A high-precision map regionalization updating method and system based on spatio-temporal correlation

Abstract

The application provides a high-precision map regionalization updating method and system based on space-time correlation, and relates to the technical field of high-precision maps. The application binds real scene image data and associated data of a shooting terminal to form an image data unit with a space-time position marker. The image data unit is compared with a pre-stored real scene three-dimensional map to determine satellite positioning information correction coordinates, identify geographical feature characteristics, and distribute the geographical feature characteristics to corresponding space-time grid units according to the correction coordinates. Spatial clustering analysis is performed according to the distribution of the space-time grid units to determine key areas that need to be traced back. Image data units and geographical feature characteristics at different historical times are obtained through time retrieval and spatial indexing to reconstruct a time sequence change state. According to the difference between the time sequence change state and the pre-stored map, the real scene three-dimensional map is updated in a regionalized attribute, which can realize regionalized attribute updating of a high-precision map and improve the accuracy and pertinence of map updating.

Description

Technical Field

[0001] This application relates to the field of high-precision map technology, and in particular to a method and system for regional updating of high-precision maps based on spatiotemporal correlation. Background Technology

[0002] With the rapid development of autonomous driving and intelligent transportation, high-precision maps need to match the dynamic changes of geographical elements such as road facilities and traffic signs in real time. Therefore, technical solutions need to efficiently integrate their spatiotemporal information to achieve accurate identification of geographical elements and tracking of temporal changes, complete the orientation enhancement of map attributes, and ensure the timeliness and data richness of high-precision maps.

[0003] Currently, the mainstream approach is to use static threshold matching for batch periodic updates. This approach first performs basic preprocessing on the crowdsourced image data, then compares it with fixed areas of existing high-precision maps based on a preset static similarity threshold to filter out images suspected of being changed; subsequently, the filtered images are manually reviewed and labeled in batches at fixed intervals, and finally the attributes of the corresponding areas on the high-precision map are updated uniformly based on the labeling results, with temporary update processes only initiated in special emergency scenarios.

[0004] However, this scheme has core flaws: it lacks spatiotemporal correlation, relies solely on static threshold matching, and cannot build dynamic correlations, thus failing to identify cross-regional and temporally continuous changes in geographical elements and making it difficult to achieve spatiotemporal retrospection; there is an imbalance between update efficiency and accuracy, with periodic updates exhibiting lag, and manual review is prone to omissions or misjudgments due to the large amount of data; the utilization rate of crowdsourced data is low, failing to explore its spatiotemporal characteristics and clustering patterns, and the coordinate calibration accuracy is insufficient, affecting the accuracy of map attribute updates. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for regional updating of high-precision maps based on spatiotemporal correlation, so as to solve the problem of imbalance between updating efficiency and accuracy in the prior art.

[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for high-precision map regionalization updating based on spatiotemporal correlation, comprising:

[0007] Acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers;

[0008] The image data unit is compared with a pre-stored real-scene 3D map. Based on the comparison results, the corrected coordinates of the satellite positioning information in the associated data are determined. Geographic feature characteristics are simultaneously identified from the image data unit, and the geographic feature characteristics are assigned to the corresponding spatiotemporal grid unit based on the corrected coordinates of the satellite positioning information.

[0009] Based on the distribution of the spatiotemporal grid units, spatial clustering analysis is performed on the geographic feature characteristics to identify key areas that need to be retrospectively analyzed. Through time series retrieval and spatial location indexing, the image data units and geographic feature characteristics corresponding to the key areas at different historical time points are obtained. The change trajectory of each type of geographic feature in the time dimension is extracted, and the temporal change status of each geographic feature in the spatiotemporal grid units of the key areas is reconstructed.

[0010] Based on the comparison results between the temporal change state and the pre-stored real-scene 3D map, the real-scene 3D map is updated with regional attributes.

[0011] Optionally, based on the distribution of the spatiotemporal grid units, spatial clustering analysis is performed on the geographic feature characteristics to identify key areas requiring retrospective analysis. Through time series retrieval and spatial location indexing, the image data units and geographic feature characteristics corresponding to the key areas at different historical time points are obtained. The change trajectory of each type of geographic feature in the time dimension is extracted, and the temporal change state of each geographic feature within the spatiotemporal grid units of the key areas is reconstructed, including:

[0012] Based on the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit, spatial clustering analysis is performed to calculate the feature density value of each spatiotemporal grid unit.

[0013] The feature density value is compared with a preset density threshold. Spatiotemporal grid cells whose feature density value exceeds the preset density threshold are marked as high-density cells, and spatially continuous high-density cells are merged as key regions.

[0014] By using time series retrieval and spatial location indexing, the image data units and geographic feature characteristics of the key area at different historical time points are obtained;

[0015] Extract the vector values ​​of each type of geographic element feature at different time points, calculate the difference between the values ​​of each feature vector, and arrange the difference values ​​in chronological order to form a change trajectory sequence;

[0016] Based on the numerical differences in the change trajectory sequence, starting from the starting time point of the change trajectory sequence, the difference values ​​are sequentially superimposed onto the baseline state data of the key area to reconstruct the temporal change state of each geographic element.

[0017] Optionally, based on the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit, spatial clustering analysis is performed to calculate the feature density value of each spatiotemporal grid unit, including:

[0018] The quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit are converted into feature vectors, which contain quantity dimension and type dimension;

[0019] The neighborhood range of each spatiotemporal grid cell is determined based on a preset distance radius, and the basic density value of each spatiotemporal grid cell is calculated based on the total number of geographic feature characteristics within the statistical cell.

[0020] Based on the feature vectors of all spatiotemporal grid cells within the neighborhood, the distance weights of the neighborhood cells to the central cell are calculated, and based on the distance weights and the feature vectors of the neighborhood cells, the weighted density value of the central cell is calculated, and the weighted density value is used as the feature density value.

[0021] Optionally, the feature density value is compared with a preset density threshold, and spatiotemporal grid cells whose feature density value exceeds the preset density threshold are marked as high-density cells. Spatially continuous high-density cells are merged as key regions, including:

[0022] The feature density value of the spatiotemporal grid cell is obtained, wherein the spatiotemporal grid cell is obtained by dividing the target region according to a predefined spatial step size and time interval, and the feature density value is compared with a preset density threshold.

[0023] Spatiotemporal grid cells whose feature density values ​​exceed the preset density threshold are marked as high-density cells. All spatiotemporal grid cells of the high-density cells are traversed to determine whether there are other high-density cells in the spatial neighborhood of the current high-density cell.

[0024] If there are other high-density cells in the spatial neighborhood of the current high-density cell, then the current high-density cell will be merged with the other high-density cells in the spatial neighborhood to form a continuous merged cell.

[0025] Repeat the above traversal, judgment and merging process until no new merges occur or all high-density cells are processed, and then identify each consecutive merged cell as the key region.

[0026] Optionally, the image data unit is compared with a pre-stored real-scene 3D map, and the corrected coordinates of the satellite positioning information in the associated data are determined based on the comparison results. Simultaneously, geographic feature characteristics are identified from the image data unit, and the geographic feature characteristics are assigned to corresponding spatiotemporal grid units based on the corrected coordinates of the satellite positioning information. This includes:

[0027] Extract the spatial coordinates from the image data unit and the coordinates of the corresponding reference points in the pre-stored real-scene 3D map, and calculate the total coordinate offset based on the difference between the spatial coordinates and the reference point coordinates on each coordinate axis;

[0028] If the total coordinate offset exceeds a preset threshold, then corrected coordinates are generated based on the reference point coordinates, and the corrected coordinates replace the satellite positioning information in the associated data;

[0029] Geographic feature characteristics are simultaneously identified from the image data unit through pixel analysis, and the feature vectors of the geographic feature characteristics are assigned to the corresponding spatiotemporal grid units according to the latitude and longitude values ​​of the corrected coordinates.

[0030] Optionally, based on the comparison results between the temporal change state and the pre-stored real-scene 3D map, the real-scene 3D map is updated with regional attributes, including:

[0031] Calculate the attribute difference matrix based on the attribute data in the time-series change state and the current attribute data of the pre-stored real-scene 3D map;

[0032] From the attribute difference matrix, select geographic features whose difference values ​​exceed the allowable error range, obtain the unique identifier and current timestamp of the selected geographic features, and calculate the updated attribute values ​​based on the attribute difference matrix;

[0033] The unique identifier, the current timestamp, and the updated attribute value are combined into an attribute update instruction, which is then grouped by spatiotemporal grid units and written in batches to the database of the real-scene 3D map to update the attribute fields.

[0034] Optionally, real-scene image data and associated data generated during the shooting process of the shooting terminal are acquired, and the real-scene image data and the associated data are bound and associated to form an image data unit with spatiotemporal location markers, including:

[0035] The camera acquires real-scene image data and associated data generated during the shooting process from the shooting terminal, assigns a unique identifier to the real-scene image data, and binds the unique identifier to the associated data.

[0036] Spatial coordinates, timestamps, attitude data, and relative distances are extracted from the bound associated data. The spatial coordinates and timestamps are merged to generate a spatiotemporal identifier code, and the spatiotemporal identifier code is embedded into the metadata area of ​​the real scene image data.

[0037] The attitude data and relative distance are converted into a binary stream, and the real-scene image data containing spatiotemporal identification codes and the binary stream are encapsulated to form an image data unit with spatiotemporal location markings.

[0038] Secondly, this application provides a high-precision map regionalization update system based on spatiotemporal correlation, including:

[0039] The association module is used to acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers.

[0040] The allocation module is used to compare the image data unit with a pre-stored real-scene 3D map, determine the corrected coordinates of satellite positioning information in the associated data based on the comparison results, simultaneously identify geographic feature characteristics from the image data unit, and allocate the geographic feature characteristics to the corresponding spatiotemporal grid unit based on the corrected coordinates of the satellite positioning information.

[0041] The reconstruction module is used to perform spatial clustering analysis on the geographic feature characteristics based on the distribution status of the spatiotemporal grid units, determine the key areas that need to be backtracked, obtain the image data units and geographic feature characteristics of the key areas at different historical time points through time series retrieval and spatial location indexing, extract the change trajectory of each type of geographic feature in the time dimension, and reconstruct the temporal change status of each geographic feature in the spatiotemporal grid units of the key areas.

[0042] The update module is used to update the regional attributes of the real-world 3D map based on the comparison results between the time-series change state and the pre-stored real-world 3D map.

[0043] Thirdly, this application provides an electronic device, comprising:

[0044] Memory, used to store computer programs;

[0045] A processor, configured to execute the computer program to implement the steps of a high-precision map regionalization update method based on spatiotemporal correlation as described in the first aspect above.

[0046] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the high-precision map regionalization update method based on spatiotemporal correlation as described in the first aspect above.

[0047] This application provides a high-precision map regionalization update method based on spatiotemporal correlation. By acquiring and binding real-scene image data and related data from a camera terminal, image data units with spatiotemporal location markers are formed. This provides foundational data with accurate spatiotemporal labels for subsequent map updates, ensuring the spatiotemporal correlation and integrity of the data source. By comparing these image data units with pre-stored real-scene 3D maps to determine corrected coordinates for satellite positioning information, and simultaneously identifying geographic feature characteristics and assigning them to corresponding spatiotemporal grid units, coordinate calibration is achieved to improve positioning accuracy. Furthermore, the spatiotemporal grid units manage geographic feature characteristics. This provides a structured data foundation for subsequent analysis; by using spatial clustering analysis based on spatiotemporal grid cell distribution to identify key areas, and then using time series retrieval and spatial location indexing to obtain relevant data for different historical times in these areas, the change trajectory can be extracted and the temporal change state reconstructed. This allows for precise identification of areas requiring key updates, complete tracing of the spatiotemporal evolution of geographic elements, and providing dynamic change basis for updates; by using regional attribute updates based on the difference between the temporal change state and the pre-stored real-scene 3D map, precise and targeted map updates can be achieved, avoiding the waste of resources in overall updates and improving the efficiency and accuracy of map updates. Attached Figure Description

[0048] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A flowchart illustrating a high-precision map regionalization update method based on spatiotemporal correlation provided in this application embodiment;

[0050] Figure 2 A flowchart illustrating a high-precision map regionalization update method based on spatiotemporal correlation provided in this application embodiment;

[0051] Figure 3 A scene diagram illustrating a high-precision map regionalization update method based on spatiotemporal correlation provided in an embodiment of this application;

[0052] Figure 4 This is a schematic diagram of the structure of a high-precision map regionalization update system based on spatiotemporal correlation, provided in an embodiment of this application. Detailed Implementation

[0053] In the development of autonomous driving and intelligent transportation, high-precision maps need to keep up with changes in geographical elements such as road facilities and traffic signs, and also need to be able to trace the geographical state at different times. However, the commonly used "static threshold matching + batch periodic update" scheme has obvious problems: it cannot dynamically combine the time and location information of the images. Changes that span time and region, such as road construction, will be identified separately, and the change process cannot be fully traced; moreover, it needs to be updated at a fixed period, such as once a month, which makes it difficult to reflect new changes in a timely manner, and errors are prone to occur when manually checking a large amount of data; at the same time, it does not make full use of the time and location patterns in crowdsourced images, and the positioning accuracy is not enough, ultimately affecting the accuracy of map updates.

[0054] To address these issues, this application proposes a high-precision map regionalization update method based on spatiotemporal correlation. This method first binds captured real-world images with associated information to form data units with time and location markers. Then, by comparing and calibrating the positioning with existing maps, the identified geographic elements are assigned to corresponding spatiotemporal grids. Next, by analyzing the grid distribution, key areas requiring attention are identified, and imagery and element data of these areas at different times are traced to reconstruct the change process of geographic elements. Finally, based on the differences between the changes and the existing map, map attributes are updated accordingly. This method can completely correlate spatiotemporal information, solving the problem of existing solutions being unable to trace changes; it does not require a fixed cycle, can accurately locate and update key areas, avoiding delays and human error; and it can fully exploit data value and improve positioning accuracy, fundamentally solving the shortcomings of existing solutions and making high-precision map updates more timely and accurate.

[0055] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] The core of this application is to provide a high-precision map regionalization update method based on spatiotemporal correlation. A flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes:

[0057] S101. Acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers.

[0058] Optionally, step S101 may specifically include the following steps:

[0059] S1011. Obtain real-scene image data and associated data generated during the shooting process from the shooting terminal, assign a unique identifier to the real-scene image data, and bind the unique identifier to the associated data;

[0060] S1012. Extract spatial coordinates, timestamps, attitude data and relative distance from the bound associated data, merge the spatial coordinates and timestamps to generate a spatiotemporal identifier code, and embed the spatiotemporal identifier code into the metadata area of ​​the real scene image data;

[0061] S1013. The attitude data and relative distance are converted into a binary stream, and the real-scene image data containing the spatiotemporal identification code and the binary stream are encapsulated to form an image data unit with spatiotemporal location marking.

[0062] In the above steps, real-scene image data refers to real-world geographic environment images or video data captured by the camera terminal, including visual information such as roads, buildings, and traffic signs; associated data is auxiliary data generated synchronously during the shooting process, including spatial coordinates, timestamps, attitude data, and relative distances; a unique identifier is a unique code assigned to each segment of real-scene image data to uniquely identify the image data and achieve correspondence with the associated data; spatial coordinates refer to the geographic coordinates of the camera terminal at the time of shooting, usually obtained from a satellite positioning system; timestamps refer to the specific time information of the captured real-scene image data, accurate to the second or millisecond; attitude data refers to the attitude parameters of the camera terminal during shooting. Examples of parameters include tilt angle and rotation direction; relative distance refers to the distance data between the shooting terminal and surrounding reference objects such as roadside guardrails and traffic signs; the spatiotemporal identifier is a code generated by combining spatial coordinates and timestamps, integrating the location and time information of the image data; the metadata area is the part of the real scene image data file used to store descriptive information, which does not affect the visual content of the image itself; the binary stream is binary format data that can be directly processed by a computer after converting the attitude data and relative distance; the image data unit is a complete data unit that encapsulates real scene image data containing spatiotemporal identifier and binary stream format attitude data and relative distance data, with a clear spatiotemporal location marker.

[0063] In this embodiment, firstly, real-scene image data and associated data are obtained from the shooting terminal in step S1011. A unique identifier is assigned to the real-scene image data and bound to the associated data. Specifically, the system reads the real-scene image data and associated data stored in the shooting terminal through the data transmission interface, generates a unique identifier for each segment of real-scene image data using a universal unique identification code generation algorithm, and then writes the unique identifier into the attribute field of the associated data, so that each segment of real-scene image data can find the corresponding associated data through the unique identifier. For example, a camera terminal captures a 5-minute real-scene video of a road in area A. This video clearly records visual information such as lane lines, type C roadside trees on both sides, type D traffic lights at intersections, and type E bus stop signs on the roadside. Simultaneously, the camera terminal generates associated data such as the spatial coordinates, timestamp, tilt angle and rotation direction of the camera terminal, and distance to the type E bus stop sign. The system uses a UUID generation algorithm to generate a unique identifier "UUID-20250902-A001" for this real-scene video and writes this identifier into the corresponding associated data attribute field. In subsequent data processing, simply inputting "UUID-20250902-A001" can quickly match all associated data corresponding to this real-scene video, avoiding mismatches between the image and auxiliary data.

[0064] Secondly, in step S1012, spatial coordinates, timestamps, attitude data, and relative distances are extracted from the bound associated data. The spatial coordinates and timestamps are merged to generate a spatiotemporal identifier code, which is then embedded into the metadata area of ​​the real-scene image data. Specifically, the system filters spatial coordinates, timestamps, attitude data, and relative distances from the associated data bound with unique identifiers. First, the spatial coordinates and timestamps are formatted to ensure that the latitude and longitude of the spatial coordinates are retained to six decimal places and the timestamps are accurate to the second. Then, a string concatenation algorithm is used to merge the two into a spatiotemporal identifier code in the form of "latitude and longitude, timestamp". Finally, using an image metadata editing tool such as ExifTool, the generated spatiotemporal identifier code is written into the user annotation field of the metadata area of ​​the real-scene image data. For example, from the associated data bound to "UUID-20250902-A001", the spatial coordinates "116.123456, 39.654321", the timestamp "2025-09-02 14:30:00", the attitude data "tilt angle 2°, rotation direction counterclockwise 3°", and the relative distance "1.8 meters from type E bus stop sign" are filtered and extracted. First, the spatial coordinates and timestamp are formatted and standardized to confirm that the spatial coordinates conform to six decimal places of latitude and longitude. After the data format and timestamp are accurate to the second, a string concatenation algorithm is used to merge the two into a spatiotemporal identifier code of "116.123456,39.654321,2025-09-0214:30:00". Then, using the ExifTool, this spatiotemporal identifier code is written into the metadata user comment field of the road real-scene video corresponding to "UUID-20250902-A001", so that when viewing the video metadata later, its shooting location and time information can be directly obtained.

[0065] Finally, in step S1013, the attitude data and relative distance are converted into a binary stream. The real-scene image data containing the spatiotemporal identification code and this binary stream are then encapsulated to form an image data unit with spatiotemporal location markers. Specifically, the system uses a data type conversion algorithm, such as Base64 encoding, to convert the attitude data and relative distance from the original text or numerical format into a binary stream that can be efficiently processed by a computer. After the conversion, a data encapsulation tool, such as a ZIP compression tool, is used to package the real-scene image data containing the spatiotemporal identification code and the converted binary stream into a complete data file with the suffix ".mapunit". This ".mapunit" file is the image data unit with spatiotemporal location markers. For example, the attitude data "tilt angle 2°, rotation direction counterclockwise 3°" and relative distance "1.8 meters from type E bus stop sign" from the above-mentioned associated data are input into a Base64 encoding conversion algorithm to convert these original text format data into a binary stream in the form of "SGVsbG8gV29ybGQh...". Then, a ZIP compression and packaging tool is started, and the road scene video with "UUID-20250902-A001" embedded with the spatiotemporal identifier "116.123456,39.654321,2025-09-0214:30:00" is added to the compression list along with the converted binary stream. After setting the compression level to normal compression, the file "UUID-20250902-A001.mapunit" is generated. This file contains the visual information, spatiotemporal information, and auxiliary attitude and distance information of the image, which can be directly used for subsequent comparison with the real scene 3D map.

[0066] In a practical application, a smart transportation data collection project deployed 20 C-brand camera terminals to collect data for the urban road network update needs of City B. These terminals were all mounted on a D-type data collection vehicle, with each terminal's lens facing directly forward of the vehicle. Each terminal was equipped with a high-precision satellite positioning module, attitude sensor, and laser ranging module, enabling simultaneous acquisition of real-world images and related data. Data collection ran daily from 8:00 AM to 6:00 PM, with the vehicle traveling along the pre-designated first and second ring roads and major arterial roads of City B. During the journey, each camera terminal automatically captured a 10-second real-world video clip every 30 seconds. Simultaneously, the satellite positioning module acquired the latitude and longitude coordinates of the captured image in real time, the system automatically recorded the exact time of capture, the attitude sensor detected and recorded the tilt angle and rotation direction of the camera terminal, and the laser ranging module measured the distance between the camera terminal and reference objects such as roadside guardrails and traffic signs. After data acquisition, the camera transmits the real-scene image data and associated data to the project's backend system via a 4G network. The backend system first executes step S1011, assigning a unique UUID identifier to each 10-second segment of real-scene road video, such as "UUID-20250902-B001" or "UUID-20250902-B002", and binding each identifier to the corresponding associated data. Next, step S1012 is executed, extracting spatial coordinates, timestamps, and other information from the bound associated data, generating spatiotemporal identification codes after unifying the format, such as "116.234567,39.765432,2025-09-0208:00:30", and then embedding these identification codes into the metadata area of ​​the corresponding video. Subsequently, step S1013 is executed, converting the attitude data and relative distance into a binary stream using the Base64 algorithm, and encapsulating it together with the video containing the spatiotemporal identification codes into an image data unit in ".mapunit" format using a ZIP tool. These image data units are stored in the distributed database of the back-end system. Staff can quickly retrieve the required data using unique identifiers or spatiotemporal identification codes. When updating the high-precision map of City B, these image data units can be called to perform comparison operations with the existing real-scene 3D map. The entire data acquisition and processing process does not require manual data matching or metadata editing and is completed automatically. More than 5,000 real-scene image data segments can be processed every day, effectively ensuring the efficiency and quality of data supply for the high-precision map update of City B.

[0067] In the overall scheme of step S101 above, precise binding between real-scene image data and associated data is achieved by assigning a unique identifier to the real-scene image data, avoiding data correspondence errors; spatial coordinates and timestamps are merged into a spatiotemporal identifier code and embedded with metadata, so that the image data carries clear spatiotemporal information, facilitating subsequent rapid positioning and tracing; attitude data and relative distance are converted into binary streams and encapsulated with the image data, improving data storage and transmission efficiency while ensuring data integrity. Overall, the scheme achieves the organic integration of real-scene image data and associated data, forming a clearly structured and information-complete image data unit. This lays a solid foundation for subsequent steps such as comparing image data units with real-scene 3D maps and identifying geographic feature characteristics, ensuring that subsequent high-precision map updates can be carried out based on accurate and complete basic data.

[0068] S102. The image data unit is compared with the pre-stored real-scene 3D map. The corrected coordinates of the satellite positioning information in the associated data are determined according to the comparison results. Geographic feature characteristics are identified from the image data unit simultaneously. The geographic feature characteristics are assigned to the corresponding spatiotemporal grid unit according to the corrected coordinates of the satellite positioning information.

[0069] Optionally, step S102 may specifically include the following steps:

[0070] S1021. Extract the spatial coordinates from the image data unit and the reference point coordinates of the corresponding positions in the pre-stored real-scene 3D map, and calculate the total coordinate offset based on the difference between the spatial coordinates and the reference point coordinates on each coordinate axis.

[0071] S1022. If the total coordinate offset exceeds a preset threshold, then generate corrected coordinates based on the reference point coordinates, and replace the satellite positioning information in the associated data with the corrected coordinates;

[0072] S1023. Simultaneously identify geographic feature characteristics from the image data unit through pixel analysis, and allocate the feature vector of the geographic feature to the corresponding spatiotemporal grid unit according to the latitude and longitude values ​​of the corrected coordinates.

[0073] In the above steps, the image data unit refers to a real-scene image data unit encapsulated with a spatiotemporal identifier, a binary stream of attitude data, and a binary stream of relative distance, bearing a clear spatiotemporal location marker; the pre-stored real-scene 3D map refers to map data pre-constructed and stored in the system, containing geographic spatial location and 3D morphological information, including a large number of reference point coordinates used for positioning calibration; spatial coordinates refer to the original geographic location coordinates extracted from the image data unit at the time of shooting by the shooting terminal; reference point coordinates refer to the standard coordinates of a specific location in the pre-stored real-scene 3D map corresponding to the spatial coordinates of the image data unit, used to calibrate the original spatial coordinates; coordinate axes refer to the X-axis, Y-axis, and Z-axis in 3D space, corresponding to the east-west direction, the north-south direction, and the vertical height direction, respectively; the difference refers to the numerical difference between the spatial coordinates of the image data unit and the reference point coordinates on the X-axis, Y-axis, and Z-axis; the total coordinate offset refers to the value obtained by calculating the comprehensive deviation of the differences between the X-axis, Y-axis, and Z-axis, reflecting... The overall deviation between the original spatial coordinates and the standard coordinates; the preset threshold refers to the critical value set in advance by the system to determine whether the total coordinate offset needs to be calibrated; the corrected coordinates refer to the accurate coordinates calculated based on the reference point coordinates when the total coordinate offset exceeds the preset threshold, used to replace the original satellite positioning information; satellite positioning information refers to the original geographical location information obtained by the shooting terminal through the satellite positioning system; pixel resolution refers to the technical means of identifying geographic elements by analyzing the pixel distribution, color, outline and other features of the image in the image data unit; geographic element features refer to the feature information of geographic objects such as roads, traffic signs, buildings, and vegetation identified from the image, usually stored in the form of feature vectors; feature vectors refer to the numerical vectors that convert the attributes of geographic element features into numerical vectors, which are convenient for computer processing; spatiotemporal grid units refer to the units formed by dividing geographic space into grids of preset size and combining them with the time dimension, used to store geographic element features, with each unit corresponding to a specific spatial range and time interval.

[0074] In this embodiment, firstly, step S1021 extracts the spatial coordinates from the image data unit and the coordinates of the corresponding reference points in the pre-stored real-world 3D map. The total coordinate offset is then calculated based on the differences between the spatial coordinates and the reference point coordinates on each coordinate axis. Specifically, the system first extracts the original spatial coordinates from the metadata area of ​​the image data unit, then locates the corresponding area in the pre-stored real-world 3D map based on these spatial coordinates, and retrieves the standard coordinates of three or more reference points within that area. Next, the differences between the original spatial coordinates and the coordinates of each reference point on the X, Y, and Z axes are calculated respectively. The total coordinate offset is then calculated using the Euclidean distance formula. ,in , , These represent the original spatial coordinates of the image data unit on the X, Y, and Z axes, respectively. , , These are the reference point coordinates on the X, Y, and Z axes, respectively; finally, the average of the calculation results from multiple reference points is taken as the final total coordinate offset. For example, the original spatial coordinates extracted from a certain image data unit are (116.123456, 39.654321, 50.0). Three reference points corresponding to this location are found in a pre-stored real-world 3D map, with coordinates of (116.123458, 39.654323, 50.1), (116.123457, 39.654322, 50.0), and (116.123459, 39.654322, 50.0). 4, 50.1); Calculate the differences between the original coordinates and the coordinates of the three reference points on each axis. Taking the first reference point as an example, the difference on the X-axis is 116.123456-116.123458=-0.000002, the difference on the Y-axis is 39.654321-39.654323=-0.000002, and the difference on the Z-axis is 50.0-50.1=-0.1; Substitute these values ​​into the Euclidean distance formula to calculate the offset corresponding to the first reference point. Similarly, the offsets of the other two reference points are calculated to be approximately 0.099 and 0.101 respectively. After averaging, the final total coordinate offset is approximately 0.1.

[0075] Secondly, if the total coordinate offset exceeds a preset threshold in step S1022, corrected coordinates are generated based on the reference point coordinates, and the corrected coordinates replace the satellite positioning information in the associated data. Specifically, the system compares the total coordinate offset calculated in step S1021 with the preset threshold. If the total coordinate offset is greater than the preset threshold, a weighted average method is used, with the reliability of the reference points as the weight, to calculate the corrected coordinates for multiple reference points. If the total coordinate offset is less than or equal to the preset threshold, no corrected coordinates need to be generated, and the original spatial coordinates are used directly. Finally, the generated corrected coordinates are written into the associated data corresponding to the image data unit, replacing the original satellite positioning information. For example, the system's preset total coordinate offset threshold is 0.05. The total coordinate offset calculated in step S1021 is approximately 0.1, exceeding the preset threshold. At this point, the confidence weights of the three reference points are retrieved, with weights of 0.4, 0.3, and 0.3 respectively. The corrected coordinates are calculated using a weighted average method. The corrected X-axis coordinate is 116.123458×0.4 + 116.123457×0.3 + 116.123459×0.3 = 116.1234. 58. The corrected Y-axis coordinates are 39.654323×0.4+39.654322×0.3+39.654324×0.3=39.654323, and the corrected Z-axis coordinates are 50.1×0.4+50.0×0.3+50.1×0.3=50.07. Replace the original satellite positioning information in the associated data of this image data unit with the corrected coordinates (116.123458, 39.654323, 50.07).

[0076] Finally, in step S1023, geographic feature characteristics are synchronously identified from the image data unit through pixel analysis. Based on the latitude and longitude values ​​of the corrected coordinates, the feature vectors of the geographic feature characteristics are allocated to the corresponding spatiotemporal grid units. Specifically, the system uses a convolutional neural network algorithm to perform pixel analysis on the real-scene images in the image data unit, identifying geographic features such as roads, traffic lights, and roadside guardrails in the images, and extracting the shape, size, color, and other attributes of each geographic feature, converting them into feature vectors. At the same time, based on the latitude and longitude values ​​of the corrected coordinates, the spatial range corresponding to the image data unit is determined, and then combined with the timestamp of the image capture, the pre-divided spatiotemporal grid unit is located. Finally, the feature vectors of the identified geographic feature characteristics are stored in the corresponding spatiotemporal grid units, completing the allocation. For example, the system performs pixel analysis on the real-scene images in the aforementioned image data unit, and uses a convolutional neural network algorithm to identify that the image contains a two-way four-lane road, two red circular traffic lights, and a section of gray wave-shaped guardrail; it extracts the road's feature attributes, such as width of 8 meters, number of lanes of 4, and road surface color of gray-black, and converts them into a feature vector (8, 4, gray-black); it extracts the traffic light's feature attributes, such as circular shape, red color, and height of 5 meters, and converts them into a feature vector (circular, red, 5); it extracts the guardrail's feature attributes, such as wave-shaped type and gray color. The image, 20 meters in length, is converted into a feature vector (waveform, gray, 20). Based on the corrected coordinates (116.123458, 39.654323, 50.07), the spatial range corresponding to the image is determined to belong to "Spatiotemporal Grid Unit G-20250902-01", where G represents the grid number, 20250902 represents the shooting date, and 01 represents the first time segment of the day. The feature vectors of the road, traffic lights, and guardrail are all assigned and stored in "Spatiotemporal Grid Unit G-20250902-01".

[0077] In a practical application, a smart map update project conducted high-precision map updates for the urban road network in region A. The system pre-stored a real-world 3D map of region A, which was divided into spatial grids with a latitude and longitude range of 0.0001 degrees and time segments in 1-hour increments, forming spatiotemporal grid units. The project used a B-brand camera terminal to collect image data units of roads in region A, and after collection, the data was transmitted to the backend system for processing. The system first performs step S1021 on a certain image data unit to extract the original spatial coordinates of the unit (116.234567, 39.765432, 45.0). In the pre-stored real-scene 3D map of area A, the system finds the area corresponding to the coordinates and retrieves the coordinates of four reference points within that area, namely (116.234569, 39.765434, 45.1), (116.234568, 39.765433, 45.0), (116.234570, 39.765435, 45.1), and (116.234566, 39.765431, 44.9). The system calculates the difference between the original coordinates and each reference point on each axis, calculates the offset corresponding to each reference point using the Euclidean distance formula, and takes the average value to obtain a total coordinate offset of approximately 0.12. Next, step S1022 is executed. The system's preset total coordinate offset threshold is 0.06. 0.12 exceeds the threshold, so the confidence weights of the four reference points are retrieved, which are 0.3, 0.25, 0.25 and 0.2 respectively. The weighted average method is used to calculate the corrected coordinates (116.234568, 39.765433, 45.02), and these corrected coordinates replace the original satellite positioning information in the associated data of the image data unit. Following step S1023, the system uses a convolutional neural network algorithm to perform pixel-by-pixel analysis on the real-world image of the image data unit, identifying a six-lane dual carriageway road, three green arrow traffic lights, and two sections of blue guardrails within the image. After converting the attributes of these geographic elements into feature vectors, the system determines the corresponding spatiotemporal grid unit as "Spatiotemporal Grid Unit H-20250902-10" based on the corrected coordinate latitude and longitude values ​​and the image capture timestamp (2025-09-02 10:30:00), where H is the grid number, 20250902 is the date, and 10 represents the 10-11 AM time segment. The feature vectors of all geographic elements are then assigned to this unit. Throughout this process, the system can process multiple image data units simultaneously, completing image data comparison, coordinate correction, and geographic element allocation within a 500 square kilometer area of ​​Region A daily, providing accurate element data support for subsequent regional updates of high-precision maps.

[0078] In the overall scheme of step S102 above, by extracting spatial coordinates and calculating the total coordinate offset with reference point coordinates, the degree of deviation of the original satellite positioning information can be accurately determined; by comparing the total coordinate offset with a preset threshold, corrected coordinates are generated, which can replace the original positioning information with large deviations and improve the accuracy of geographic feature location; by identifying geographic feature features through pixel analysis and assigning them to corresponding spatiotemporal grid units, the structured storage and management of geographic features are realized. The overall process combines image data with real-world 3D maps to complete coordinate calibration and geographic feature allocation, providing a precise and orderly data foundation for subsequent spatial clustering analysis and map attribute updates, effectively avoiding geographic feature identification errors caused by positioning deviations, and enabling geographic feature data to be classified according to spatiotemporal grid units, facilitating subsequent tracking of geographic feature changes by region and time.

[0079] This application provides a flowchart illustrating a high-precision map regionalization update method based on spatiotemporal correlation. Figure 2 As shown, it includes the following:

[0080] S103. Based on the distribution status of the spatiotemporal grid units, perform spatial clustering analysis on the geographic feature characteristics to determine the key areas that need to be backtracked. Through time series retrieval and spatial location indexing, obtain the image data units and geographic feature characteristics of the key areas at different historical time points, extract the change trajectory of each type of geographic feature in the time dimension, and reconstruct the temporal change status of each geographic feature in the spatiotemporal grid units of the key areas.

[0081] Optionally, step S103 may specifically include the following steps:

[0082] S1031. Based on the quantity and type distribution of geographic element features in the spatiotemporal grid unit, perform spatial clustering analysis and calculate the feature density value of each spatiotemporal grid unit;

[0083] Specifically, step S1031 may include the following processes: converting the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit into feature vectors, wherein the feature vectors contain quantity dimension and type dimension; determining the neighborhood range of each spatiotemporal grid unit according to a preset distance radius; calculating the basic density value of each spatiotemporal grid unit according to the total quantity of geographic feature characteristics within the statistical unit; calculating the distance weight of the neighborhood unit to the central unit according to the feature vectors of all spatiotemporal grid units within the neighborhood range; and calculating the weighted density value of the central unit based on the distance weight and the feature vectors of the neighborhood units, and using the weighted density value as the feature density value.

[0084] S1032. Compare the feature density value with a preset density threshold, mark the spatiotemporal grid cells whose feature density value exceeds the preset density threshold as high-density cells, and merge the spatially continuous high-density cells as key regions;

[0085] Specifically, step S1032 may include the following process: obtaining the feature density value of a spatiotemporal grid cell, wherein the spatiotemporal grid cell is obtained by dividing the target region according to a predefined spatial step size and time interval; comparing the feature density value with a preset density threshold; marking spatiotemporal grid cells whose feature density value exceeds the preset density threshold as high-density cells; traversing all spatiotemporal grid cells of the high-density cells; determining whether there are other high-density cells in the spatial neighborhood of the current high-density cell; if there are other high-density cells in the spatial neighborhood of the current high-density cell, merging the current high-density cell with the other high-density cells in the spatial neighborhood to form a continuous merged cell; repeating the above traversal, determination and merging process until no new merging occurs or all high-density cells are processed; and determining each of the obtained continuous merged cells as a key region.

[0086] S1033. Obtain the image data units and geographic feature characteristics of the key area at different historical time points through time series retrieval and spatial location indexing;

[0087] S1034. Extract the values ​​of the vectors of each type of geographic element feature at different time points, calculate the difference between the values ​​of each feature vector, and arrange the difference values ​​in chronological order to form a change trajectory sequence.

[0088] S1035. Based on the numerical differences in the change trajectory sequence, starting from the starting time point of the change trajectory sequence, the difference values ​​are sequentially superimposed onto the baseline state data of the key area to reconstruct the temporal change state of each geographic element.

[0089] In the above steps, a spatiotemporal grid cell refers to a cell used to store geographic feature characteristics, divided according to a predefined spatial step size and time interval. Each cell corresponds to a specific spatial range and time segment. Geographic feature characteristics refer to the attribute information of objects such as roads and traffic signs identified from images, such as shape and size. Spatial clustering analysis refers to an analysis technique that groups spatiotemporal grid cells with similar features into one category based on the distribution patterns of geographic feature characteristics. Feature density value refers to a numerical value reflecting the density of geographic feature characteristics within a spatiotemporal grid cell, including basic density and weighted density. Feature vector refers to a numerical vector that transforms the quantity and type of geographic feature characteristics within a spatiotemporal grid cell. Preset distance radius refers to a pre-set spatial distance used to determine the neighborhood range of a certain spatiotemporal grid cell. Neighborhood range refers to the set of surrounding grid cells divided according to the preset distance radius with the target spatiotemporal grid cell as the center. Basic density value refers to the density calculated only based on the total number of geographic feature characteristics within the target spatiotemporal grid cell. Distance weight refers to the weight calculated based on the spatial distance between neighboring cells and the central cell. The closer the distance, the greater the weight; the weighted density value refers to the comprehensive density value calculated by combining the basic density of the central unit with the feature vectors of neighboring units and distance weights; the preset density threshold refers to the pre-set critical value for judging whether a spatiotemporal grid unit is a high-density unit; a high-density unit refers to a spatiotemporal grid unit whose feature density value exceeds the preset density threshold; a key area refers to an area formed by merging multiple spatially continuous high-density units; time series retrieval refers to the technology of querying data of a key area at different historical time points in chronological order; spatial location index refers to the indexing technology of quickly locating the data storage location of the corresponding data of a key area based on spatial coordinates; the change trajectory sequence refers to the sequence formed by arranging the difference values ​​of a certain type of geographic feature at different time points in chronological order; the difference value refers to the difference in the feature vector values ​​of the same geographic feature at different time points; the baseline state data refers to the original state data of geographic features in a key area at a certain initial time point; the temporal change state refers to the complete change process of a geographic feature from the initial to the present by superimposing the difference values ​​of different time points.

[0090] In this embodiment, firstly, spatial clustering analysis is performed in step S1031 based on the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit to calculate the feature density value of each spatiotemporal grid unit. Specifically, the quantity and type of geographic feature characteristics within each spatiotemporal grid unit are first converted into feature vectors. For example, the feature vector of a certain unit might be (quantity: 3, type: road-1, traffic light-2). Then, the neighborhood range of the unit is determined according to a preset distance radius, the total quantity of geographic feature characteristics within the unit is counted, and the basic density value is calculated. The basic density = total quantity / grid unit spatial area. If the unit area is 1 square kilometer, then the basic density = 3 / 1 = 3. Next, based on the spatial distance between neighboring units and the central unit, the distance weight is calculated using the distance decay formula. The formula is: Distance Weight The distance is the spatial distance between the neighboring unit and the central unit (unit: kilometers). Then, the weight of unit A, which is 1 kilometer away, is 0.5, and the weight of unit B, which is 2 kilometers away, is ≈0.33. Finally, the weighted density value of the central unit is calculated by combining the feature vectors of the neighboring units. The weighted density = basic density + Σ (number of feature vectors of neighboring units × distance weight). This value is the feature density value. For example, the geographic feature of "Spatiotemporal grid unit G-20250902-01" is one two-way four-lane road and two circular traffic lights, with feature vectors of (quantity: 3, type: road-1, traffic light-2); the preset distance radius covers three surrounding grid units, with a unit area of ​​1 square kilometer and a base density of 3; neighboring units G02 (distance 1 km, feature vector quantity 4, weight 0.5), G03 (distance 1.5 km, quantity 3, weight 0.4), and G04 (distance 2 km, quantity 2, weight 0.33); the weighted density = 3 + (4 × 0.5 + 3 × 0.4 + 2 × 0.33) = 6.86, that is, the feature density value is 6.86.

[0091] Secondly, by comparing the feature density values ​​with the preset density threshold in step S1032, high-density units are marked and merged into key regions. The specific process is as follows: First, the feature density values ​​of all spatiotemporal grid units are obtained and compared with the preset density threshold. If they exceed the threshold, they are marked as high-density units. Then, all high-density units are traversed to determine whether there are other high-density units in the spatial neighborhood of the current unit. If so, they are merged. The operation is repeated until no new merging occurs. Finally, the continuously merged units are the key regions. For example, the preset density threshold is 5. The density of "spatial grid unit G-20250902-01" is 6.86, and G02 (5.9) and G03 (5.5) are all high-density units. When traversing G01, the neighboring G02 and G03 are merged. The density of the neighboring G04 is 4.8, which does not reach the threshold. Merging is stopped, and "key region A1" is formed.

[0092] Next, through step S1033, image data units and geographic feature characteristics of the key area at different historical time points are obtained by time series retrieval and spatial location indexing. Specifically, a spatial location index is established based on the spatial coordinates of the key area to locate the data storage path; then, by time series retrieval, the image data units and geographic feature characteristics corresponding to each time point are extracted. For example, for "key area A1", the "key area A1 data folder" is located by establishing an index using spatial coordinates; data from 8:00 AM on September 1, 2025 to 8:00 AM on September 2, 2025 is retrieved by time series, extracting data such as 8:00 AM on September 1 (image "UUID-20250901-08-G01.mapunit", features 1 road and 1 traffic light) and 12:00 PM on September 1 (image "UUID-20250901-12-G01.mapunit", features 1 road and 2 traffic lights).

[0093] Then, step S1034 extracts the vector values ​​of each type of geographic feature at different time points, calculates the difference values, and forms a change trajectory sequence. Specifically, for a given type of geographic feature, the feature vector values ​​at different time points are extracted; the numerical differences between adjacent time points are calculated; and the difference values ​​are arranged in chronological order to form a change trajectory sequence. For example, for traffic lights within "Key Area A1", the values ​​are extracted as follows: 1 at 8:00 AM on September 1st, 2 at 12:00 PM, 2 at 6:00 PM, and 2 at 8:00 AM on September 2nd; the difference values ​​are calculated as 1, 0, and 0; thus forming the change trajectory sequence [1, 0, 0].

[0094] Finally, in step S1035, the temporal change state is reconstructed by overlaying the difference values ​​in the change trajectory sequence onto the baseline state data. Specifically, the baseline state data of the geographic elements is determined; starting from the initial time point, the difference values ​​are sequentially overlaid onto the baseline data; and finally, the temporal change state is restored. For example, the baseline state of the traffic lights in "Key Area A1" is "Quantity 1, Type Circular" at 8:00 AM on September 1st; the overlay sequence [1,0,0]: the quantity changes to 2 at 12:00 PM on September 1st, and remains at 2 thereafter; the reconstructed temporal change state is "1 at 8:00 AM on September 1st → 2 added at 12:00 PM → no change from 6:00 PM to 8:00 AM on September 2nd".

[0095] In practical applications, a smart transportation project conducted high-precision map spatiotemporal retrospective updates for roads in the core business district of City B. The system divided the space into spatial grids at 0.0001 degrees of latitude and longitude and into hourly time intervals to form a spatiotemporal grid unit library, collecting crowdsourced image data from September 1st to 7th. During execution S1031, "Spatiotemporal grid unit C-20250903-14" (14:00 on September 3rd) contained 2 roads, 3 signs, and 1 traffic light, with a feature vector (number 6). The preset distance radius covered 4 neighboring units, and the basic density = 6 / 0.000001 = 6000. The weighted density, calculated based on the neighboring unit weights, was 6009.05. In S1032, a preset threshold of 5000 was used to mark the four high-density units, including C14, and merge them into "Key Area B1". In S1033, the data was located using an index, retrieving data from 8:00 on September 1st to 20:00 on September 7th and extracting features. S1034 extracts numerical values ​​from traffic signs and calculates the difference values, forming a sequence [0,1,0,0,0]. S1035 uses the two signs at 8:00 AM on September 1st as a baseline, overlays the difference values, and reconstructs the change status: 2 signs on September 1st → 3 signs added at 12:00 PM on September 2nd → no further changes. This process provides accurate time-series data for map updates, improving update efficiency and accuracy.

[0096] In the overall scheme of step S103 above, spatial clustering analysis is used to calculate feature density values, which can accurately identify spatiotemporal grid units with dense geographic feature characteristics. By marking high-density units and merging them into key areas, the focus can be placed on areas requiring key backtracking, avoiding resource waste caused by indiscriminate analysis. Through time series retrieval and spatial location indexing, historical data of key areas can be quickly obtained, improving data acquisition efficiency. By extracting difference values ​​to form a change trajectory sequence and overlaying it with benchmark data, the temporal change process of geographic elements can be completely restored, realizing spatiotemporal backtracking. The overall process, from area selection to data retrieval and then to change reconstruction, forms a complete spatiotemporal backtracking mechanism for key area geographic elements. It effectively solves the problem that traditional methods are difficult to accurately locate key areas and cannot completely trace element changes, providing detailed and accurate temporal change basis for subsequent high-precision map regional attribute updates, ensuring that map updates can accurately reflect the dynamic changes of geographic elements.

[0097] S104. Based on the comparison results between the time-series change state and the pre-stored real-scene 3D map, the real-scene 3D map is updated with regional attributes.

[0098] Optionally, step S104 may specifically include the following steps:

[0099] S1041. Calculate the attribute difference matrix based on the attribute data in the time-series change state and the current attribute data of the pre-stored real-scene 3D map;

[0100] S1042. Filter out geographic features whose difference values ​​exceed the allowable error range from the attribute difference matrix, obtain the unique identifier and current timestamp of the filtered geographic features, and calculate the updated attribute values ​​based on the attribute difference matrix.

[0101] S1043. Combine the unique identifier, the current timestamp, and the updated attribute value into an attribute update instruction, group them by spatiotemporal grid unit, and write them in batches into the database of the real-scene 3D map to update the attribute fields.

[0102] In the above steps, the temporal change status refers to the complete change process data of geographic elements in the key area at different points in time, obtained through backtracking analysis, including attribute update information of geographic elements; the pre-stored real-scene 3D map refers to the 3D map that is pre-built and saved in the system and contains the current attribute data of geographic elements; the attribute difference matrix refers to the matrix-form data formed by calculating the difference between the attribute data in the temporal change status and the current attribute data of the real-scene 3D map, used to visually present the attribute deviation between the two; attribute data refers to specific information describing the characteristics of geographic elements, such as the number of traffic lights, the number of lanes on a road, the type of guardrail, etc.; the allowable error range refers to the acceptable deviation range of attribute data set in advance by the system, and attribute updates are required if this range is exceeded; geographic elements The unique identifier refers to the unique code assigned to each geographic feature to uniquely identify the feature and ensure the accuracy of the updated object; the current timestamp refers to the specific time information when the attribute update operation is performed, used to record the update time; the updated attribute value refers to the latest attribute data of the geographic feature that needs to be written into the real-world 3D map, calculated based on the attribute difference matrix; the attribute update instruction refers to the instruction data containing the geographic feature's unique identifier, current timestamp, and updated attribute value, used to trigger map attribute updates; spatiotemporal grid cell grouping refers to classifying attribute update instructions according to their corresponding spatiotemporal grid cells for easy batch processing; the real-world 3D map database refers to the database that stores all data of the real-world 3D map; the attribute field refers to the specific field in the database used to store geographic feature attribute data.

[0103] In this embodiment, step S1041 first calculates an attribute difference matrix based on the attribute data in the time-varying state and the current attribute data of the pre-stored real-scene 3D map. Specifically, the attribute data of each geographic element in the time-varying state and the current attribute data of the corresponding geographic element in the real-scene 3D map are extracted. Then, a matrix construction algorithm is used, with geographic elements as rows and attribute types as columns, to fill the two types of attribute data into the corresponding positions in the matrix. Finally, the difference between the data at the corresponding positions is calculated to form the attribute difference matrix. For example, for the geographic element "Key Area A1", the traffic light attribute data in the time-series change state is "Quantity 2, Type: Circular", and the road attribute data is "Number of Lanes 4, Width 8 meters". In the pre-stored real-scene 3D map, the current attribute data of the corresponding element is "Quantity of Traffic Lights 1, Type: Circular, Number of Road Lanes 4, Width 8 meters". When constructing the attribute difference matrix, the rows are set as "Traffic Lights, Roads", and the columns are set as "Quantity, Type, Number of Lanes, Width". After filling in the two types of data, the difference is calculated, and the attribute difference matrix is ​​obtained as follows: Traffic Lights row (Quantity: 2-1=1, Type: Circular-Circular=0), Roads row (Number of Lanes: 4-4=0, Width: 8-8=0).

[0104] Secondly, in step S1042, geographic features whose difference values ​​exceed the allowable error range are filtered from the attribute difference matrix. The unique identifier and current timestamp of the filtered geographic features are obtained, and the updated attribute values ​​are calculated based on the attribute difference matrix. Specifically, the attribute difference value of each geographic feature in the attribute difference matrix is ​​compared with the allowable error range to filter out geographic features whose difference values ​​exceed the range. Then, the unique identifier of the geographic feature is queried from the system database, and the current timestamp at the time of the filtering operation is recorded. Finally, based on the attribute difference matrix, the current attribute value of the real-world 3D map is added to the difference value to obtain the updated attribute value. For example, if the allowable error range is set to "0 for quantity and 0.5 meters for size", the difference in the number of traffic lights in the attribute difference matrix is ​​1, which exceeds the allowable error of 0. The differences in all attributes of the road are 0 and do not exceed the range. The geographic element of traffic lights is selected. Its unique identifier is "TL-2025-A1-001" and the current timestamp is "2025-09-02 16:00:00". The updated attribute values ​​are calculated as follows: number of traffic lights = 1 + 1 = 2, and the type remains circular. That is, the updated attribute value is "2 for quantity and circular for type".

[0105] Finally, in step S1043, the unique identifier, current timestamp, and updated attribute value are combined into an attribute update instruction, which is then grouped by spatiotemporal grid unit and batch-written into the database of the real-scene 3D map to update the attribute fields. Specifically, the unique identifier, current timestamp, and updated attribute value of the selected geographic features are combined according to a preset format to form an attribute update instruction; then, the spatiotemporal grid units corresponding to each attribute update instruction are grouped, with all attribute update instructions for the same spatiotemporal grid unit grouped together; finally, using batch database writing technology, the updated data corresponding to each group of attribute update instructions is written into the attribute fields of the corresponding geographic features in the real-scene 3D map database, completing the attribute update. For example, the unique identifier of the traffic light "TL-2025-A1-001", the current timestamp "2025-09-02 16:00:00", and the updated attribute values ​​"Quantity 2, Type: Circle" are combined into an attribute update instruction. This instruction corresponds to "Spatiotemporal Grid Unit G-20250902-01", and is included in the update instruction group of this unit. Through a database batch writing tool, the updated attribute values ​​in this instruction are written into the traffic light attribute field corresponding to "TL-2025-A1-001" in the database, updating the original "Quantity 1" to "Quantity 2", thus completing the attribute update of the traffic light in the real-scene 3D map.

[0106] In practical application, a smart map update project carried out high-precision map attribute updates for the "key area B1" in the core business district of City B. First, step S1041 was executed to extract the attribute data of the geographic elements in this area under time-varying conditions: 3 traffic signs, square type, 12-meter width of a six-lane road; the corresponding attribute data from the pre-stored real-scene 3D map was also analyzed: 2 traffic signs, square type, 12-meter road width; an attribute difference matrix was calculated using a matrix construction algorithm, resulting in a matrix with a difference of 1 for the number of traffic signs and 0 for other attributes. Next, step S1042 was executed, setting the allowable error range to 0 for the number of signs, and filtering out the traffic sign element; its unique identifier "SI-2025-B1-003" was queried, and the current timestamp "2025-09-07 18:30:00" was recorded; the updated attribute values ​​were calculated: number of signs = 2 + 1 = 3, type remains square. Finally, S1043 was executed, combining "SI-2025-B1-003", "2025-09-07 18:30:00", and "Quantity 3, Type Square" into an attribute update instruction, which was then placed into the instruction group corresponding to "Spatiotemporal Grid Unit C-20250907-18". Using a database batch write tool, the updated data was written to the corresponding attribute fields in the real-scene 3D map database, completing the attribute update for the traffic sign. Simultaneously, the project team performed the same process on other selected elements requiring updates within the area, such as the newly added section of blue guardrail. The attribute update for all geographic elements in "Key Area B1" was completed in just 2 hours, ensuring that the high-precision map data of the core business district of City B was consistent with the actual geographic environment.

[0107] In the overall scheme of step S104 above, by calculating the attribute difference matrix, the deviation between the temporal change state and the attribute of the real-world 3D map can be intuitively and accurately identified, providing a clear basis for subsequent updates. By filtering geographic elements that exceed the allowable error range and calculating the updated attribute values, the objects to be updated can be accurately identified, ensuring the accuracy of the updated data. By combining attribute update instructions and writing them to the database in batches according to spatiotemporal grid units, the efficiency of map attribute updates is greatly improved, avoiding the tedious operation of updating each element individually. The entire process, from difference identification to update execution, forms a precise and efficient high-precision map regional attribute update mechanism, effectively solving the problems of object ambiguity and low efficiency in traditional map updates, ensuring that the real-world 3D map can reflect the latest state of geographic elements in a timely manner, and providing reliable map data support for fields such as autonomous driving and intelligent transportation.

[0108] The following is a complete embodiment for steps S101 to S104:

[0109] like Figure 3As shown, a smart transportation technology company initiated a quarterly update to ensure the timeliness of high-precision maps for the approximately 80-kilometer-long ring road of City A. The project deployed 60 Brand B vehicle-mounted camera terminals, equipped with BeiDou positioning modules, 1080P high-definition cameras, and attitude sensors. Following a "two-way, four-lane segmented data collection" strategy, data was collected daily from 9:00-11:00 and 14:00-16:00 along the ring road. Each camera terminal generates one frame of real-world road image every 2 seconds, simultaneously outputting associated data including latitude and longitude accuracy of ±0.5 meters, shooting timestamp accuracy to milliseconds, and device tilt angle range of ±5°. The system assigns a unique identifier to each image frame using UUID encoding, merges latitude and longitude with the timestamp into a spatiotemporal identifier code embedded in the image metadata, and then encapsulates the attitude data to form an image data unit. The system then compares the image data units with a pre-stored 3D map of the A city ring expressway to calculate the positioning deviation. For example, if the original positioning of a certain image segment deviates from the map reference point by 1.2 meters, the deviation is reduced to 0.3 meters after correction. Simultaneously, a convolutional neural network is used to identify geographic elements such as road markings, emergency telephone poles, and guardrails, and these are assigned to spatiotemporal grid units of "0.0001 degrees latitude and longitude + 10 minutes" according to the corrected coordinates. Based on the distribution of elements within the grid, a feature density threshold is set to "a single grid containing more than 5 types of elements". Grids that continuously exceed the threshold are merged into key areas, such as the service area expansion section from K35 to K37. By retrieving historical image data units of this area over the past 30 days through time series analysis, the system extracts the change trajectory of service area entrance signs and the number of parking spaces, and reconstructs the temporal change status of "sign addition - parking space expansion - marking adjustment". Finally, by comparing the time-series changes with the pre-stored map, an attribute update instruction was generated to "add 3 entrance signs and update the number of parking spaces to 120". This instruction was then written into the map database in batches to complete the regional update of this road segment. The entire process took approximately 4 hours.

[0110] The high-precision map regionalization update method based on spatiotemporal correlation provided in this application ensures the spatiotemporal correlation of data by binding real-scene imagery with related data, providing a reliable foundation for subsequent analysis. Map comparison and coordinate calibration, along with the allocation of geographic elements to a spatiotemporal grid, improves the accuracy of element positioning and the standardization of management. Cluster analysis identifies key areas, and historical data is used to reconstruct temporal changes, accurately capturing dynamic changes in geographic elements. Regionalization updates are achieved based on difference comparisons, avoiding the waste of resources in overall updates and significantly improving the efficiency and accuracy of high-precision map updates, effectively meeting the timeliness and accuracy requirements of autonomous driving and intelligent transportation.

[0111] Figure 4 This application provides a schematic diagram of a specific implementation of a high-precision map regionalization update system based on spatiotemporal correlation, referring to... Figure 4 The system may include:

[0112] The association module 41 is used to acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers.

[0113] The allocation module 42 is used to compare the image data unit with a pre-stored real-scene 3D map, determine the corrected coordinates of satellite positioning information in the associated data according to the comparison results, simultaneously identify geographic feature characteristics from the image data unit, and allocate the geographic feature characteristics to the corresponding spatiotemporal grid unit according to the corrected coordinates of the satellite positioning information.

[0114] The reconstruction module 43 is used to perform spatial clustering analysis on the geographic feature features based on the distribution status of the spatiotemporal grid units, determine the key areas that need to be backtracked, obtain the image data units and geographic feature features corresponding to the key areas at different historical time points through time series retrieval and spatial location indexing, extract the change trajectory of each type of geographic feature in the time dimension, and reconstruct the temporal change status of each geographic feature in the spatiotemporal grid units of the key areas.

[0115] The update module 44 is used to update the regional attributes of the real-world 3D map based on the comparison results between the time-series change state and the pre-stored real-world 3D map.

[0116] The high-precision map regionalization update system based on spatiotemporal correlation in this application is used to implement the aforementioned high-precision map regionalization update method based on spatiotemporal correlation. Therefore, the specific implementation of the high-precision map regionalization update system based on spatiotemporal correlation can be found in the embodiment section of the high-precision map regionalization update method based on spatiotemporal correlation in the preceding text. The specific implementation can be referred to the description of the corresponding embodiments, which will not be repeated here.

[0117] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for executing the computer program to implement the steps of any of the above-described high-precision map regionalization update methods based on spatiotemporal correlation.

[0118] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of any of the above-described high-precision map regionalization update methods based on spatiotemporal correlation.

[0119] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.

[0120] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the embodiments of the high-precision map regionalization update method based on spatiotemporal correlation.

[0121] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0122] The above provides a detailed description of a high-precision map regionalization update method and system based on spatiotemporal correlation provided in this application. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.

Claims

1. A method for regional updating of high-precision maps based on spatiotemporal correlation, characterized in that, include: Acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers; The image data unit is compared with a pre-stored real-scene 3D map. Based on the comparison results, the corrected coordinates of the satellite positioning information in the associated data are determined. Geographic feature characteristics are simultaneously identified from the image data unit, and the geographic feature characteristics are assigned to the corresponding spatiotemporal grid unit based on the corrected coordinates of the satellite positioning information. Based on the distribution of the spatiotemporal grid units, spatial clustering analysis is performed on the geographic feature characteristics to identify key areas that need to be retrospectively analyzed. Through time series retrieval and spatial location indexing, the image data units and geographic feature characteristics corresponding to the key areas at different historical time points are obtained. The change trajectory of each type of geographic feature in the time dimension is extracted, and the temporal change status of each geographic feature in the spatiotemporal grid units of the key areas is reconstructed. Based on the comparison results between the temporal change state and the pre-stored real-scene 3D map, the real-scene 3D map is updated with regional attributes. Specifically, based on the distribution of the spatiotemporal grid units, spatial clustering analysis is performed on the geographic feature characteristics to identify key areas requiring retrospective analysis, including: Based on the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit, spatial clustering analysis is performed to calculate the feature density value of each spatiotemporal grid unit. The feature density value is compared with a preset density threshold. Spatiotemporal grid cells whose feature density value exceeds the preset density threshold are marked as high-density cells. Spatially continuous high-density cells are merged as key regions.

2. The method according to claim 1, characterized in that, By using time-series retrieval and spatial location indexing, the image data units and geographic feature characteristics of the key area at different historical time points are obtained. The change trajectory of each type of geographic feature in the time dimension is extracted, and the temporal change state of each geographic feature within the spatiotemporal grid unit of the key area is reconstructed, including: By using time series retrieval and spatial location indexing, the image data units and geographic feature characteristics of the key area at different historical time points are obtained; Extract the vector values ​​of each type of geographic element feature at different time points, calculate the difference between the values ​​of each feature vector, and arrange the difference values ​​in chronological order to form a change trajectory sequence; Based on the numerical differences in the change trajectory sequence, starting from the starting time point of the change trajectory sequence, the difference values ​​are sequentially superimposed onto the baseline state data of the key area to reconstruct the temporal change state of each geographic element.

3. The method according to claim 2, characterized in that, Based on the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid cells, spatial clustering analysis is performed to calculate the feature density value of each spatiotemporal grid cell, including: The quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit are converted into feature vectors, which contain quantity dimension and type dimension; The neighborhood range of each spatiotemporal grid cell is determined based on a preset distance radius, and the basic density value of each spatiotemporal grid cell is calculated based on the total number of geographic feature characteristics within the statistical cell. Based on the feature vectors of all spatiotemporal grid cells within the neighborhood, the distance weights of the neighborhood cells to the central cell are calculated, and based on the distance weights and the feature vectors of the neighborhood cells, the weighted density value of the central cell is calculated, and the weighted density value is used as the feature density value.

4. The method according to claim 2, characterized in that, The feature density values ​​are compared with a preset density threshold. Spatiotemporal grid cells whose feature density values ​​exceed the preset density threshold are marked as high-density cells. Spatially continuous high-density cells are merged into key regions, including: The feature density value of the spatiotemporal grid cell is obtained, wherein the spatiotemporal grid cell is obtained by dividing the target region according to a predefined spatial step size and time interval, and the feature density value is compared with a preset density threshold. Spatiotemporal grid cells whose feature density values ​​exceed the preset density threshold are marked as high-density cells. All spatiotemporal grid cells of the high-density cells are traversed to determine whether there are other high-density cells in the spatial neighborhood of the current high-density cell. If there are other high-density cells in the spatial neighborhood of the current high-density cell, then the current high-density cell will be merged with the other high-density cells in the spatial neighborhood to form a continuous merged cell. Repeat the above traversal, judgment and merging process until no new merges occur or all high-density cells are processed, and then identify each consecutive merged cell as the key region.

5. The method according to claim 1, characterized in that, The image data unit is compared with a pre-stored real-scene 3D map. Based on the comparison results, the corrected coordinates of the satellite positioning information in the associated data are determined. Simultaneously, geographic feature characteristics are identified from the image data unit, and the geographic feature characteristics are assigned to corresponding spatiotemporal grid units according to the corrected coordinates of the satellite positioning information. This includes: Extract the spatial coordinates from the image data unit and the coordinates of the corresponding reference points in the pre-stored real-scene 3D map, and calculate the total coordinate offset based on the difference between the spatial coordinates and the reference point coordinates on each coordinate axis; If the total coordinate offset exceeds a preset threshold, then corrected coordinates are generated based on the reference point coordinates, and the corrected coordinates replace the satellite positioning information in the associated data; Geographic feature characteristics are simultaneously identified from the image data unit through pixel analysis, and the feature vectors of the geographic feature characteristics are assigned to the corresponding spatiotemporal grid units according to the latitude and longitude values ​​of the corrected coordinates.

6. The method according to claim 1, characterized in that, Based on the comparison results between the temporal change state and the pre-stored real-scene 3D map, the real-scene 3D map is updated with regional attributes, including: Calculate the attribute difference matrix based on the attribute data in the time-series change state and the current attribute data of the pre-stored real-scene 3D map; Geographic features whose difference values ​​exceed the allowable error range are filtered from the attribute difference matrix. The unique identifier and current timestamp of the filtered geographic features are obtained, and the updated attribute values ​​are calculated based on the attribute difference matrix. The unique identifier, the current timestamp, and the updated attribute value are combined into an attribute update instruction, which is then grouped by spatiotemporal grid units and written in batches to the database of the real-scene 3D map to update the attribute fields.

7. The method according to claim 1, characterized in that, Acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers, including: The camera acquires real-scene image data and associated data generated during the shooting process from the shooting terminal, assigns a unique identifier to the real-scene image data, and binds the unique identifier to the associated data. Spatial coordinates, timestamps, attitude data, and relative distances are extracted from the bound associated data. The spatial coordinates and timestamps are merged to generate a spatiotemporal identifier code, and the spatiotemporal identifier code is embedded into the metadata area of ​​the real scene image data. The attitude data and relative distance are converted into a binary stream, and the real-scene image data containing spatiotemporal identification codes and the binary stream are encapsulated to form an image data unit with spatiotemporal location markings.

8. A high-precision map regionalization update system based on spatiotemporal correlation, characterized in that, include: The association module is used to acquire real-scene image data and associated data generated during the shooting process of the shooting terminal, and bind and associate the real-scene image data with the associated data to form an image data unit with spatiotemporal location markers; The allocation module is used to compare the image data unit with a pre-stored real-scene 3D map, determine the corrected coordinates of satellite positioning information in the associated data based on the comparison results, simultaneously identify geographic feature characteristics from the image data unit, and allocate the geographic feature characteristics to the corresponding spatiotemporal grid unit based on the corrected coordinates of the satellite positioning information. The reconstruction module is used to perform spatial clustering analysis on the geographic feature characteristics based on the distribution status of the spatiotemporal grid units, determine the key areas that need to be backtracked, obtain the image data units and geographic feature characteristics of the key areas at different historical time points through time series retrieval and spatial location indexing, extract the change trajectory of each type of geographic feature in the time dimension, and reconstruct the temporal change status of each geographic feature in the spatiotemporal grid units of the key areas. The update module is used to update the regional attributes of the real-world 3D map based on the comparison results between the time-series change state and the pre-stored real-world 3D map. Specifically, based on the distribution of the spatiotemporal grid units, spatial clustering analysis is performed on the geographic feature characteristics to identify key areas requiring retrospective analysis, including: Based on the quantity and type distribution of geographic feature characteristics in the spatiotemporal grid unit, spatial clustering analysis is performed to calculate the feature density value of each spatiotemporal grid unit. The feature density value is compared with a preset density threshold. Spatiotemporal grid cells whose feature density value exceeds the preset density threshold are marked as high-density cells. Spatially continuous high-density cells are merged as key regions.

9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the high-precision map regionalization update method based on spatiotemporal correlation as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the implementation of the high-precision map regionalization update method based on spatiotemporal correlation as described in any one of claims 1 to 7.

Images