphics Processing Unit), or one or more integrated circuits, to implement relevant programs and achieve the technical solutions provided in the embodiments of this specification.

[0056] The memory 1020 can be implemented in the form of ROM (Read-Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1020 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 1020 and called by the processor 1010. The input / output interface 1030 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, projectors, speakers, vibrators, indicator lights, etc.

[0057] The communication interface 1040 is used to connect the communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB (Universal Serial Bus), network cable, etc.) or wireless means (such as mobile network, WIFI (Wireless Fidelity), Bluetooth, etc.).

[0058] Bus 1050 includes a pathway for transmitting information between various components of the device, such as processor 1010, memory 1020, input / output interface 1030, and communication interface 1040.

[0059] It should be noted that although the above-described device only shows the processor 1010, memory 1020, input / output interface 1030, communication interface 1040, and bus 1050, in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0060] The electronic devices described above are used to implement the multi-source spatial data management method for high-precision maps in any of the foregoing embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0061] Based on the same inventive concept, corresponding to any of the above embodiments, this application also discloses a non-transitory computer-readable storage medium that stores computer instructions for enabling a computer to implement the multi-source spatial data management method for high-precision maps as described above.

[0062] The computer-readable storage medium of this embodiment includes permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium, which can be used to store information accessible by a computing device. The computer instructions stored in the storage medium of the above embodiments are used to enable the computer to implement the multi-source spatial data management method for high-precision maps as described in any of the above embodiments, and have the beneficial effects of the corresponding method embodiments, which will not be repeated here.

[0063] The above description is merely a preferred embodiment and the technical principles employed in this application. This application is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions that can be made by those skilled in the art will not depart from the scope of protection of this application. Therefore, although this application has been described in detail through the above embodiments, this application is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of this application. The scope of this application is determined by the scope of the claims.

Claims

1. A multi-source spatial data management method for high-precision maps, characterized by, include: Spatial data from different data platforms are converted into metadata containing terminal type identifiers and sensor extrinsic parameters; Based on the terminal type identifier and the sensor extrinsic parameters, the metadata is converted to the world coordinate system; Remove metadata that does not meet one or more of the preset conditions, such as coverage rate, positioning signal quality, and data continuity. The remaining metadata after removal is decoupled from the different data platforms through sensor extrinsic calibration and converted into metadata files with a fixed directory structure; Generate a unique identifier for the corresponding end-to-end task in the metadata file.

2. The method for managing multi-source spatial data of high-precision maps according to claim 1, characterized in that, Also includes: The metadata file containing the unique identifier of the entire task is input into the high-precision map platform to perform the data feature extraction task, thereby forming a map data file.

3. The method for managing multi-source spatial data of high-precision maps according to claim 2, characterized in that, The metadata file includes a JSON-formatted lineage subfile for data lineage tracing; After the high-precision map platform executes each sub-step of the data feature extraction task, it synchronously adds the processing records of each sub-step of the task to the lineage sub-file.

4. The multi-source spatial data management method for high-precision maps according to claim 3, characterized in that, Record the completion status of sub-steps of the data feature extraction task in the lineage sub-file; If a task is interrupted or terminated abnormally, the data feature extraction task will be restarted from the unfinished sub-steps based on the completion status of the sub-steps.

5. The method for managing multi-source spatial data of high-precision maps according to claim 3, characterized in that, Using the unique task identifier across the entire chain as the root node, after each sub-step of the data element extraction task is completed, add one or more of the following to the root node in the lineage sub-file: the corresponding processing tool identifier, algorithm version, execution parameters, output file path, and quality indicators for each sub-step.

6. The multi-source spatial data management method for high-precision maps according to claim 2, characterized in that, When the processing result of the data feature extraction task needs to be modified, the geometric difference between the corresponding data features in the map data file and the old version of the high-precision map is used as the processing record, and the version submission identifier is written into the lineage sub-file. The geometric difference was calculated using the Hausdorff algorithm.

7. The multi-source spatial data management method of a high-precision map according to claim 6, characterized in that, When modifying the processing results of the data feature extraction task, the geometric difference of the corresponding data features is used as a weighted training sample. When the amount of data in the training samples is greater than or equal to the first preset threshold, the training samples are used to train the deep learning segmentation model. When the amount of training sample data is greater than or equal to the second preset threshold, the training sample is used to train the end-to-end vectorized model. The deep learning segmentation model adopts a multimodal fusion model, and the end-to-end vectorization model adopts either the CLRNet model or the U-Net model.

8. A multi-source spatial data management system for high-precision maps, characterized by, include: The spatial data receiving module is used to convert spatial data from different data platforms into metadata that includes terminal type identifiers and sensor extrinsic parameters. The coordinate transformation module is used to transform the metadata into the world coordinate system based on the terminal type identifier and the sensor extrinsic parameters. The data quality control module is used to remove metadata that does not meet one or more of the preset conditions, such as coverage, positioning signal quality, and data continuity. The metadata file generation module is used to decouple the remaining metadata after removal from the different data platforms through sensor extrinsic calibration and convert it into a metadata file with a fixed directory structure. The task identifier generation module is used to generate a unique identifier for the corresponding end-to-end task in the metadata file.

9. An electronic device, comprising: The method includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the multi-source spatial data management method for the high-precision map according to any one of claims 1-7.

10. A storage medium, being a computer readable storage medium, characterized in that, The system stores computer instructions for causing a computer to implement the multi-source spatial data management method for the high-precision map according to any one of claims 1-7.

Images