A GIS buffer parameter inference method and system based on deep learning

Abstract

The application relates to the technical field of geographic information systems, in particular to a GIS buffer parameter inference method based on deep learning, which comprises the following steps: acquiring a natural language instruction input by a user; analyzing the natural language instruction to extract a geographic entity keyword and a fuzzy description word; determining an initial distance parameter through a pre-trained language-space mapping model based on the fuzzy description word; querying a spatial database based on the geographic entity keyword to acquire geographic entity attributes; correcting the initial distance parameter according to the geographic entity attributes to obtain a corrected distance parameter; using a field programmable gate array (FPGA) to perform parallel calculation on a spatial relationship matrix to generate an optimized buffer; and outputting the optimized buffer. The application introduces the FPGA to perform parallel calculation on the spatial relationship matrix, significantly improves the data processing speed and energy efficiency ratio, and realizes real-time generation of the buffer.

Description

Technical Field

[0001] This application relates to the technical field of geographic information systems, and in particular to a method and system for inferring GIS buffer parameters based on deep learning. Background Technology

[0002] Hardware acceleration technologies are becoming increasingly prevalent in high-performance GIS computing, especially the parallel application of FPGAs in spatial computation. Existing technologies mostly employ general-purpose GPUs to accelerate geographic computations, but these suffer from low energy efficiency and difficulty adapting to irregular spatial data flows. Some research has attempted to accelerate spatial indexing using FPGAs, but no dedicated designs have been developed for matrix computations and topology optimization during buffer generation.

[0003] While related technologies have proposed GPU-based parallel buffer generation methods that compute spatial relation matrices using CUDA kernels, these methods, while improving computational speed, suffer from high power consumption and cannot flexibly adapt to dynamic parameter adjustment processes.

[0004] Therefore, the existing technology has the following drawbacks: the lack of a dedicated FPGA architecture and low-power parallelization scheme for spatial relationship calculation in buffer generation, resulting in insufficient real-time performance and low energy efficiency. Summary of the Invention

[0005] To address at least one of the aforementioned technical problems, this application provides a method and system for inferring GIS buffer parameters based on deep learning.

[0006] Firstly, this application provides a GIS buffer parameter inference method based on deep learning, comprising the following steps:

[0007] Obtain natural language commands input by the user;

[0008] Parse the natural language instructions to extract geographic entity keywords and fuzzy descriptive terms;

[0009] Based on the fuzzy descriptive words, the initial distance parameters are determined using a pre-trained language-space mapping model;

[0010] Based on the geographic entity keywords, query the spatial database to obtain the geographic entity attributes;

[0011] Based on the geographic entity attributes, the initial distance parameters are corrected to obtain the corrected distance parameters;

[0012] Using a field-programmable gate array (FPGA), the spatial relation matrix is ​​computed in parallel to generate an optimized buffer.

[0013] Output the optimized buffer.

[0014] By adopting the above technical solution and introducing FPGA for parallel computation of spatial relationship matrix, the data processing speed and energy efficiency ratio are significantly improved, real-time generation of buffer is realized, and the problems of high computational latency and high power consumption of traditional software solutions are solved.

[0015] In one possible implementation, the step of using a field-programmable gate array (FPGA) to perform parallel computation of the spatial relation matrix includes:

[0016] The terrain elevation data is stored in the on-chip RAM cache and the spatial relationship matrix is ​​transmitted to the FPGA via the PCIe interface.

[0017] By adopting the above technical solutions, through on-chip RAM cache and high-speed PCIe interface, data transmission latency is reduced, computing efficiency is improved, and real-time processing capability is ensured.

[0018] In one possible implementation, the step of using a field-programmable gate array (FPGA) to perform parallel computation of the spatial relation matrix includes:

[0019] A pulsating array structure is used to perform convolution operations to optimize the calculation of topological constraint weights.

[0020] By adopting the above technical solutions, the pulsating array structure achieves efficient parallel processing of data streams, accelerates convolution operations, and improves the accuracy and speed of spatial relationship calculations.

[0021] In one possible implementation, the step of correcting the initial distance parameter based on the geographic entity attribute includes:

[0022] The calculation is based on a weighted average of semantic intensity correction, scale correction coefficient, and terrain correction coefficient, where each coefficient has a unified physical meaning and unit definition.

[0023] By adopting the above technical solution and clearly defining the physical meaning and unit of the coefficients, the interpretability and repeatability of parameter correction are enhanced, and the accuracy and consistency of buffer generation are improved.

[0024] In one possible implementation, the step of parsing the natural language instruction includes:

[0025] When natural language instructions contain multiple conflicting intentions, a constraint satisfaction model is used for multi-objective optimization, and the Pareto optimal solution is output as the corrected distance parameter.

[0026] By adopting the above technical solution and optimizing conflicting intentions through multi-objective optimization, an adaptive balance between semantics and spatial logic is achieved, generating a buffer that better suits user preferences.

[0027] In one possible implementation, the step of using a field-programmable gate array (FPGA) to perform parallel computation on the spatial relation matrix to generate an optimized buffer includes:

[0028] Terrain data is obtained based on a digital elevation model, and a three-dimensional buffer surface is generated using a slope-sensitive interpolation algorithm.

[0029] By adopting the above technical solution and introducing slope-sensitive interpolation, the actual distance accuracy of the buffer zone on the three-dimensional terrain is ensured, which significantly improves the buffer effect of complex terrains such as steep slopes and valleys.

[0030] One possible implementation also includes:

[0031] Use a graph neural network model to predict the probability of collisions at buffer boundaries;

[0032] The expansion parameters of the buffer are dynamically adjusted based on the collision probability.

[0033] By adopting the above technical solution, and using graph neural networks to predict conflicts in real time and adjust parameters, adaptive optimization of the buffer is achieved, reducing manual intervention and improving the reliability and intelligence level of the system.

[0034] Secondly, this application provides a GIS buffer parameter inference method system, comprising:

[0035] The instruction acquisition module is used to acquire natural language instructions input by the user.

[0036] The parsing module is used to parse the natural language instructions to extract geographic entity keywords and fuzzy descriptive terms;

[0037] The parameter determination module is used to determine the initial distance parameters based on the fuzzy descriptive words using a pre-trained language-space mapping model.

[0038] The query module is used to query the spatial database based on the geographic entity keywords to obtain geographic entity attributes;

[0039] The correction module is used to correct the initial distance parameter based on the geographic entity attributes to obtain the corrected distance parameter;

[0040] The FPGA processing module is used to perform parallel computation on the spatial relation matrix using a field-programmable gate array (FPGA) to generate an optimized buffer.

[0041] The output module is used to output the optimized buffer.

[0042] By adopting the above technical solution, the system integrates an FPGA processing module, which realizes efficient parallel computing of the spatial relationship matrix through hardware acceleration, greatly improving the real-time performance and energy efficiency of buffer generation, and is suitable for large-scale geographic data processing.

[0043] Thirdly, this application provides an electronic device including a memory and a processor, wherein the memory is used to store computer program code, and the processor is used to execute the computer program code stored in the memory to implement the methods in the first aspect and any one of the first aspects, or in the second aspect and any possible implementation of the second aspect.

[0044] Fourthly, this application provides a computer-readable storage medium storing a computer program or instructions that, when executed, implement the methods described in the first aspect and any one thereof, or the second aspect and any possible implementation thereof. Attached Figure Description

[0045] Figure 1 This is a flowchart illustrating a deep learning-based GIS buffer parameter inference method provided in an embodiment of this application.

[0046] Figure 2 This is a schematic diagram of the structure of a GIS buffer parameter inference system provided in an embodiment of this application.

[0047] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0048] The technical solutions in this application will now be described with reference to all the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them.

[0049] In the description of the embodiments of this application, unless otherwise stated, " / " means "or". For example, A / B can mean A or B. "And / or" in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Furthermore, in the description of the embodiments of this application, "plural" or "multiple" refers to two or more than two.

[0050] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this embodiment, unless otherwise stated, "a plurality of" means two or more.

[0051] The terminology used in the following embodiments is for the purpose of describing particular embodiments only and is not intended to be limiting of this application. As used in the specification and appended claims of this application, the singular expressions “a,” “an,” “the,” “the,” “the,” and “this” are intended to also include expressions such as “one or more,” unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of this application, “at least one” and “one or more” refer to one, two, or more than two.

[0052] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "one embodiment," "some embodiments," "another embodiment," "other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0053] This application provides a deep learning-based GIS buffer parameter inference method, executed by an electronic device. This electronic device can be a standalone physical electronic device, a cluster of multiple physical electronic devices, a distributed system, or a cloud electronic device providing cloud computing services. This application does not impose any limitations on this method. Figure 1 As shown, the method includes the following steps:

[0054] S100: Obtain natural language instructions input by the user.

[0055] Specifically, users input natural language commands containing geographic semantics through terminals such as graphical user interfaces or voice input devices. These commands are then transmitted to the command acquisition module of the central control host. This module includes a command verification unit that performs format verification and character encoding conversion on the input commands to ensure that the commands conform to encoding standards.

[0056] Furthermore, the natural language command acquisition process adopts a double buffering mechanism: the front-end buffer receives real-time input, and the back-end buffer performs command preprocessing, thereby ensuring that command data is not lost in high-concurrency scenarios.

[0057] Furthermore, the communication logic adopts a publish-subscribe pattern. When a new instruction arrives, the parsing module is automatically triggered to start, and the reliability of instruction transmission is ensured through a message queue.

[0058] Based on this, through multimodal input adaptation and redundancy design, the success rate of command reception is greatly improved, supporting the concurrent processing of hundreds of commands and providing a stable data source for subsequent processing; and it breaks through the limitation of the single input method of traditional GIS systems, significantly improving system compatibility and processing efficiency.

[0059] S200: Parse natural language instructions to extract geographic entity keywords and fuzzy descriptive terms.

[0060] Specifically, the parsing module of the central control host is based on a deep learning semantic parsing engine, which includes three core components: a lexical analyzer, a syntax analyzer, and a semantic annotator.

[0061] Among them, the lexical analyzer segments lexical units through a bidirectional long short-term memory network; the syntax analyzer constructs a syntax tree using dependency parsing; and the semantic annotation unit identifies geographic entity types and spatial relation words through a conditional random field model.

[0062] Furthermore, through part-of-speech tagging and named entity recognition, geographic entity keywords such as "highway" and fuzzy descriptive words such as "far away" are extracted.

[0063] Furthermore, the parsing process employs a multi-level verification mechanism. The first round of parsing obtains preliminary results, the second round performs semantic correction using a rule base, and the third round combines a geographic knowledge graph for entity disambiguation. This allows for the design of a divide-and-conquer strategy specifically for complex instructions, breaking them down into multiple simple clauses for separate processing.

[0064] Based on this, a multi-level verification and divide-and-conquer strategy is introduced into geographic semantic parsing, which significantly improves the accuracy of natural language parsing, especially in handling nested structures and ambiguous expressions. This provides reliable input for subsequent parameter mapping, effectively solves the problems of decomposing complex instructions and eliminating ambiguity, and achieves a qualitative improvement in semantic parsing accuracy.

[0065] In some embodiments, in order to achieve an adaptive balance between semantics and spatial logic and generate a buffer that better suits user preferences, S200 further includes the following steps:

[0066] S201. When a natural language instruction contains multiple conflicting intentions, a constraint satisfaction model is used for multi-objective optimization, and the Pareto optimal solution is output as the corrected distance parameter.

[0067] Specifically, the constraints satisfy the three objective optimization functions for model construction, namely minimizing semantic bias, minimizing scale mismatch, and minimizing terrain conflict.

[0068] Furthermore, the model employs an improved NSGA-III algorithm to solve for the Pareto front, using reference points to guide the solution set and maintain its diversity.

[0069] Furthermore, the constraints of the model include spatial topological relationships between geographic entities, domain rules, and user preference constraints.

[0070] Furthermore, the optimization process adopts a hierarchical solution strategy, namely, the first layer quickly selects feasible solution regions, the second layer refines the quality of the solution set, and the third layer selects the final solution based on user preferences.

[0071] Furthermore, the strategy supports an incremental optimization mechanism, whereby only the affected region is re-optimized when new constraints are added.

[0072] Based on this, by combining multi-objective optimization theory with geospatial constraint solving, an adaptive balance between semantics and spatial logic is achieved, the problem of resolving complex semantic conflicts is solved, intelligent decision-making is realized, and a buffer that is more in line with user preferences is generated.

[0073] In this embodiment, the method further includes the following steps:

[0074] S300: Based on fuzzy descriptive words, the initial distance parameters are determined through a pre-trained language-space mapping model.

[0075] Specifically, the language-space mapping model adopts a deep neural network architecture. Its input layer receives fuzzy descriptive words and geographic entity type embedding vectors, the hidden layer contains three fully connected layers, and the output layer outputs the absolute distance value and the scaling factor through a dual-channel structure.

[0076] The model pre-training uses a dataset of millions of geographic texts and spatial correspondences, and employs a contrastive learning strategy to optimize feature representation.

[0077] Furthermore, the model inference employs a dynamic routing mechanism, automatically selecting the output channel based on the geographic entity type. Point entities are preferentially output using absolute distance, while linear and areal entities are preferentially output using scale factors. It also supports online learning, continuously optimizing model parameters based on user feedback data.

[0078] Based on this, the model is designed with a dual-channel output structure to achieve accurate mapping from fuzzy semantics to spatial parameters, which greatly improves the matching accuracy of the test set, effectively eliminates the semantic gap in traditional methods, provides accurate basic parameters for buffer generation, and thus adapts to the parameter generation needs of different types of geographic entities, breaking through the limitations of a single output mode.

[0079] S400: Based on geographic entity keywords, query the spatial database to obtain geographic entity attributes.

[0080] Specifically, the spatial database adopts a distributed architecture, with the master node storing metadata indexes and the slave nodes storing entity attribute data.

[0081] Among them, entity attribute data includes multi-dimensional attributes such as the geometric features, scale attributes, topological relationships, and time-varying characteristics of geographic entities.

[0082] Furthermore, the query process implements a three-level caching strategy: the first level caches hot entity data, the second level caches region-related data, and the third level caches historical query results. Specifically, the LRU eviction algorithm is used to maintain cache validity, while a cache prefetching mechanism improves query efficiency.

[0083] Based on this, through the collaborative design of distributed architecture and multi-level caching mechanism, the average query response time is effectively shortened, supporting tens of thousands of concurrent queries per second, providing efficient data services for large-scale geographic applications, and solving the performance bottleneck problem of massive spatial data query.

[0084] S500. Based on the geographic entity attributes, the initial distance parameters are corrected to obtain the corrected distance parameters.

[0085] Specifically, the correction process is based on the correction module in the central control unit, which contains three parallel correction units: the semantic strength correction unit obtains the correction coefficient by querying the semantic strength dictionary; the scale correction unit calculates the scaling factor based on the ratio of the actual scale to the typical scale of the entity; and the terrain correction unit extracts slope and aspect features through the digital elevation model to generate terrain factors.

[0086] Furthermore, the correction process implements a feedback adjustment mechanism, which monitors the deviation between the correction result and the expected target in real time and dynamically adjusts the weight allocation of each correction unit.

[0087] Among them, a correction rule library is set up for special geographical scenarios such as geological disaster areas to ensure that the correction results comply with geographical constraints.

[0088] Based on this, the accuracy of distance parameter correction is improved by fusing and adjusting multi-dimensional correction factors, especially when dealing with complex geographic constraints. It provides accurate distance parameters for buffer generation, realizes adaptive optimization of distance parameters, and breaks through the accuracy limitations of traditional single correction methods.

[0089] In some embodiments, to enhance the interpretability and repeatability of parameter correction and improve the accuracy and consistency of buffer generation, S500 further includes the following steps:

[0090] S501. Weighted calculation is performed based on semantic strength correction, scale correction coefficient, and terrain correction coefficient.

[0091] Specifically, each coefficient has a unified physical meaning and unit definition.

[0092] The semantic strength correction value is obtained by querying a pre-built fuzzy semantic strength dictionary, which contains 2,000+ geospatial terms and their corresponding strength value ranges.

[0093] The size correction factor is based on the ratio of the actual size of an entity to the benchmark size of similar entities, and is normalized using a logarithmic function.

[0094] The terrain correction coefficient is generated by extracting multi-dimensional features such as slope, aspect, and curvature from the digital elevation model and combining them with a machine learning model.

[0095] Furthermore, all coefficients are represented by dimensionless standardized numerical values, and their physical meanings can be clearly defined as follows: the semantic intensity correction reflects the intensity of the user's intent, and this correction is 0.5-2.0; the scale correction coefficient characterizes the entity scale effect, and this coefficient is 0.1-10.0; the terrain correction coefficient reflects the degree of influence of terrain, and this coefficient is 0.3-3.0.

[0096] Furthermore, the weighted calculation employs an adaptive weight allocation algorithm, dynamically adjusting the weights of each coefficient based on the characteristics of the geographical scene.

[0097] Based on this, a multi-dimensional parameter standardization system is established, the physical meaning and unit of the coefficients are clearly defined, and the unified quantification and integration of semantic, scale and terrain factors are realized. This enhances the interpretability and repeatability of parameter correction, improves the accuracy and consistency of buffer generation in different geographical environments, and solves the problem of unclear parameter meaning in traditional methods.

[0098] In this embodiment, the method further includes the following steps:

[0099] The S600 uses a field-programmable gate array (FPGA) to perform parallel computation on the spatial relation matrix to generate an optimized buffer.

[0100] Specifically, the FPGA chip is equipped with a dedicated computing architecture, which includes a data preprocessing unit, a matrix calculation unit, and a post-processing unit.

[0101] The data preprocessing unit is responsible for the format conversion and block division of the spatial relationship matrix; the matrix calculation unit adopts a pulsating array structure, which contains 64 processing units for parallel calculation; and the post-processing unit is responsible for result aggregation and buffer boundary optimization.

[0102] Furthermore, the computation process employs a pipelined parallel strategy, dividing the spatial relation matrix into multiple sub-matrices and processing them continuously through a data flow-driven approach.

[0103] This strategy supports dynamic precision adjustment, allowing users to select single-precision or half-precision floating-point operations based on application requirements, thereby balancing computational precision and resource consumption.

[0104] Based on this, a dedicated FPGA architecture for spatial relationship computation is designed, which effectively improves the computing performance compared to the CPU solution and supports the real-time generation of buffers for complex geographic scenes, meeting the needs of high-concurrency GIS applications. By breaking through the computing performance bottleneck through hardware-level parallelization, a leapfrog improvement in buffer generation efficiency is achieved.

[0105] In some embodiments, in order to reduce data transmission latency, improve computational efficiency, and ensure real-time processing capabilities, S600 further includes the following steps:

[0106] S601 stores terrain elevation data in on-chip RAM cache and transmits the spatial relationship matrix to the FPGA via PCIe interface.

[0107] Specifically, after the terrain elevation data is read from the digital elevation model database, it is converted into a fixed-point format that can be processed by the data preprocessing unit of the central control host, and then written to the Block RAM cache array configured on the FPGA through the direct memory access controller.

[0108] The spatial relationship matrix is ​​stored in the main memory in a sparse matrix format and transferred to the FPGA's DDR controller via a PCIe 4.0 x16 interface using DMA.

[0109] Furthermore, the on-chip RAM adopts a multi-body parallel architecture, storing elevation data in blocks according to a spatial grid, supporting simultaneous read and write operations.

[0110] Furthermore, PCIe transmission adopts a descriptor chain DMA mode, supports multi-matrix parallel transmission and interrupt aggregation technology, which can minimize transmission overhead.

[0111] Furthermore, ECC verification and retransmission mechanisms are implemented during data transmission to ensure data integrity.

[0112] Based on this, through collaborative innovation of on-chip RAM cache and high-speed PCIe interface, data transmission latency is reduced, computing efficiency is improved, and reliable data supply is provided for real-time buffer generation.

[0113] In some embodiments, to improve the accuracy and speed of spatial relationship calculations, S600 further includes the following steps:

[0114] S602. A pulsating array structure is used to perform convolution operations to optimize the calculation of topological constraint weights.

[0115] Specifically, the FPGA internally features a configurable systolic processing array containing 8×8 processing units, each equipped with a multiply-accumulate unit and a local register file.

[0116] The array employs a two-dimensional grid interconnection structure, with data flowing along the row direction and weights broadcast along the column direction.

[0117] The topological constraint weight calculation uses an improved sliding window algorithm, supporting various convolutional kernel sizes from 3×3 to 7×7.

[0118] Furthermore, the pulsating array operates at a clock frequency of 500MHz, and can complete 64 multiply-accumulate operations per cycle.

[0119] The operation employs a dynamic precision adjustment mechanism, automatically selecting 16-bit or 32-bit floating-point operations based on the complexity of the topological relationship.

[0120] The array controller monitors data stream conflicts in real time and dynamically adjusts the data transmission timing.

[0121] Based on this, the pulsed computing architecture is applied to the field of geospatial computing, realizing efficient parallel processing of data streams, accelerating convolution operations, improving the accuracy and speed of spatial relationship calculations, and solving the performance bottleneck problem of complex topological relationship calculations through hardware-level parallelization, perfectly adapting to the needs of real-time GIS applications.

[0122] In some embodiments, to ensure the accuracy of the actual distance of the buffer zone on the three-dimensional terrain and significantly improve the buffer effect on complex terrains such as steep slopes and valleys, S600 includes the following steps:

[0123] S603: Obtain terrain data based on digital elevation model, and generate a three-dimensional buffer surface through slope-sensitive interpolation algorithm.

[0124] Specifically, topographic point cloud data is extracted from a high-precision digital elevation model to construct an irregular triangular network surface model.

[0125] Furthermore, the slope-sensitive interpolation algorithm can dynamically adjust the sampling density according to the surface curvature. Here, it can be specified that a 1-meter resolution sampling is used in areas with a slope greater than 25 degrees, and a 5-meter resolution sampling is used in flat areas.

[0126] Furthermore, the interpolation calculation can employ an improved radial basis function, introducing a slope weighting factor to enhance the interpolation accuracy in steep slope areas.

[0127] Specifically, the three-dimensional buffer can be generated using a normal offset algorithm, which involves projecting the two-dimensional buffer polygon onto the surface model and then offsetting it by a specified distance along the surface normal direction to generate the three-dimensional buffer.

[0128] Furthermore, in steep slope areas, a distance integration algorithm along the ground surface can be used to ensure the consistency of the actual buffer distance.

[0129] Furthermore, the algorithm supports multi-level detail rendering, dynamically adjusting surface details based on view distance.

[0130] Based on this, the slope-sensitive interpolation and surface offset algorithm are combined to ensure the accuracy of the actual distance of the buffer zone on the three-dimensional terrain, significantly improving the buffer effect of complex terrains such as steep slopes and valleys, and achieving a balance between accuracy and efficiency.

[0131] In this embodiment, the method further includes the following steps:

[0132] S700, Output Optimization Buffer.

[0133] Specifically, the output process is based on the output module of the central control host, which includes a data formatting unit, a quality inspection unit, and a transmission interface.

[0134] The data formatting unit converts buffer data into the standard GeoJSON format; the quality inspection unit verifies the topological consistency of the buffer and the accuracy of the spatial reference system; and the transmission interface supports both application programming interface and message queue output methods to adapt to different application scenarios.

[0135] Furthermore, the output process implements a multi-verification mechanism: geometric verification ensures that the buffer polygons are closed and have no self-intersections; attribute verification verifies the integrity of metadata; and topological verification ensures the consistency of relationships with surrounding geographic entities.

[0136] This mechanism supports incremental output mode, pushing the buffer generation status in real time.

[0137] Based on this, by combining standardized output format with multiple verification mechanisms, the accuracy of buffer output is greatly improved. The output data can be directly used for spatial analysis and visualization, ensuring the reliability and availability of buffer data and forming a complete processing loop.

[0138] In summary, by introducing FPGA for parallel computation of spatial relationship matrices, the data processing speed and energy efficiency are significantly improved, real-time buffer generation is achieved, and the problems of high computational latency and high power consumption in traditional software solutions are solved.

[0139] In some embodiments, in order to achieve adaptive optimization of the buffer, reduce manual intervention, and improve the reliability and intelligence of the system, this method further includes the following steps:

[0140] S800 uses a graph neural network model to predict the collision probability of buffer boundaries.

[0141] Specifically, the graph neural network model can adopt a gated graph attention network architecture, whose input is a spatial relationship graph of geographic entities. Its node features include entity type, scale attribute, time-varying characteristics, etc., and its edge features include distance, orientation, topological relationship, etc.

[0142] Furthermore, the network can contain three graph convolutional layers and one prediction layer, which outputs a collision probability value between 0 and 1.

[0143] Furthermore, its training data comes from a historical buffer zone conflict case library.

[0144] Furthermore, the network training can employ a multi-task learning framework, which can simultaneously predict conflict probability and conflict type.

[0145] Furthermore, the framework can employ graph data augmentation techniques to expand the training samples, thereby improving the model's generalization ability.

[0146] Furthermore, a subgraph sampling strategy is adopted in the inference phase to enable real-time processing of large-scale spatial graphs.

[0147] Based on this, graph neural networks are applied to geospatial conflict prediction, which improves the prediction accuracy of the model on the test set. This is a significant improvement over traditional rule-based methods. Deep learning breaks through the accuracy limitations of traditional methods and provides reliable predictions for buffer optimization.

[0148] S900: Dynamically adjust the expansion parameters of the buffer based on the collision probability.

[0149] Specifically, the dynamic adjustment system can use a fuzzy PID controller, whose input is the conflict probability value and output is the extended parameter adjustment amount.

[0150] Furthermore, the controller parameters can be adaptively tuned based on the conflict type and historical adjustment effects. Its adjustment strategy can employ a tiered response mechanism, fine-tuning parameters when the conflict probability is low and initiating major adjustments when the conflict probability is high.

[0151] Furthermore, the adjustment process can implement a predictive correction mechanism, which involves first predicting the effect after the adjustment, correcting the adjustment amount based on the prediction results, evaluating the actual effect after implementation, and updating the controller parameters.

[0152] Based on this, by integrating intelligent control theory with geospatial analysis across disciplines, adaptive optimization of buffer parameters is achieved, which reduces the occurrence rate of buffer conflicts, decreases the average time spent in the adjustment process, and significantly reduces the need for manual intervention, breaking through the limitations of traditional manual adjustment and improving the level of system automation.

[0153] In summary, by using graph neural networks to predict conflicts in real time and adjust parameters, adaptive optimization of the buffer is achieved, reducing manual intervention and improving the reliability and intelligence of the system.

[0154] The GIS buffer parameter inference system provided in the embodiments of this application will be described below. The GIS buffer parameter inference system described below can be referred to in correspondence with the deep learning-based GIS buffer parameter inference method described above.

[0155] refer to Figure 2 The GIS buffer parameter inference system adopts a hierarchical distributed architecture, with each module interconnected via a high-speed data bus, specifically including:

[0156] Instruction Acquisition Module 1 is used to acquire natural language instructions input by the user.

[0157] The instruction acquisition module 1 includes a multimodal interface unit and an instruction verification unit, which supports multiple input methods such as text and voice, and performs standardized processing on the input instructions.

[0158] Parsing module 2 is used to parse natural language instructions to extract geographic entity keywords and fuzzy descriptive terms.

[0159] The parsing module 2 includes a lexical analysis unit, a syntax analysis unit, and a semantic annotation unit. It uses a deep learning model to achieve accurate parsing of natural language instructions.

[0160] Parameter determination module 3 is used to determine the initial distance parameters based on fuzzy descriptive words and through a pre-trained language-space mapping model.

[0161] Among them, the parameter determination module 3 is equipped with a dual-channel output mechanism, including an absolute distance calculation unit and a scale factor generation unit, which can automatically select the output mode according to the geographic entity type.

[0162] Query module 4 is used to query the spatial database based on geographic entity keywords to obtain geographic entity attributes.

[0163] Among them, the query module 4 establishes a distributed index system, including spatial index units and attribute retrieval units, which supports efficient geographic entity attribute queries.

[0164] Correction module 5 is used to correct the initial distance parameters based on the geographic entity attributes to obtain the corrected distance parameters.

[0165] Among them, the correction module 5 adopts a multi-factor fusion architecture, including a semantic intensity correction unit, a scale correction unit, and a terrain correction unit. The standardized coefficients of each unit are then weighted and fused.

[0166] FPGA processing module 6 is used to perform parallel computation on the spatial relation matrix using a field-programmable gate array (FPGA) to generate an optimized buffer.

[0167] The FPGA processing module 6 includes a matrix calculation unit and a topology optimization unit, which realizes parallel processing of spatial relationship matrices through hardware acceleration.

[0168] Output module 7 is used to output the optimized buffer.

[0169] The output module 7 constructs a dual output channel, including a data formatting unit and a quality detection unit, to ensure that the output buffer conforms to geographic information standards.

[0170] Furthermore, during system operation, an event-driven mechanism can be adopted, with each module exchanging data through a message middleware, and subsequent modules being automatically triggered to start after the preceding module has finished processing.

[0171] In some embodiments, the system may further include three auxiliary processing units: a semantic conflict detection unit, responsible for identifying multi-target command conflicts; a three-dimensional terrain processing unit, for realizing elevation data fusion; and a dynamic adjustment unit, for optimizing parameters based on real-time feedback.

[0172] Furthermore, the data processing flow employs end-to-end monitoring, with each module equipped with a performance acquisition unit and an anomaly handling unit to ensure stable system operation.

[0173] Furthermore, the system supports an online update mechanism, allowing model parameters and rule bases to be dynamically loaded without interrupting service.

[0174] In summary, the system integrates an FPGA processing module, which accelerates the efficient parallel computation of the spatial relationship matrix through hardware, significantly improving the real-time performance and energy efficiency of buffer generation, making it suitable for large-scale geographic data processing.

[0175] This application provides an electronic device, such as... Figure 3 As shown, Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 The illustrated electronic device 300 includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, via a bus 302. Optionally, the electronic device 300 may also include a transceiver 304. It should be noted that in practical applications, the transceiver 304 is not limited to one type, and the structure of this electronic device 300 does not constitute a limitation on the embodiments of this application.

[0176] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in connection with the embodiments of this application. Processor 301 may also be a combination that implements computing functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0177] Bus 302 may include a pathway for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 302 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 3 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0178] The memory 303 may be a ROM (Read-Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or it may be an EEPROM (Electrically Erasable Programmable Read-Only Memory), a CD-ROM (Compact Disc Read-Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.

[0179] The memory 303 is used to store application code that executes the scheme of the embodiments of this application, and its execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the content shown in the foregoing method embodiments.

[0180] Among them, electronic devices include, but are not limited to: mobile terminals such as mobile phones, laptops, digital radio receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), and in-vehicle terminals (such as in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 3 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0181] This application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it implements the steps of the GIS buffer parameter inference method described above.

[0182] Since the embodiments of the computer-readable storage medium portion correspond to the embodiments of the method portion, please refer to the description of the embodiments of the method portion for the embodiments of the computer-readable storage medium portion.

[0183] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0184] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A method for inferring GIS buffer parameters based on deep learning, characterized in that, Includes the following steps: Obtain natural language commands input by the user; Parse the natural language instructions to extract geographic entity keywords and fuzzy descriptive terms; Based on the fuzzy descriptive words, an initial distance parameter is determined through a pre-trained language-space mapping model, wherein the language-space mapping model is a deep neural network model configured to output an absolute distance value or a scaling factor based on the fuzzy descriptive words and geographic entity types; Based on the geographic entity keywords, query the spatial database to obtain the geographic entity attributes; Based on the geographic entity attributes, the initial distance parameters are corrected to obtain the corrected distance parameters; Using a Field Programmable Gate Array (FPGA), the spatial relation matrix is ​​computed in parallel to generate an optimized buffer. This parallel computation of the spatial relation matrix using the FPGA includes: storing terrain elevation data in on-chip RAM cache and transmitting the spatial relation matrix to the FPGA via a PCIe interface; performing convolution operations using a pulsating array structure to optimize the calculation of topological constraint weights; acquiring terrain data based on a digital elevation model and generating a three-dimensional buffer surface using a slope-sensitive interpolation algorithm. Output the optimized buffer.

2. The method according to claim 1, characterized in that, The step of correcting the initial distance parameter based on the geographic entity attributes includes: The calculation is based on a weighted average of semantic intensity correction, scale correction coefficient, and terrain correction coefficient, where each coefficient has a unified physical meaning and unit definition.

3. The method according to claim 1, characterized in that, The step of parsing the natural language instruction includes: When natural language instructions contain multiple conflicting intentions, a constraint satisfaction model is used for multi-objective optimization, and the Pareto optimal solution is output as the corrected distance parameter.

4. The method according to claim 1, characterized in that, Also includes: Use a graph neural network model to predict the probability of collisions at buffer boundaries; The expansion parameters of the buffer are dynamically adjusted based on the collision probability.

5. A GIS buffer parameter inference system, characterized in that, include: The instruction acquisition module is used to acquire natural language instructions input by the user. The parsing module is used to parse the natural language instructions to extract geographic entity keywords and fuzzy descriptive terms; The parameter determination module is used to determine the initial distance parameter based on the fuzzy descriptive words and through a pre-trained language-space mapping model, wherein the language-space mapping model is a deep neural network model and is configured to output an absolute distance value or a scaling factor based on the fuzzy descriptive words and geographic entity types. The query module is used to query the spatial database based on the geographic entity keywords to obtain geographic entity attributes; The correction module is used to correct the initial distance parameter based on the geographic entity attributes to obtain the corrected distance parameter; An FPGA processing module is used to perform parallel computation on the spatial relation matrix using a field-programmable gate array (FPGA) to generate an optimized buffer. The FPGA processing module is configured to: store terrain elevation data in on-chip RAM cache and transmit the spatial relation matrix to the FPGA via a PCIe interface; perform convolution operations using a systolic array structure to optimize the calculation of topological constraint weights; acquire terrain data based on a digital elevation model and generate a three-dimensional buffer surface using a slope-sensitive interpolation algorithm. The output module is used to output the optimized buffer.

6. An electronic device, characterized in that, include: One or more processors; One or more memory units; And one or more computer programs, wherein the one or more computer programs are stored in the one or more memories, the one or more computer programs including instructions that, when executed by the one or more processors, cause the electronic device to perform the method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, The storage medium stores a program or instructions that, when executed, implement the method as described in any one of claims 1 to 4.

Images