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

Abstract

The invention relates to the technical field of geographic information systems, in particular to a GIS buffer zone parameter inference method based on deep learning, and the method comprises the following steps: obtaining a natural language instruction input by a user; analyzing the natural language instruction to extract a geographic entity keyword and a fuzzy description word; based on the fuzzy description word, determining an initial distance parameter through a pre-trained language-space mapping model; querying a spatial database based on the geographic entity keyword to obtain geographic entity attributes; according to the geographic entity attribute, correcting the initial distance parameter to obtain a corrected distance parameter; a field programmable gate array (FPGA) is used for carrying out parallel calculation on the spatial relation matrix to generate an optimized buffer area; and outputting the optimized buffer area. According to the method, the FPGA is introduced to carry out parallel calculation of the spatial relation matrix, so that the data processing speed and the energy efficiency ratio are remarkably improved, and the real-time generation of the buffer area is realized.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of geographic information systems, and in particular to a GIS buffer parameter inference method and system based on deep learning. BACKGROUND

[0002] Hardware acceleration technology is gradually popularized in GIS high-performance computing, especially the parallel application of FPGA in spatial operation. Existing technologies mostly use general-purpose GPUs to accelerate geographic computing, but the energy efficiency ratio is low and it is difficult to adapt to irregular spatial data flow. Some research attempts to realize spatial index acceleration through FPGA, but does not make special design for matrix calculation and topology optimization in buffer generation.

[0003] Although a GPU-based buffer parallel generation method is proposed in the related art, which calculates the spatial relationship matrix through the CUDA kernel. This method improves the calculation speed, but has high power consumption and cannot flexibly adapt to dynamic parameter adjustment processes.

[0004] Therefore, the existing technology has the following defects: there is a lack of special FPGA architecture and low-power parallelization scheme for spatial relationship calculation in buffer generation, resulting in insufficient real-time performance and low energy efficiency ratio. SUMMARY

[0005] To solve at least one of the above technical problems, the application provides a GIS buffer parameter inference method and system based on deep learning.

[0006] In a first aspect, the application provides a GIS buffer parameter inference method based on deep learning, comprising the following steps: obtaining a natural language instruction input by a user; parsing the natural language instruction to extract geographical entity keywords and fuzzy description words; determining an initial distance parameter based on the fuzzy description words through a pre-trained language-space mapping model; querying a spatial database based on the geographical entity keywords to obtain geographical entity attributes; correcting the initial distance parameter according to the geographical 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; outputting the optimized buffer.

[0007] By adopting the above technical solution, the FPGA is introduced to perform parallel calculation on the spatial relationship matrix, which significantly improves the data processing speed and energy efficiency ratio, realizes real-time generation of the buffer, and solves the problems of high calculation delay and high power consumption of the traditional software solution.

[0008] In a possible implementation, the step of performing parallel computation on the spatial relationship matrix using the field programmable gate array (FPGA) includes: The terrain elevation data is stored in an on-chip RAM cache, and the spatial relationship matrix is transmitted to the FPGA through a PCIe interface.

[0009] By using the above technical solutions, the data transmission delay is reduced, the computing efficiency is improved, and the real-time processing capability is ensured through the on-chip RAM cache and the high-speed PCIe interface.

[0010] In a possible implementation, the step of performing parallel computation on the spatial relationship matrix using the field programmable gate array (FPGA) includes: The convolution operation is performed using a systolic array structure to optimize the calculation of the topological constraint weight.

[0011] By using the above technical solutions, the systolic array structure realizes efficient parallel processing of data streams, accelerates the convolution operation, and improves the accuracy and speed of spatial relationship calculation.

[0012] In a possible implementation, the step of correcting the initial distance parameter according to the geographic entity attribute includes: The weighted calculation is performed based on the semantic intensity correction amount, the scale correction coefficient, and the terrain correction coefficient, wherein each coefficient has a unified physical meaning and unit definition.

[0013] By using the above technical solutions, the physical meaning and unit of the coefficient are defined, the explainability and repeatability of parameter correction are enhanced, and the accuracy and consistency of the buffer generation are improved.

[0014] In a possible implementation, the step of analyzing the natural language instruction includes: When the natural language instruction contains multiple conflicting intentions, a constraint satisfaction model is used for multi-objective optimization, and a Pareto optimal solution is output as a corrected distance parameter.

[0015] By using the above technical solutions, the conflicting intentions are processed through multi-objective optimization, the adaptive balance between semantics and spatial logic is realized, and a buffer that is more in line with user preferences is generated.

[0016] In a possible implementation, the step of performing parallel computation on the spatial relationship matrix using the field programmable gate array (FPGA) to generate an optimized buffer includes: Terrain data is obtained based on a digital elevation model, and a three-dimensional buffer surface is generated through a slope-sensitive interpolation algorithm.

[0017] By adopting the above technical solution and introducing slope-sensitive interpolation, the actual distance accuracy of the buffer zone on three-dimensional terrain is ensured, and the buffer zone effect on complex terrains such as steep slopes and valleys is significantly improved.

[0018] In one possible implementation, it also includes: Use a graph neural network model to predict the conflict probability of buffer zone boundaries; The expansion parameters of the buffer zone are dynamically adjusted according to the conflict probability.

[0019] By adopting the above technical solution, conflicts are predicted and parameters are adjusted in real time through graph neural networks, which achieves adaptive optimization of the buffer zone, reduces manual intervention, and improves the reliability and intelligence level of the system.

[0020] In a second aspect, the present application provides a GIS buffer parameter inference method system, comprising: An instruction acquisition module is used to obtain natural language instructions input by the user; A parsing module, configured to parse the natural language instructions to extract geographic entity keywords and fuzzy descriptive words; A parameter determination module, configured to determine an initial distance parameter based on the fuzzy description word through a pre-trained language-space mapping model; A query module, configured to query a spatial database based on the geographic entity keywords to obtain geographic entity attributes; A correction module, configured to correct the initial distance parameter according to the geographic entity attribute to obtain a corrected distance parameter; an FPGA processing module, configured to perform parallel calculations on the spatial relationship matrix using a field programmable gate array (FPGA) to generate an optimized buffer; An output module is used to output the optimized buffer.

[0021] By adopting the above technical solution, the system integrates the FPGA processing module and realizes efficient parallel calculation of the spatial relationship matrix through hardware acceleration, which greatly improves the real-time performance and energy efficiency of buffer generation and is suitable for large-scale geographic data processing.

[0022] In a third aspect, the present application provides an electronic device comprising 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 method in the above-mentioned first aspect and any one of the first aspect, or the above-mentioned second aspect and any possible implementation of the second aspect.

[0023] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program or instructions, when the computer program or instructions are executed, the method in the first aspect and any possible implementation of the first aspect, or the second aspect and any possible implementation of the second aspect is implemented. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1 A flowchart of a GIS buffer parameter inference method based on deep learning provided by an embodiment of the present application.

[0025] Figure 2 A structural diagram of a GIS buffer parameter inference system provided by an embodiment of the present application.

[0026] Figure 3 A structural diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0027] The technical solutions in the present application will be described below with reference to all the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all the embodiments.

[0028] In the description of the embodiments of the present application, unless otherwise specified, " / " represents the meaning of "or", for example, A / B can represent A or B; "and / or" in this document is only a description of the association relationship of the associated objects, which means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B together, and B alone. In addition, in the description of the embodiments of the present application, "plurality" or "multiple" means two or more than two.

[0029] Hereinafter, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include one or more features. In the description of the embodiments, unless otherwise specified, the meaning of "multiple" is two or more than two.

[0030] The terms used in the following embodiments are only for the purpose of describing specific embodiments and are not intended to be limiting on the present application. As used in the specification and the appended claims of the present application, the singular forms "a", "an" and "the" are intended to include, for example, the expression "one or more", unless the context clearly indicates otherwise. It should also be understood that in the following embodiments of the present application, "at least one" and "one or more" mean one, two or more than two.

[0031] References to "one embodiment" or "some embodiments" in this specification mean that a particular feature, structure, or characteristic described in conjunction with that embodiment is included in one or more embodiments of the present application. Thus, phrases such as "one embodiment," "some embodiments," "another embodiment," and "other embodiments" appearing in various places in this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically stated. The terms "including," "comprising," "having," and variations thereof mean "including but not limited to," unless otherwise specifically stated.

[0032] The embodiment of the present application provides a GIS buffer parameter inference method based on deep learning, which is executed by an electronic device, wherein the electronic device can be an independent physical electronic device, or an electronic device cluster or distributed system composed of multiple physical electronic devices, or a cloud electronic device providing cloud computing services. The embodiment of the present application is not limited here, such as Figure 1 As shown, the method includes the following steps: S100: Obtain a natural language instruction input by a user.

[0033] 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 verifies the format of the input commands and converts the character encoding to ensure that the commands comply with the encoding specifications.

[0034] Furthermore, the acquisition process of natural language instructions adopts a double buffering mechanism, with the foreground buffer receiving real-time input and the background buffer performing instruction preprocessing, thereby ensuring that instruction data is not lost in high-concurrency scenarios.

[0035] Furthermore, the communication logic adopts a publish-subscribe model. When a new instruction arrives, it automatically triggers the start of the parsing module and ensures the reliability of instruction transmission through the message queue.

[0036] Based on this, through multimodal input adaptation and redundant design, the success rate of command reception has been greatly improved, supporting the concurrent processing of hundreds of commands, providing a stable data source for subsequent processing; and breaking through the limitations of the single input method of traditional GIS systems, significantly improving system compatibility and processing efficiency.

[0037] S200: Parse natural language instructions to extract geographic entity keywords and fuzzy description words.

[0038] Specifically, the parsing module of the central control host is based on a deep learning semantic parsing engine and includes three core components: a lexical analyzer, a syntactic analyzer, and a semantic tagger.

[0039] Among them, the lexical analyzer segments the lexical units through a bidirectional long short-term memory network; the syntax analyzer adopts dependency syntax analysis to construct a syntax tree; and the semantic tagger identifies geographical entity types and spatial relationship words through a conditional random field model.

[0040] Further, through part-of-speech tagging and named entity recognition, geographical entity keywords such as "expressway" and ambiguous description words such as "far away" are extracted.

[0041] Further, the parsing process adopts a multi-level verification mechanism, the first round of parsing obtains a preliminary result, the second round performs semantic correction through a rule base, and the third round performs entity disambiguation in combination with a geographical knowledge graph. Furthermore, a divide-and-conquer strategy can be designed specifically for complex instructions, which decomposes complex instructions into multiple simple clauses for separate processing.

[0042] Based on this, multi-level verification and the divide-and-conquer strategy are introduced into geographical semantic parsing, which greatly 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 complex instruction decomposition and ambiguity elimination, and realizes the qualitative improvement of semantic parsing accuracy.

[0043] In some embodiments, in order to achieve adaptive balance of semantics and spatial logic and generate a buffer that is more in line with user preferences, S200 further includes the following steps: S201, when the natural language instruction contains multiple conflicting intents, a constraint satisfaction model is used for multi-objective optimization, and a Pareto optimal solution is output as a revised distance parameter.

[0044] Specifically, the constraint satisfaction model constructs a three-objective optimization function, which respectively minimizes semantic deviation, minimizes scale mismatch degree, and minimizes topological conflict degree.

[0045] Further, the model uses an improved NSGA-III algorithm to solve the Pareto front, and maintains the diversity of the solution set through reference point guidance.

[0046] Further, the constraints of the model include spatial topological relationships between geographical entities, domain rules, and user preference constraints.

[0047] Further, the optimization process adopts a hierarchical solution strategy, which respectively selects a feasible solution region in the first layer, fine-tunes the solution set quality in the second layer, and selects the final solution based on user preferences in the third layer.

[0048] Further, the strategy supports an incremental optimization mechanism, when new constraints are added, only the affected area is re-optimized.

[0049] Based on this, the multi-objective optimization theory is combined with the geographical space constraint solution to realize the adaptive balance of semantics and spatial logic, solve the complex semantic conflict resolution problem, realize the intelligent decision-making, and generate a buffer zone that is more in line with the user's preference.

[0050] In this embodiment, the method further comprises the following steps: S300, based on the fuzzy description word, determining an initial distance parameter through a pre-trained language-space mapping model.

[0051] Specifically, the language-space mapping model adopts a deep neural network architecture, the input layer of which receives the fuzzy description word and the geographical entity type embedding vector, the hidden layer of which contains three fully connected layers, and the output layer of which outputs the absolute distance value and the scale factor through a double-channel structure respectively.

[0052] Among them, the model pre-training uses a million-level geographical text and space corresponding dataset, and adopts a contrast learning strategy to optimize the feature representation.

[0053] Further, the model inference adopts a dynamic routing mechanism to automatically select the output channel according to the geographical entity type. Among them, the point entity preferentially uses the absolute distance output, and the line and surface entities preferentially use the scale factor output. And support online learning function, through user feedback data to continuously optimize the model parameters.

[0054] Based on this, the model designs a double-channel output structure to realize the accurate mapping of fuzzy semantics to spatial parameters, greatly improves the matching accuracy on the test set, effectively eliminates the semantic gap existing in the traditional method, provides accurate basic parameters for buffer zone generation, and further adapts to the parameter generation needs of different types of geographical entities, breaking through the limitations of single output mode.

[0055] S400, based on the geographical entity keyword, querying the spatial database to obtain the geographical entity attribute.

[0056] Specifically, the spatial database adopts a distributed architecture, and the master node stores metadata index and the slave node stores entity attribute data.

[0057] Among them, the entity attribute data includes the geometric characteristics, scale attribute, topological relationship and time-varying characteristics of the geographical entity and other multi-dimensional attributes.

[0058] Further, the query process implements a three-level cache strategy, the first level cache hot entity data, the second level cache regional associated data, and the third level cache historical query results. Among them, the LRU eviction algorithm is used to maintain the cache effectiveness, and at the same time, the cache prefetch mechanism is used to improve the query efficiency.

[0059] Based on this, through the collaborative design of distributed architecture and multi-level cache mechanism, the average query response time is effectively shortened, supporting 10,000 concurrent queries per second, providing efficient data services for large-scale geographic applications, and solving the performance bottleneck problem of massive spatial data query.

[0060] S500, according to the geographical entity attribute, the initial distance parameter is corrected to obtain the corrected distance parameter.

[0061] Specifically, the correction process is based on a correction module in the central control host, which includes three parallel processing correction units. The semantic intensity correction unit obtains the correction coefficient by querying the semantic intensity dictionary. The scale correction unit calculates the scaling coefficient based on the ratio of the actual size of the entity to the typical size. The terrain correction unit extracts slope, slope direction features through digital elevation model, and generates terrain factors.

[0062] Further, the correction process implements a feedback adjustment mechanism that monitors the deviation of the correction result from the expected target in real time and dynamically adjusts the weight distribution of each correction unit.

[0063] Among them, the correction rule base is set for special geographic scenes such as geological disaster areas to ensure that the correction result meets the geographical constraints.

[0064] Based on this, through the fusion and feedback adjustment of multi-dimensional correction factors, the accuracy of distance parameter correction is improved, especially in handling complex geographic constraint scenarios. It performs well and provides accurate distance parameters for buffer generation, realizes adaptive optimization of distance parameters, and breaks through the precision limit of traditional single correction method.

[0065] In some embodiments, to enhance the explainability and repeatability of parameter correction and improve the accuracy and consistency of buffer generation, S500 further includes the following steps: S501, based on the semantic intensity correction amount, the scale correction coefficient and the terrain correction coefficient, weighted calculation is performed.

[0066] Specifically, each coefficient has a unified physical meaning and unit definition.

[0067] Among them, the semantic intensity correction amount is obtained by querying the pre-constructed fuzzy semantic intensity dictionary, which contains 2000+ geographic spatial terms and their corresponding intensity value intervals.

[0068] The scale correction coefficient is based on the ratio of the actual size of the entity to the reference size of the same type of entity, and is normalized by using a logarithmic function.

[0069] The terrain correction coefficient extracts multi-dimensional features such as slope, slope direction, curvature, etc. through digital elevation model, and is generated in combination with machine learning model.

[0070] Further, each coefficient adopts a dimensionless standardized numerical value, and its physical meaning can be explicitly defined as follows: the semantic intensity correction reflects the intensity of user intention, and the correction takes a value of 0.5-2.0; the scale correction coefficient represents the scale effect of the entity, and the coefficient takes a value of 0.1-10.0; and the terrain correction coefficient reflects the degree of topographic influence, and the coefficient takes a value of 0.3-3.0.

[0071] Further, the weighted calculation adopts an adaptive weight distribution algorithm to dynamically adjust the weight of each coefficient according to the characteristics of the geographic scene.

[0072] Based on this, a multi-dimensional parameter standardization system is established to explicitly define the physical meaning and unit of the coefficient, realize the unified quantification and fusion of semantic, scale and terrain factors, enhance the interpretability and repeatability of parameter correction, improve the accuracy and consistency of buffer generation in different geographic environments, and solve the problem of unclear meaning of parameters in traditional methods.

[0073] In the embodiment, the method further includes the following steps: S600, using a field programmable gate array (FPGA), performing parallel calculation on the spatial relationship matrix to generate an optimized buffer.

[0074] Specifically, the FPGA chip is equipped with a special computing architecture, including a data preprocessing unit, a matrix calculation unit and a post-processing unit.

[0075] The data preprocessing unit is responsible for format conversion and blocking of the spatial relationship matrix; the matrix calculation unit adopts a systolic array structure and includes 64 processing units for parallel calculation; and the post-processing unit is responsible for result aggregation and buffer boundary optimization.

[0076] Further, the calculation process adopts a pipeline parallel strategy to divide the spatial relationship matrix into multiple sub-matrices and continuously process them through a data flow driven mode.

[0077] The strategy supports dynamic precision adjustment, and selects single-precision or half-precision floating-point operation according to application requirements, thereby balancing the calculation accuracy and resource consumption.

[0078] Based on this, a special FPGA architecture for spatial relationship calculation is designed, the calculation performance is effectively improved compared with the CPU scheme, and the buffer of a complex geographic scene can be generated in real time, meeting the high-concurrency GIS application requirements. Through hardware-level parallelization, the calculation performance bottleneck is broken, and the efficiency of buffer generation is greatly improved.

[0079] In some embodiments, in order to reduce data transmission delay, improve calculation efficiency and ensure real-time processing capability, S600 further includes the following steps: S601, store the terrain elevation data in the on-chip RAM cache, and transmit the spatial relationship matrix to the FPGA through the PCIe interface.

[0080] Specifically, after the terrain elevation data is read from the digital elevation model database, it is converted to fixed-point number format that can be processed by the FPGA through the data preprocessing unit of the central control host, and written into the Block RAM cache array configured on the FPGA chip through the direct memory access controller.

[0081] Among them, the spatial relationship matrix is stored in the main memory in the form of a sparse matrix, and is transmitted to the FPGA DDR controller through the PCIe 4.0 x16 interface using the DMA method.

[0082] Further, the on-chip RAM adopts a multi-body parallel architecture, and the elevation data is stored in spatial grid blocks to support simultaneous read and write operations.

[0083] Further, the PCIe transmission adopts a descriptor chain DMA mode, supports multi-matrix parallel transmission and interrupt aggregation technology, and can minimize transmission overhead.

[0084] Further, the ECC check and retransmission mechanism is implemented during data transmission to ensure data integrity.

[0085] Based on this, through the cooperation of on-chip RAM cache and high-speed PCIe interface, the data transmission delay is reduced, the calculation efficiency is improved, and reliable data supply guarantee is provided for real-time buffer generation.

[0086] In some embodiments, in order to improve the accuracy and speed of spatial relationship calculation, S600 further includes the following steps: S602, perform convolution operation using a systolic array structure to optimize the calculation of topological constraint weights.

[0087] Specifically, a configurable systolic processing array is constructed inside the FPGA, which contains 8x8 processing units, each unit is equipped with a multiplication accumulator and a local register file.

[0088] Among them, the array adopts a two-dimensional grid interconnection structure, and the data flows along the row direction and the weight is broadcasted along the column direction.

[0089] Among them, the topological constraint weight calculation adopts an improved sliding window algorithm, which supports multiple convolution kernel sizes from 3x3 to 7x7.

[0090] Further, the systolic array runs at a clock frequency of 500MHz, and can complete 64 multiplication and addition operations per cycle.

[0091] Among them, the operation adopts a dynamic precision adjustment mechanism, which automatically selects 16-bit or 32-bit floating point operation according to the complexity of the topological relationship.

[0092] The array controller monitors data flow conflicts in real time and dynamically adjusts data transmission timing.

[0093] Based on this, the pulsatile computing architecture is applied to the geospatial computing field, realizing efficient parallel processing of data flow, accelerating convolution operation, improving the accuracy and speed of spatial relationship calculation, solving the performance bottleneck problem of complex topological relationship calculation through hardware-level parallelization, and perfectly adapting to the real-time GIS application demand.

[0094] In some embodiments, in order to ensure the actual distance accuracy of the buffer on the three-dimensional terrain and significantly improve the buffer effect of complex terrains such as steep slopes and valleys, the following steps are included in S600: S603, obtaining terrain data based on a digital elevation model, and generating a three-dimensional buffer surface through a slope-sensitive interpolation algorithm.

[0095] Specifically, from a high-precision digital elevation model, terrain point cloud data is extracted to construct an irregular triangular mesh surface model.

[0096] Further, the slope-sensitive interpolation algorithm can dynamically adjust the sampling density according to the surface curvature, which can be specified as 1-meter resolution sampling in areas with a slope greater than 25 degrees, and 5-meter resolution sampling in flat areas.

[0097] Further, the interpolation calculation can use an improved radial basis function, introducing a slope weight factor to further enhance the interpolation accuracy in steep slope areas.

[0098] Specifically, the three-dimensional buffer generation can use a normal offset algorithm, which specifically projects the two-dimensional buffer polygon to the surface model and offsets a specified distance along the surface normal direction to generate a three-dimensional buffer.

[0099] Further, in steep slope areas, the along-surface distance integration algorithm can be used to ensure the consistency of the actual buffer distance.

[0100] Further, the algorithm can support multi-detail level rendering and dynamically adjust the surface details according to the view distance.

[0101] Based on this, the slope-sensitive interpolation and surface offset algorithm are combined to ensure the actual distance accuracy of the buffer on the three-dimensional terrain and significantly improve the buffer effect of complex terrains such as steep slopes and valleys, realizing the unity of precision and efficiency.

[0102] In this embodiment, the method further includes the following steps: S700, outputting the optimized buffer.

[0103] Specifically, the output process is based on an output module of the central host, which includes a data formatting unit, a quality checking unit and a transmission interface.

[0104] The data formatting unit converts the buffer data into a standard GeoJSON format; the quality checking 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 modes to adapt to different application scenarios.

[0105] Further, the output process implements a multi-checking mechanism, including geometry checking to ensure that the buffer polygons are closed and have no self-intersections, attribute checking to verify the completeness of the metadata, and topology checking to ensure the consistency of the relationship with surrounding geographic entities.

[0106] The mechanism supports an incremental output mode, which pushes the buffer generation status in real time.

[0107] Based on this, by combining the standardized output format and the multi-checking mechanism, the accuracy of the buffer output is greatly improved, and the output data can be directly used for spatial analysis and visualization, ensuring the reliability and usability of the buffer data and forming a complete processing closed loop.

[0108] In summary, by introducing FPGA for parallel computation of spatial relationship matrices, the data processing speed and energy efficiency ratio are significantly improved, real-time generation of buffers is achieved, and the problems of high computational delay and high power consumption of traditional software solutions are solved.

[0109] In some embodiments, to achieve adaptive optimization of the buffer and reduce human intervention, improve the reliability and intelligent level of the system, the method further includes the following steps: S800, use a graph neural network model to predict the conflict probability of the buffer boundary.

[0110] Specifically, the graph neural network model can use a gated graph attention network architecture, the input of which is a spatial relationship graph of geographic entities, the node features of which include entity type, size attribute, time-varying characteristics, etc., and the edge features of which include distance, orientation, topological relationship, etc.

[0111] Further, the network can include 3 graph convolution layers and 1 prediction layer, which outputs a conflict probability value between 0 and 1.

[0112] Further, the training data of the network comes from a historical buffer conflict case library.

[0113] Further, the network training can use a multi-task learning framework, which can simultaneously predict the conflict probability and the conflict type.

[0114] Further, the framework can use graph data augmentation techniques to expand the training samples and improve the model generalization ability.

[0115] Further, the reasoning stage adopts a subgraph sampling strategy to realize real-time processing of large-scale spatial graphs.

[0116] Based on this, the graph neural network is applied to the geographic spatial conflict prediction, so that the prediction accuracy of the model on the test set is improved, and the prediction accuracy is significantly improved compared with the traditional rule-based method. Through deep learning, the precision limit of the traditional method is broken through, and reliable prediction is provided for buffer optimization.

[0117] S900, according to the conflict probability, dynamically adjusting the expansion parameter of the buffer area.

[0118] Specifically, the dynamic adjustment system can adopt a fuzzy PID controller, and the input of the fuzzy PID controller is the conflict probability value, and the output of the fuzzy PID controller is the expansion parameter adjustment amount.

[0119] Further, the controller parameters can be adaptively set according to the conflict type and historical adjustment effect. The adjustment strategy can adopt a hierarchical response mechanism, which respectively fine-tunes the parameters when the conflict probability is low and starts a major adjustment when the conflict probability is high.

[0120] Further, the adjustment process can implement a predictive correction mechanism, which respectively predicts the effect after adjustment, corrects the adjustment amount according to the prediction result, and evaluates the actual effect after implementation, and updates the controller parameters.

[0121] Based on this, through the cross-field fusion of intelligent control theory and geographic spatial analysis, the adaptive optimization of the buffer area parameters is realized, the buffer area conflict occurrence rate is reduced, the average time consumption of the adjustment process is reduced, and the demand for manual intervention is greatly reduced. Break through the limitations of traditional manual adjustment, and improve the system automation level.

[0122] In summary, through the real-time prediction of conflicts and the adjustment of parameters by the graph neural network, the adaptive optimization of the buffer area is realized, the manual intervention is reduced, and the reliability and intelligent level of the system are improved.

[0123] The GIS buffer parameter inference system provided by the embodiments of the present application will be introduced below. The GIS buffer parameter inference system described below can be mutually corresponding to the GIS buffer parameter inference method based on deep learning described above.

[0124] Reference Figure 2 , the GIS buffer parameter inference system adopts a hierarchical distributed architecture, and each module is interconnected through a high-speed data bus, specifically including: The instruction acquisition module 1 is used to acquire the natural language instruction input by the user.

[0125] The instruction obtaining module 1 comprises a multi-modal interface unit and an instruction checking unit, supports multiple input modes such as text and voice, and performs standardized processing on the input instruction.

[0126] The parsing module 2 is used for parsing the natural language instruction to extract a geographical entity keyword and a fuzzy description word.

[0127] The parsing module 2 comprises a lexical analysis unit, a syntax analysis unit and a semantic labeling unit, adopts a deep learning model, and realizes accurate parsing of the natural language instruction.

[0128] The parameter determining module 3 is used for determining an initial distance parameter based on the fuzzy description word by using a pre-trained language-space mapping model.

[0129] The parameter determining module 3 is equipped with a double-channel output mechanism, comprising an absolute distance calculation unit and a scale factor generation unit, which can automatically select an output mode according to the geographical entity type.

[0130] The query module 4 is used for querying a spatial database based on the geographical entity keyword to obtain geographical entity attributes.

[0131] The query module 4 establishes a distributed index system, comprising a spatial index unit and an attribute retrieval unit, which supports efficient geographical entity attribute query.

[0132] The correction module 5 is used for correcting the initial distance parameter according to the geographical entity attribute to obtain a corrected distance parameter.

[0133] The correction module 5 adopts a multi-factor fusion architecture, comprising a semantic intensity correction unit, a scale correction unit and a terrain correction unit, and each unit outputs a standardized coefficient and then performs weighted fusion.

[0134] The FPGA processing module 6 is used for using a field programmable gate array (FPGA) to perform parallel calculation on a spatial relationship matrix to generate an optimized buffer.

[0135] The FPGA processing module 6 comprises a matrix calculation unit and a topological optimization unit, and realizes parallel processing of the spatial relationship matrix through hardware acceleration.

[0136] The output module 7 is used for outputting the optimized buffer.

[0137] The output module 7 constructs a double output channel, comprising a data formatting unit and a quality detection unit, to ensure that the output buffer meets the geographic information standard.

[0138] Further, the system runs in an event-driven mechanism, and each module exchanges data through a message middleware, and the subsequent module is automatically triggered to start after the prequel module is processed.

[0139] In some embodiments, the system can further include three auxiliary processing units, namely a semantic conflict detection unit responsible for identifying multi-target instruction conflicts, a three-dimensional terrain processing unit for implementing elevation data fusion, and a dynamic adjustment unit for optimizing parameters according to real-time feedback.

[0140] Further, the data processing flow adopts full-link monitoring, and each module is equipped with a performance acquisition unit and an exception handling unit to ensure stable operation of the system.

[0141] Further, the system supports an online update mechanism, and model parameters and rule libraries can be dynamically loaded without interrupting services.

[0142] In summary, the system integrates an FPGA processing module to realize efficient parallel computing of spatial relationship matrices through hardware acceleration, greatly improving the real-time performance and energy efficiency ratio of buffer generation, and is suitable for large-scale geographic data processing.

[0143] Embodiments of the present application provide an electronic device, such as Figure 3 as shown, Figure 3 a structural schematic diagram of an electronic device provided by the embodiments of the present application, Figure 3 The electronic device 300 shown in the figure includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, such as through a bus 302. Optionally, the electronic device 300 can also include a transceiver 304. It should be noted that in actual applications, the transceiver 304 is not limited to one, and the structure of the electronic device 300 does not constitute a limitation on the embodiments of the present application.

[0144] The processor 301 can be a CPU (Central Processing Unit, central processor), a general-purpose processor, a DSP (Digital Signal Processor, data signal processor), an ASIC (Application Specific Integrated Circuit, application-specific integrated circuit), an FPGA (Field Programmable Gate Array, field programmable gate array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can realize or execute various exemplary logical blocks, modules and circuits described in combination with the disclosure of the embodiments of the present application. The processor 301 can also be a combination of computing functions, such as one or more microprocessor combinations, combinations of DSP and microprocessor, etc.

[0145] The bus 302 can include a path that transmits information between the above-described components. The bus 302 can be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus 302 can be divided into an address bus, a data bus, a control bus, etc. For convenience of representation, Figure 3 Only one thick line is used in the middle, but it does not mean that there is only one bus or one type of bus.

[0146] The memory 303 can be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited thereto.

[0147] The memory 303 is used to store application program codes for implementing the embodiments of the present application, and is controlled by the processor 301 to perform. The processor 301 is used to execute the application program codes stored in the memory 303 to realize the contents shown in the foregoing method embodiments.

[0148] The electronic device includes, but is not limited to, a mobile terminal such as a mobile phone, a notebook computer, a digital broadcast receiver, a PDA (Personal Digital Assistant), a PAD (Tablet Personal Computer), a PMP (Portable Multimedia Player), a car terminal (for example, a car navigation terminal), and the like, and a fixed terminal such as a digital TV, a desktop computer, and the like. Figure 3 The electronic device shown is only an example, and should not bring any limitation to the functions and use range of the embodiments of the present application.

[0149] The embodiments of the present application provide a computer readable storage medium, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the steps of the GIS buffer parameter inference method.

[0150] Since the embodiments of the computer readable storage medium part correspond to the embodiments of the method part, the embodiments of the computer readable storage medium part refer to the description of the embodiments of the method part.

[0151] It should be understood that although the steps in the flowcharts of the drawings are shown in a certain order according to the arrows, the steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of the steps is not strictly limited in order, and the steps can be executed in other orders. Moreover, at least part of the steps in the flowcharts of the drawings can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.

[0152] The above is only some embodiments of the present application, and it should be pointed out that for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which should also be considered as the protection scope of the present application.

Claims

1. A GIS buffer parameter inference method based on deep learning, characterized in that: The following steps are involved: Obtain natural language instructions input by the user; Parsing the natural language instructions to extract geographic entity keywords and fuzzy descriptive words; Based on the fuzzy description words, an initial distance parameter is determined through a pre-trained language-space mapping model; Based on the geographic entity keywords, query the spatial database to obtain geographic entity attributes; Correcting the initial distance parameter according to the geographic entity attribute to obtain a corrected distance parameter; Using a field programmable gate array (FPGA), the spatial relationship matrix is ​​parallelized to generate an optimized buffer; The optimized buffer is output.

2. The method according to claim 1, characterized in that The step of using a field programmable gate array (FPGA) to perform parallel calculations on the spatial relationship matrix includes: The terrain elevation data is stored in the on-chip RAM cache and the spatial relationship matrix is ​​transmitted to the FPGA through the PCIe interface.

3. The method according to claim 1, characterized in that The step of using a field programmable gate array (FPGA) to perform parallel calculations on the spatial relationship matrix includes: A systolic array structure is used to perform convolution operations to optimize the calculation of topological constraint weights.

4. The method according to claim 1, wherein The step of correcting the initial distance parameter according to the geographic entity attributes includes: A weighted calculation is performed based on the semantic strength modifier, the scale correction coefficient, and the terrain correction coefficient, where each coefficient has a unified physical meaning and unit definition.

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

6. The method according to claim 1, characterized in that The step of using a field programmable gate array (FPGA) to perform parallel calculations on the spatial relationship matrix to generate an optimized buffer zone includes: Terrain data is obtained based on the digital elevation model, and a three-dimensional buffer surface is generated through a slope-sensitive interpolation algorithm.

7. The method according to claim 1, characterized in that Also includes: Use a graph neural network model to predict the conflict probability of buffer zone boundaries; The expansion parameters of the buffer zone are dynamically adjusted according to the conflict probability.

8. A GIS buffer parameter inference system, characterized in that: include: An instruction acquisition module is used to obtain natural language instructions input by the user; A parsing module, configured to parse the natural language instructions to extract geographic entity keywords and fuzzy descriptive words; A parameter determination module, configured to determine an initial distance parameter based on the fuzzy description word through a pre-trained language-space mapping model; A query module, configured to query a spatial database based on the geographic entity keywords to obtain geographic entity attributes; A correction module, configured to correct the initial distance parameter according to the geographic entity attribute to obtain a corrected distance parameter; an FPGA processing module, configured to perform parallel calculations on the spatial relationship matrix using a field programmable gate array (FPGA) to generate an optimized buffer; An output module is used to output the optimized buffer.

9. An electronic device, characterized in that: include: one or more processors; one or more memories; and one or more computer programs, wherein the one or more computer programs are stored in the one or more memories, and the one or more computer programs include 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 7.

10. A computer-readable storage medium, characterized in that The storage medium stores a program or instruction, and when the program or instruction is executed, the method according to any one of claims 1 to 7 is implemented.

Images