Major language model-based map plotting method, apparatus and device, and medium

Abstract

The embodiment of the invention provides a map plotting method and device based on a large language model, equipment and a medium, and relates to the technical field of map plotting based on the large language model, and the method comprises the steps: obtaining raster image data and vector road data, so as to build a geographic space reference coordinate system; collecting a positioning information text corpus based on natural language description; building a spatial semantic processing model based on a self-attention mechanism according to the geographic space reference coordinate system and the positioning information text corpus; receiving a mixed instruction input by a user, and converting the mixed instruction into a structured instruction tree; generating a coordinate set according to the structured instruction tree, and generating a coordinate set scheme after verifying the coordinate set; plotting the coordinate set scheme into an interactive map based on a spatial semantic processing model; a user is supported to input a mixed data instruction, complex semantics in a natural language can be analyzed, map drawing is automatically carried out, and map drawing efficiency is improved.

Description

Technical Field

[0001] The present application relates to the technical field of map drawing based on a large language model, and in particular to a map drawing method, device, equipment and medium based on a large language model. Background Art

[0002] Mapping is the process of marking geographic information, target locations or dynamic events on a map in a graphical manner. It is widely used in military, emergency command, traffic planning, resource management and other fields. Traditional map-making systems have the following technical pain points: existing automated tools only support structured data input, cannot parse complex semantics in natural language, and rely on manual operations for symbol drawing and annotation, which is inefficient. Summary of the invention

[0003] The purpose of the embodiments of the present application is to provide a map drawing method, device, equipment and medium based on a large language model, so as to solve the problem that the existing map drawing method only supports structured data input, cannot parse the complex semantics in natural language, and relies on manual operation for symbol drawing and annotation, which is inefficient.

[0004] In a first aspect, an embodiment of the present application provides a map drawing method based on a large language model, the method comprising: Obtain raster image data and vector road data to establish a geographic spatial reference coordinate system; Collecting a text corpus of positioning information based on natural language description, wherein the text corpus of positioning information includes place name entities, location descriptions, and spatial relationship expressions; According to the geographic spatial reference coordinate system and the positioning information text corpus, a spatial semantic processing model is built based on the self-attention mechanism; Receive mixed instructions input by a user, and convert the mixed instructions into a structured instruction tree; Generate a coordinate set according to the structured instruction tree, and after verifying the coordinate set, generate a coordinate set solution; Based on the spatial semantic processing model, the coordinate set scheme is plotted as an interactive map.

[0005] In the above implementation process, raster image data and vector road data are obtained to establish a geographic spatial reference coordinate system; a positioning information text corpus based on natural language description is collected, wherein the positioning information text corpus contains place name entities, orientation descriptions and spatial relationship expressions; a spatial semantic processing model is built based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus; mixed instructions input by the user are received, and the mixed instructions are converted into a structured instruction tree; a coordinate set is generated according to the structured instruction tree, and after the coordinate set is verified, a coordinate set scheme is generated; based on the spatial semantic processing model, the coordinate set scheme is plotted as an interactive map; supporting users to input mixed data instructions, it is possible to parse complex semantics in natural language, automatically draw maps, and improve map drawing efficiency.

[0006] Furthermore, it also includes: Verify the conversion results of the interactive map and process the spatial semantic processing model based on the verification results; The step of verifying the conversion result of the interactive map and processing the spatial semantic processing model according to the verification result includes: Get the difference between the marked position of the interactive map and the actual measured value; If the difference exceeds the set threshold, the retraining process of the spatial semantic processing model is triggered; the retraining process includes: automatically labeling erroneous samples, adjusting the weight parameters of the spatial decoder, and updating the geographic entity knowledge graph; If the difference is less than or equal to the set threshold, the interactive map and the spatial semantic processing model are retained.

[0007] In the above implementation process, the interactive map is verified and feedback processed to achieve an adaptive error detection and model update closed loop.

[0008] Furthermore, the acquisition of raster image data and vector road data to establish a geographic spatial reference coordinate system includes: Obtain raster image data of the target area through the satellite remote sensing data interface; Parse the longitude and latitude coordinate system or access the open source geographic information database to obtain vector road network data; Convert raster image data and vector road data into data in a unified standard format, and fuse indexes to form a geographic spatial reference system.

[0009] In the above implementation process, a geographic spatial reference system in a unified standard format is established to facilitate the subsequent establishment of a spatial semantic processing model.

[0010] Furthermore, the spatial semantic processing model is built based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus, including: A text encoder is used to process the location information text corpus, an image encoder is used to parse the building outlines and road topology features in satellite images, and a spatial decoder is used to generate geometric elements in the geographic spatial reference coordinate system; The attention mechanism is used to achieve cross-modal feature alignment and obtain a spatial semantic processing model; The cross-modal feature alignment is achieved by using the attention mechanism, including: Establish the mapping relationship between the semantic elements in the text description and the spatial coordinates, and the corresponding relationship between the annotation entities and the plotted icons; Build dynamic spatial reasoning capabilities and optimize spatial reasoning capabilities through user feedback data based on reinforcement learning mechanisms.

[0011] In the above implementation process, a spatial semantic processing model is built based on the self-attention mechanism to automatically draw maps and improve map drawing efficiency.

[0012] Furthermore, the receiving of the mixed instruction input by the user and converting the mixed instruction into a structured instruction tree includes: Receiving a mixed instruction input by a user; wherein the mixed instruction includes: text, voice and sketch; Identify the spatial operation type, geographic references, spatial relationship descriptions, and operation objects in the instructions; Convert mixed instructions into a structured instruction tree.

[0013] In the above implementation process, users are supported to input mixed instruction data, parse the complex semantics in natural language, and convert it into a structured instruction tree.

[0014] Further, the generating of the coordinate set according to the structured instruction tree and the generating of the coordinate set solution after verifying the coordinate set include: Generate a coordinate set according to a structured instruction tree; Verify the validity of geographic entities in coordinate sets based on geographic spatial reference coordinate systems, solve the problem of homonymous entities, and perform conflict detection in combination with spatial semantic processing models; Generate multiple candidate coordinate set solutions with confidence scores; The conflict detection is performed by combining the spatial semantic processing model, including: The coordinate set is verified in combination with the spatial semantic processing model, wherein the verification of map information includes verifying the topological consistency between the newly annotated elements and the existing map elements, detecting the adaptability of the scale, and verifying the rationality of the element distribution.

[0015] In the above implementation process, multiple candidate coordinate set solutions with confidence scores are generated so that the coordinate set solutions can be screened according to the confidence scores.

[0016] Furthermore, the coordinate set scheme is plotted as an interactive map based on the spatial semantic processing model, including: Filter out the coordinate set solutions whose confidence scores are greater than the set score threshold; Based on the spatial semantic processing model, the coordinate set scheme is generated into a vector map file that meets the set standards; Overlay time dimension information on vector map files; Render 3D terrain effects on vector map files through the graphics library, and adjust lighting parameters and material parameters; Deploy differential privacy mechanisms for vector map files to blur the annotation information of sensitive areas; Get the interactive map.

[0017] In the above implementation process, interactive maps are automatically generated to improve map drawing efficiency.

[0018] In a second aspect, an embodiment of the present application further provides a map drawing device based on a large language model, the device comprising: A data acquisition module is used to acquire raster image data and vector road data to establish a geographic spatial reference coordinate system; A corpus collection module is used to collect a text corpus of positioning information based on natural language description, wherein the text corpus of positioning information includes place name entities, location descriptions and spatial relationship expressions; The model building module is used to build a spatial semantic processing model based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus; An instruction processing module, used for receiving mixed instructions input by a user and converting the mixed instructions into a structured instruction tree; A scheme generating module is used to generate a coordinate set according to the structured instruction tree, and generate a coordinate set scheme after verifying the coordinate set; The map drawing module is used to draw the coordinate set scheme into an interactive map based on the spatial semantic processing model.

[0019] In a third aspect, an embodiment of the present application provides an electronic device, including: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the map drawing method based on the large language model as described above.

[0020] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a server, the map drawing method based on the large language model as described above is implemented.

[0021] In a fifth aspect, an embodiment of the present invention provides a computer program product, which includes instructions, and when the instructions are executed by a computer, the computer implements the map drawing method based on the large language model as described above. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments of the present application will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present application and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0023] Figure 1 A flowchart of a map drawing method based on a large language model provided in an embodiment of the present application; Figure 2 A schematic diagram of a flow chart of a map drawing device based on a large language model provided in an embodiment of the present application; Figure 3 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0024] The technical solutions in the embodiments of the present application will be described below in conjunction with the drawings in the embodiments of the present application.

[0025] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. At the same time, in the description of this application, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0026] Traditional map marking systems have the following technical pain points: they rely on manual operations for symbol drawing and labeling, which is inefficient; existing automated tools only support structured data input and cannot parse complex semantics in natural language (such as "mark a temporary checkpoint 200 meters east of intersection A"); they lack intelligent optimization capabilities in dynamic environments, and marking conflicts require manual intervention. Based on this, an embodiment of the present application provides a map drawing method based on a large language model to solve the above problems.

[0027] Please see Figure 1 , Figure 1 A flowchart of a map drawing method based on a large language model is provided in an embodiment of the present application. The map drawing method based on a large language model includes: 100. Obtain raster image data and vector road data to establish a geographic spatial reference coordinate system.

[0028] Specifically, a unified geographic spatial reference coordinate system is established by establishing a multi-source data fusion index. Optionally, raster image data of the target area is obtained through a satellite remote sensing data interface; the latitude and longitude coordinate system is parsed or the open source geographic information database is accessed to obtain vector road network data; the raster image data and vector road data are converted into data in a unified standard format, and the index is fused to form a geographic spatial reference system; a geographic spatial reference system in a unified standard format is established to facilitate the subsequent establishment of a spatial semantic processing model.

[0029] 200. Collect a text corpus of positioning information based on natural language description, wherein the text corpus of positioning information includes place name entities, location descriptions, and spatial relationship expressions.

[0030] Specifically, in order to build a comprehensive and representative location information text corpus, data can be collected from multiple data sources; the collected raw text data often contains noise and redundant information, and strict data cleaning is required, including removing irrelevant advertisements, links, special symbols, etc., correcting spelling errors and grammatical errors, and unifying text formats. On the basis of data cleaning, place name entities, location descriptions, and spatial relationship expressions in the text are annotated.

[0031] It can be understood that place name entity is one of the core elements in the positioning information text, which covers geographical names of various scales; the orientation description is used to clarify the relative position relationship between geographical entities and is an important way to express spatial information in natural language; the expression of spatial relationship is more complex and diverse, which not only involves the orientation relationship between geographical entities, but also includes topological relationship, order relationship, etc.

[0032] For example, the directional description is used to clarify the relative position relationship between geographical entities, and is an important way to express spatial information in natural language. It can be achieved through a variety of words and phrases, such as using the basic directional words "east", "south", "west", "north" and their combinations, such as "southeast", "northwest", etc., to describe the relative direction between two places, such as "the school is on the east side of the park"; more specific directional phrases, such as "in front", "behind", "left", "right", "opposite", etc., can also be used to further refine the position relationship in combination with specific reference objects, such as "the supermarket is opposite the bookstore". In addition, there are some descriptions that express distance, such as "nearby", "not far away", "a few kilometers away", etc., which are used to illustrate the degree of spatial separation between geographical entities.

[0033] For example, in the expression of spatial relationships, topological relationships describe the spatial structural features such as connection, inclusion, and adjacency between geographic entities. For example, "a river runs through a city" reflects the crossing relationship between the river and the city, and "a community contains multiple buildings" expresses the inclusion relationship between the community and the buildings. Sequential relationships emphasize the spatial arrangement order of geographic entities, such as "from left to right are the library, teaching building, and gymnasium." These spatial relationship expressions can more comprehensively and accurately describe the spatial associations between geographic entities, and provide rich information for spatial positioning and spatial reasoning.

[0034] 300. Based on the geographic spatial reference coordinate system and positioning information text corpus, a spatial semantic processing model is built based on the self-attention mechanism.

[0035] Specifically, a text encoder is used to process the positioning information text corpus (natural language instructions), an image encoder is used to parse the building outlines and road topology features in satellite images, and a spatial decoder is used to generate geometric elements in the geographic spatial reference coordinate system; the attention mechanism is used to achieve cross-modal feature alignment and obtain a spatial semantic processing model.

[0036] Exemplarily, the positioning information text corpus usually contains various forms of natural language instructions. The word segmentation technology is used to divide the text into meaningful vocabulary units, perform part-of-speech tagging and named entity recognition, and clarify key information such as location, direction, and distance in the text. Select a suitable text encoder model and use the positioning information text corpus to fine-tune the encoder. During the training process, design appropriate task objectives, such as sequence labeling tasks (identifying location entities in text) and text classification tasks (determining the type of instructions, such as navigation instructions, location description instructions, etc.), so that the encoder can better understand the semantic features of the positioning information text. The trained text encoder can map the input positioning information text into a high-dimensional semantic vector space. Each dimension in the vector space represents the characteristics of the text at different semantic levels, such as location relevance, direction indication, distance, etc.

[0037] For example, satellite image data has the characteristics of high resolution and multi-spectral, but before directly inputting into the image encoder, a series of preprocessing operations are required. Including image correction (geometric correction, radiation correction) to eliminate geometric deformation and radiation error in the image; image enhancement (contrast enhancement, sharpening) to improve the visual effect and feature recognizability of the image. Convolutional neural network (CNN) is used as the basic architecture of the image encoder, and the image encoder is trained using a large-scale geospatial image dataset, which contains various types of geographical scenes, such as urban buildings, rural roads, natural landforms, etc. During the training process, a suitable loss function is designed, such as the cross entropy loss function (for image classification tasks) or the mean square error loss function (for image regression tasks), so that the encoder can accurately parse the building outline and road topology features in the satellite image. The trained image encoder can map the input satellite image into a feature map or feature vector. Each position in the feature map corresponds to a local area in the image, and its feature value represents the feature intensity of the area; the feature vector is a global feature representation of the entire image. These features include information such as the height, shape, texture of the building, and the direction, width, and connection relationship of the road.

[0038] Exemplarily, the feature representations output by the text encoder and the image encoder are in different semantic spaces and visual spaces respectively. In order to achieve effective spatial semantic fusion, these features need to be converted to a unified geographic spatial reference coordinate system. For example, a geographic coordinate system (such as WGS84) is used as a unified reference frame, and the image features and text features are spatially aligned through a coordinate conversion algorithm. The role of the spatial decoder is to generate geometric elements in the geographic spatial reference coordinate system, such as building outlines, road centerlines, etc., based on the fused feature representations. The trained spatial decoder can generate initial geometric elements based on the input fused features. However, the generated geometric elements may have some inaccuracies or incompleteness, which need to be further optimized. Post-processing algorithms, such as curve fitting and topological correction, can be used to smooth the generated geometric elements and optimize the topological structure to improve their accuracy and reliability.

[0039] Among them, the use of attention mechanism to achieve cross-modal feature alignment includes: establishing a mapping relationship between semantic elements in text descriptions and spatial coordinates, and a corresponding relationship between labeled entities and mapped icons; building dynamic spatial reasoning capabilities, and optimizing spatial reasoning capabilities through user feedback data based on a reinforcement learning mechanism.

[0040] For example, a spatial knowledge graph is constructed to organize spatial entities, relationships, attributes and other information in the form of a graph structure; nodes in the knowledge graph represent entities, and edges represent relationships between entities, such as adjacent relationships, inclusion relationships, functional association relationships, etc. Through the knowledge graph, structured representation and efficient reasoning of spatial information can be achieved. According to the laws of spatial cognition and domain knowledge, a series of reasoning rules are defined, for example, the reachability of an entity can be inferred based on its location relationship, and its interaction relationship with other entities can be inferred based on its functional attributes; these reasoning rules can be defined based on logical rules, probability models and other methods.

[0041] In addition, in the spatial reasoning system, define the appropriate state space, action space and reward function; select the reinforcement learning algorithm suitable for the spatial reasoning task, and use a large amount of user feedback data to train the reinforcement learning model.

[0042] 400. Receive a mixed instruction input by a user, and convert the mixed instruction into a structured instruction tree.

[0043] Specifically, receiving mixed instructions input by a user; wherein the mixed instructions include: text, voice and sketch; identifying the spatial operation type, geographic reference, spatial relationship description and operation object in the instructions; and converting the mixed instructions into a structured instruction tree.

[0044] Exemplarily, the following items are identified in the instruction: spatial operation type (create / modify / delete map features), geographic references (landmark buildings, POI points), spatial relationship descriptions (adjacent, contained, distance constraints), and operation objects (fire hydrants, houses, etc.).

[0045] 500. Generate a coordinate set according to the structured instruction tree, and after verifying the coordinate set, generate a coordinate set solution.

[0046] Specifically, a coordinate set is generated according to a structured instruction tree; the legitimacy of the geographic entity of the coordinate set is verified based on the geographic spatial reference coordinate system to solve the problem of homonymous locations (such as the coordinate matching of "Beijing Road" in different cities), and conflict detection is performed in combination with a spatial semantic processing model; and multiple candidate coordinate set solutions with confidence scores are generated.

[0047] As you can understand, the confidence score is a numerical indicator used to measure the reliability of the model's prediction results, usually between 0 and 1 (or 0% to 100%); the higher the score, the greater the model's confidence in the current prediction.

[0048] The performing of conflict detection in combination with the spatial semantic processing model includes: The coordinate set is verified in combination with the spatial semantic processing model. Verifying map information includes verifying the topological consistency between the newly annotated features and the existing map features, detecting the adaptability of the scale (such as the matching degree between the road width annotation and the current zoom level), and verifying the rationality of the feature distribution.

[0049] 600. Based on the spatial semantic processing model, the coordinate set scheme is plotted as an interactive map.

[0050] 610. Filter out coordinate set solutions whose confidence scores are greater than a set score threshold.

[0051] It can be understood that the coordinate sets with confidence scores greater than the set score threshold are screened out from a large number of coordinate data sets, for example, the coordinate data with confidence scores greater than 0.8 are screened out, and the screened coordinate sets are output in a suitable data format (such as a list, DataFrame, etc.) to facilitate subsequent processing and analysis.

[0052] 620. Based on the spatial semantic processing model, the coordinate set scheme is generated into a vector map file that meets the set standards.

[0053] Optionally, use a spatial semantic processing model (such as a deep learning model, spatial clustering algorithm, etc.) to extract features from the coordinate data, identify key points, lines, surfaces and other elements in the space, and semantically annotate the coordinate data based on the feature extraction results, such as annotating different types of geographic entities such as buildings, roads, and waters. Based on the results of spatial semantic processing, construct geometric elements (points, lines, and surfaces) in the vector map. For example, coordinate points of the same category are aggregated into polygons to represent buildings or areas; add attribute information to the geometric elements, such as name, type, area, etc., which can be extracted from the original data or the results of spatial semantic processing; use a map drawing library (such as Matplotlib, Folium, Geopandas, etc.) to draw the geometric elements and attribute information into a vector map.

[0054] 630. Overlay time dimension information on the vector map file.

[0055] Specifically, determine the type of time dimension information, expand the structure of the vector map file (add time attribute fields, use time format standards), and select a suitable data model to achieve the overlay of time dimension information.

[0056] 640. Render three-dimensional terrain effects on vector map files through the graphics library and adjust lighting parameters and material parameters.

[0057] Specifically, prepare vector map data, select a graphics library to render 3D terrain, adjust lighting parameters, adjust material parameters, add terrain textures and details, and finally optimize performance.

[0058] 650. Deploy a differential privacy mechanism for vector map files and blur the annotation information of sensitive areas.

[0059] In geographic information systems (GIS), vector map files (such as Shapefile, GeoJSON) often contain sensitive information (such as military bases, residents' privacy data, etc.). To protect privacy, sensitive areas need to be differentially privately processed to achieve fuzzification or disturbance. Fuzzification processing includes identifying sensitive areas, selecting fuzzification technology, achieving differential privacy (adding noise or disturbance), and generating fuzzified vector files.

[0060] 660. Get the interactive map.

[0061] It can be understood that based on the spatial semantic processing model, the coordinate set scheme is generated into a vector map file that conforms to the OGC standard (such as Shapefile / KML format); time dimension information is superimposed to support historical version comparison and change trajectory visualization; three-dimensional terrain effects are rendered through graphics libraries such as WebGL, and lighting and material parameters are dynamically adjusted; differential privacy mechanisms are deployed to fuzzy the annotation information of sensitive areas.

[0062] As described above, the embodiment of the present application obtains raster image data and vector road data to establish a geographic spatial reference coordinate system; collects a positioning information text corpus based on natural language descriptions, wherein the positioning information text corpus contains place name entities, orientation descriptions, and spatial relationship expressions; builds a spatial semantic processing model based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus; receives mixed instructions input by the user, and converts the mixed instructions into a structured instruction tree; generates a coordinate set according to the structured instruction tree, and generates a coordinate set scheme after verifying the coordinate set; based on the spatial semantic processing model, plots the coordinate set scheme as an interactive map; supports users to input mixed data instructions, can parse complex semantics in natural language, automatically draw maps, and improve map drawing efficiency.

[0063] 700. Verify the conversion result of the interactive map, and process the spatial semantic processing model according to the verification result.

[0064] 710. Obtain the difference between the marked position of the interactive map and the actual measurement value.

[0065] 720. If the difference exceeds the set threshold, the retraining process of the spatial semantic processing model is triggered; wherein the retraining process includes: automatically labeling erroneous samples, adjusting the weight parameters of the spatial decoder, and updating the geographic entity knowledge graph.

[0066] 730. If the difference is less than or equal to the set threshold, the interactive map and the spatial semantic processing model are retained.

[0067] It is understandable that by accessing the GNSS positioning system to obtain real information to build verification data, the difference between the marked position and the actual measurement value is quantified by a certain indicator during the test phase, such as the root mean square error. When the error exceeds the threshold, the model retraining process is triggered.

[0068] As mentioned above, the multimodal fusion mechanism of the embodiment of the present application realizes bidirectional mapping of text-spatial data. By constructing a multi-task learning model based on the Transformer architecture, integrating the text encoder, image encoder and spatial decoder, cross-modal feature alignment is achieved. The text encoder parses the semantic elements (such as place names, direction descriptions) in natural language instructions, the image encoder extracts spatial features (such as building outlines, road topology) in satellite images, and the spatial decoder dynamically maps the semantic elements to a unified geographic coordinate system through an attention mechanism to form a bidirectional data association. At the same time, a reinforcement learning mechanism is introduced to continuously optimize the matching accuracy of text descriptions and spatial coordinates using user feedback, effectively solving the problem of accurate conversion of complex semantics (such as "200 meters from the east gate") to geometric coordinates.

[0069] The dynamic spatial reasoning engine of the embodiment of the present application supports the processing of complex constraints. In the instruction parsing stage, the system identifies spatial relationships (including, distance constraints), geographic references and operation objects through a structured instruction tree, combines geographic entity legitimacy verification and homonymous heteronym resolution, and generates a candidate coordinate set; the spatial reasoning engine verifies the topological consistency of element distribution (such as the adjacency relationship between fire hydrants and roads), scale adaptability (matching degree between annotation size and map level) and operation conflicts (such as duplicate annotation detection) in real time, and uses the knowledge graph to dynamically load regional geographic rules (such as firefighting facility layout specifications) to achieve joint reasoning of multi-dimensional constraints, and finally outputs an optimization plan with confidence score to ensure that the annotation results meet the needs of actual business scenarios.

[0070] The embodiment of the present application introduces a security enhancement layer for geographic information privacy protection. In the mapping result output stage, the system uses a differential privacy mechanism to inject Gaussian noise or perform regional fuzzification processing on the coordinate data of sensitive areas (such as military facilities and personal addresses), and uses attribute generalization technology to desensitize metadata such as POI names. The data encryption module implements hierarchical permission control on the generated vector files (such as KML) to ensure that different user roles can only access geographic information at the authorized level. In addition, through dynamic watermark embedding and access log tracking, a full-link security protection system is constructed from data generation, transmission to storage, taking into account both geographic information sharing needs and privacy protection compliance.

[0071] The embodiments of the present application realize an adaptive closed loop of error detection and model update. A quantitative evaluation index is established by calculating the root mean square error (RMSE) between the GNSS measured data and the annotated coordinates. When the error exceeds the threshold, the closed-loop optimization process is automatically triggered: the error sample annotation module identifies the combination of semantic description and spatial features of the positioning deviation, the knowledge graph update module corrects the mapping relationship of geographic entity coordinates, and the spatial decoder dynamically adjusts the cross-modal attention weights through back propagation. This mechanism realizes an automated link from error detection, sample mining to model iteration, ensuring that the system continues to improve the accuracy of spatial reasoning when regional geographic data is updated (such as road reconstruction) or when user annotation patterns evolve.

[0072] The above steps are not to be performed in a strict order as described in the numbers, but should be understood as an overall solution.

[0073] In the second aspect, based on the above embodiment, the embodiment of the present application also provides a map drawing device based on a large language model, referring to Figure 2 The map drawing device based on the large language model provided in this embodiment specifically includes: a data acquisition module 201, a corpus collection module 202, a model building module 203, an instruction processing module 204, a solution generation module 205 and a map drawing module 206.

[0074] Among them, the data acquisition module 201 is used to acquire raster image data and vector road data to establish a geographic spatial reference coordinate system; the corpus acquisition module 202 is used to collect a positioning information text corpus based on natural language descriptions, wherein the positioning information text corpus contains place name entities, direction descriptions and spatial relationship expressions; the model building module 203 is used to build a spatial semantic processing model based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus; the instruction processing module 204 is used to receive mixed instructions input by the user and convert the mixed instructions into a structured instruction tree; the solution generation module 205 is used to generate a coordinate set according to the structured instruction tree, and generate a coordinate set solution after verifying the coordinate set; the map drawing module 206 is used to draw the coordinate set solution as an interactive map based on the spatial semantic processing model.

[0075] As described above, the embodiment of the present application obtains raster image data and vector road data to establish a geographic spatial reference coordinate system; collects a positioning information text corpus based on natural language descriptions, wherein the positioning information text corpus contains place name entities, orientation descriptions, and spatial relationship expressions; builds a spatial semantic processing model based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus; receives mixed instructions input by the user, and converts the mixed instructions into a structured instruction tree; generates a coordinate set according to the structured instruction tree, and generates a coordinate set scheme after verifying the coordinate set; based on the spatial semantic processing model, plots the coordinate set scheme as an interactive map; supports users to input mixed data instructions, can parse complex semantics in natural language, automatically draw maps, and improve map drawing efficiency.

[0076] In a third aspect, an embodiment of the present application further provides an electronic device, which can integrate the map drawing device based on the large language model with the user-mode polling mechanism provided in the embodiment of the present application. Figure 3 Schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3 The electronic device includes: an input device 43, an output device 44, a memory 42 and one or more processors 41; the memory 42 is used to store one or more programs; when the one or more programs are executed by the one or more processors 41, the one or more processors 41 implement the map drawing method based on the large language model of the user-mode polling mechanism provided in the above embodiment. The input device 43, the output device 44, the memory 42 and the processor 41 can be connected by a bus or other means. Figure 3 The example of connecting through bus is taken in the following.

[0077] The processor 41 executes various functional applications and data processing of the device by running the software programs, instructions and modules stored in the memory 42, that is, implements the map drawing method based on the large language model of the above-mentioned user-mode polling mechanism.

[0078] The electronic device provided above can be used to execute the map drawing method based on the large language model of the user-mode polling mechanism provided in the above embodiment, and has corresponding functions and beneficial effects.

[0079] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, the computer-readable storage medium including a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the map drawing method based on the large language model as described above, and the same beneficial effects can be achieved.

[0080] Of course, the storage medium containing computer executable instructions provided in an embodiment of the present application is not limited to the map drawing method based on a large language model as described above, and the computer executable instructions can also execute related operations in the map drawing method based on a large language model provided in any embodiment of the present application.

[0081] In a fifth aspect, the embodiments of the present application also provide a computer program product, and the methods described in the various embodiments of the present application can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer programs or instructions. When the computer program or instruction is loaded and executed on a computer, the processes or functions described in the various embodiments of the present application are executed in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, a network device, a user device, a core network device, an OAM (Open Application Model) or other programmable device.

[0082] The computer program or instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer program or instructions may be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired or wireless means. The computer-readable storage medium may be any available medium that can be accessed by a computer or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium, such as a floppy disk, a hard disk, or a magnetic tape; it may also be an optical medium, such as a digital video disk; it may also be a semiconductor medium, such as a solid-state drive. The computer-readable storage medium may be a volatile or non-volatile storage medium, or may include both volatile and non-volatile types of storage media.

[0083] In several embodiments provided in the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.

[0084] In addition, the functional modules in the various embodiments of the present application may be integrated together to form an independent part, or each module may exist separately, or two or more modules may be integrated to form an independent part.

[0085] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0086] The above description is only an embodiment of the present application and is not intended to limit the scope of protection of the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application. It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings.

[0087] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0088] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

Claims

1. A map drawing method based on a large language model, characterized in that: The method comprises: Obtain raster image data and vector road data to establish a geographic spatial reference coordinate system; Collecting a text corpus of positioning information based on natural language description, wherein the text corpus of positioning information includes place name entities, location descriptions, and spatial relationship expressions; According to the geographic spatial reference coordinate system and the positioning information text corpus, a spatial semantic processing model is built based on the self-attention mechanism; Receive mixed instructions input by a user, and convert the mixed instructions into a structured instruction tree; Generate a coordinate set according to the structured instruction tree, and after verifying the coordinate set, generate a coordinate set solution; Based on the spatial semantic processing model, the coordinate set scheme is plotted as an interactive map.

2. The map drawing method based on a large language model according to claim 1, characterized in that: Also includes: Verify the conversion results of the interactive map and process the spatial semantic processing model based on the verification results; The step of verifying the conversion result of the interactive map and processing the spatial semantic processing model according to the verification result includes: Get the difference between the marked position of the interactive map and the actual measured value; If the difference exceeds the set threshold, the retraining process of the spatial semantic processing model is triggered; the retraining process includes: automatically labeling erroneous samples, adjusting the weight parameters of the spatial decoder, and updating the geographic entity knowledge graph; If the difference is less than or equal to the set threshold, the interactive map and the spatial semantic processing model are retained.

3. The map drawing method based on a large language model according to claim 1, characterized in that: The obtaining of raster image data and vector road data to establish a geographic spatial reference coordinate system includes: Obtain raster image data of the target area through the satellite remote sensing data interface; Parse the longitude and latitude coordinate system or access the open source geographic information database to obtain vector road network data; Convert raster image data and vector road data into data in a unified standard format, and fuse indexes to form a geographic spatial reference system.

4. The map drawing method based on a large language model according to claim 1, characterized in that: The spatial semantic processing model is built based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus, including: A text encoder is used to process the location information text corpus, an image encoder is used to parse the building outlines and road topology features in satellite images, and a spatial decoder is used to generate geometric elements in the geographic spatial reference coordinate system; The attention mechanism is used to achieve cross-modal feature alignment and obtain a spatial semantic processing model; The cross-modal feature alignment is achieved by using the attention mechanism, including: Establish the mapping relationship between the semantic elements in the text description and the spatial coordinates, and the corresponding relationship between the annotation entities and the plotted icons; Build dynamic spatial reasoning capabilities and optimize spatial reasoning capabilities through user feedback data based on reinforcement learning mechanisms.

5. The map drawing method based on a large language model according to claim 1, characterized in that: The receiving of the mixed instruction input by the user and converting the mixed instruction into a structured instruction tree includes: Receiving a mixed instruction input by a user; wherein the mixed instruction includes: text, voice and sketch; Identify the spatial operation type, geographic references, spatial relationship descriptions, and operation objects in the instructions; Convert mixed instructions into a structured instruction tree.

6. The map drawing method based on a large language model according to claim 1, characterized in that: The generating of the coordinate set according to the structured instruction tree and the generating of the coordinate set scheme after verifying the coordinate set include: Generate a coordinate set according to a structured instruction tree; Verify the validity of geographic entities in coordinate sets based on geographic spatial reference coordinate systems, solve the problem of homonymous entities, and perform conflict detection in combination with spatial semantic processing models; Generate multiple candidate coordinate set solutions with confidence scores; The conflict detection is performed by combining the spatial semantic processing model, including: The coordinate set is verified in combination with the spatial semantic processing model, wherein the verification of map information includes verifying the topological consistency between the newly annotated elements and the existing map elements, detecting the adaptability of the scale, and verifying the rationality of the element distribution.

7. The map drawing method based on a large language model according to claim 1, characterized in that: The method of plotting the coordinate set scheme as an interactive map based on the spatial semantic processing model includes: Filter out the coordinate set solutions whose confidence scores are greater than the set score threshold; Based on the spatial semantic processing model, the coordinate set scheme is generated into a vector map file that meets the set standards; Overlay time dimension information on vector map files; Render 3D terrain effects on vector map files through the graphics library, and adjust lighting parameters and material parameters; Deploy differential privacy mechanisms for vector map files to blur the annotation information of sensitive areas; Get the interactive map.

8. A map drawing device based on a large language model, characterized in that: The device comprises: A data acquisition module is used to acquire raster image data and vector road data to establish a geographic spatial reference coordinate system; A corpus collection module is used to collect a text corpus of positioning information based on natural language description, wherein the text corpus of positioning information includes place name entities, location descriptions and spatial relationship expressions; The model building module is used to build a spatial semantic processing model based on the self-attention mechanism according to the geographic spatial reference coordinate system and the positioning information text corpus; An instruction processing module, used for receiving mixed instructions input by a user and converting the mixed instructions into a structured instruction tree; A scheme generating module is used to generate a coordinate set according to the structured instruction tree, and generate a coordinate set scheme after verifying the coordinate set; The map drawing module is used to draw the coordinate set scheme into an interactive map based on the spatial semantic processing model.

9. An electronic device, characterized in that: include: A processor, a memory and a bus, wherein the processor is connected to the memory via the bus, and the memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, they are used to implement the map drawing method based on a large language model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a server, implements the map drawing method based on a large language model as described in any one of claims 1 to 7.

Images