72 results about "Semantic interpretation" patented technology

A real-time system for the multimodal generation of virtual reality scenes based on artificial intelligence for the creation of immersive three-dimensional environments from natural language narratives, consisting of: a speech capture module configured to continuously record a user's spoken narrative via one or more directional microphones, preprocesses the captured signal by noise reduction and temporal alignment, and outputs a digital speech stream; A speech-to-text processing unit that is operationally coupled to the speech capture module and configured for real-time speech recognition using a continuous neural transformer model. The unit is trained to transcribe natural language utterances into structured text data while maintaining contextual continuity throughout the evolving narrative. a semantic interpretation processing unit that is communicatively linked to the speech recognition unit and configured to perform natural language understanding techniques to extract contextual entities, spatial references, temporal relationships, and object attributes from the transcribed narrative; the engine includes a large language model that is fine-tuned for spatial reasoning tasks; a scene graph generation module configured to transform the interpreted semantic data into a structured, hierarchical representation that defines nodes for identified entities and edges for corresponding relationships, with each node associated with metadata describing geometry, position, orientation, texture, and linking attributes between objects; a multimodal image-language model processor coupled with the scene graph generation module, wherein the processor is configured to retrieve, adapt, or synthesize appropriate three-dimensional elements from a pre-trained visual-lexical embedding space and align these elements with their semantic and spatial definitions derived from the scene graph; a scene assembly and rendering controller configured to create a cohesive virtual scene from the aligned assets, perform real-time rendering using a GPU-accelerated ray tracing pipeline, and produce a stereoscopic visual output that corresponds to the evolving narrative; A head-mounted virtual reality visualization device connected to the rendering engine and configured to display the generated immersive environment to the user in real time. The device features motion sensors and inside-out tracking cameras to detect head and body movements, dynamically updating viewing angles and perspective within the rendered scene; and a bidirectional feedback module integrated into the head-mounted device and connected to the semantic interpretation processing unit; the module is configured to interpret corrective commands, gestures, or supplementary comments from the user to refine or modify specific scene elements without interrupting the real-time visualization; The system continuously updates the virtual scene as the narrative develops, ensuring temporal synchronization between speech input and rendered output below a defined latency threshold, thus enabling a natural, dialogic construction of complex three-dimensional virtual environments.

The invention relates to an automatic assembly system and method for multi-scale digital twin components, and the system comprises a semantic relationship construction module which is used for carrying out explicit definition on the structural features, functional attributes, spatial layout requirements and logic dependency relationships of the multi-scale components, and constructing semantic relationships among the three types of components; the semantic reasoning and constraint engine module is used for carrying out logical reasoning through the semantic relationship constructed by the semantic relationship construction module and judging whether the component combination meets the assembly constraint or not; the assembly generation and configuration module is used for generating an assembly topological structure and a connection sequence of a system shelf according to the component candidate set output by the semantic reasoning and constraint engine module; and the man-machine interaction and visualization module is used for supporting a feedback closed loop between the engineer and the system and providing a visual display interface. The problems that an assembly method depends on artificial experience and is difficult to support high-frequency and multi-scene production line reconstruction are solved, and the method has higher semantic interpretation capacity, automatic combination capacity and context adaptive capacity.

The invention discloses a remote sensing image three-dimensional reconstruction method based on semantic information, and relates to the technical field of remote sensing image three-dimensional modeling, and the method comprises the steps: carrying out the processing of a remote sensing image covering a target region, obtaining point cloud data, and carrying out the semantic segmentation based on a semantic segmentation model, and determining a semantic category label; dividing voxel units based on a three-dimensional sparse point cloud distribution condition, and constructing to obtain an initial anchor point; the distribution density of the initial anchor points is adjusted in combination with the semantic category labels, and the adjusted initial anchor points are obtained; iteratively training the three-dimensional Gaussian splash model for multiple times to update the anchor points to obtain scene anchor points; and obtaining a three-dimensional reconstruction model of the target area based on scene anchor point rendering. According to the method, semantic information is introduced and a 3DGS three-dimensional modeling technology is fused, so that the densification quality of anchor points and the geometric boundary definition of the model are effectively improved, the modeling efficiency can be improved, and the structural rationality and semantic interpretation of a three-dimensional reconstruction result can be enhanced.

A task generation method performed by a computer device includes: obtaining task requirement information inputted based on a task generation interface; calling a large language model, and inputting the task requirement information into the large language model for performing semantic interpretation on the task requirement information and outputting an executable structural body of a target task that matches the task requirement information; performing graphic rendering on the target task based on the executable structural body of the target task, to obtain a task execution flowchart of the target task, and displaying the task execution flowchart in the task generation interface, the target task being formed according to one or more target atomic tasks; and displaying an execution progress viewing link of the target task in the task generation interface, the execution progress viewing link being configured for viewing an execution progress of the target task.

The invention provides a multi-language cross-modal information retrieval method, device and equipment. The method comprises the following steps: acquiring first to-be-retrieved data of a first modal; preprocessing the first to-be-retrieved data to obtain first data; inputting the first data into a corresponding type of encoder in a multi-modal information retrieval model for encoding to obtain a first type of encoded data; the encoder comprises a Chinese encoding unit; inputting the first type of coding data into a projection layer of a multi-modal retrieval model for projection processing to obtain a first target coding vector; and according to the first target coding vector, performing at least one type of other modal information retrieval in a vector database to obtain at least one type of other modal retrieval result data with the same semantics as the first data. According to the method, the multi-modal information retrieval efficiency, the semantic interpretation capability and the multi-modal alignment capability can be improved.

The invention belongs to the technical field of geographic information, and discloses a city update planning scheme optimization method and system driven by a large language model, a terminal and a storage medium, and the method comprises the steps: obtaining multi-source heterogeneous data, carrying out site context interpretation according to the multi-source heterogeneous data, and generating site semantic representation; according to a semantic text input by a user, converting the semantic text into a planning parameter vector set by utilizing regulation constraints and the site semantic representation; based on the regulation constraints, the site semantic representation and the planning parameter vector set, generating a space planning scheme conforming to a city updating target; and performing multi-dimensional index detection, semantic interpretation and parameter feedback optimization on the spatial planning scheme based on an index evaluation and semantic feedback optimization large language model to obtain an optimized city update planning scheme. According to the method, the problems of disjunction between scheme generation and optimization and insufficient interactive interpretation in the prior art are effectively solved.

The invention discloses a difference interaction method for a visual editor, which comprises the following steps: converting bottom data of the visual editor into a tree structure, and setting a unique identifier and a path index for each node in the tree structure; comparing the original tree structure with the newly generated tree structure through a tree difference algorithm according to the path index to obtain difference nodes, generating atomized Patches in one-to-one correspondence, and mapping the atomized Patches and the nodes through unique identifiers; and according to the atomized Patch, semantic description is obtained through a semantic interpretation engine so as to generate semantic Patches, and the semantic Patches are visually displayed so as to respond to user confirmation to perform dynamic updating. Through tree structure modeling, atomization difference analysis and semantic interaction, the problems of black box, version control deficiency, insufficient interaction experience and the like of a generative AI auxiliary visual editor in the prior art are solved.

The invention provides an automatic driving track generation method and device, a storage medium and an electronic device, and the method comprises the steps: obtaining multi-view image data collected by a camera of a vehicle, and obtaining the attribute parameters of the camera; the multi-view image data is adopted to generate two-dimensional image features, and the two-dimensional image features comprise local visual features of multiple views; the two-dimensional image features and the attribute parameters are adopted to generate bird's-eye view BEV features, the two-dimensional image features are adopted to generate high-level semantic information, the high-level semantic information comprises meta-actions and enhanced semantics, the meta-actions are used for describing rough actions of path planning, and the enhanced semantics are used for conducting semantic interpretation on the meta-actions; and generating an automatic driving path of the vehicle by adopting the BEV features and the high-level semantic information. According to the embodiment of the invention, the technical problem of low reliability of an automatic driving path generated by an end-to-end driving model in related technologies is solved, and the safety and robustness of an automatic driving system are improved.

The governed interaction interface maintains an identity-preserving semantic-state for a knowledge entity and updates it only through validated semantic evidence generated autonomously by execution mechanisms. Semantic intentions describe semantic needs and are transformed into capability-requests containing no operational commands. Execution mechanisms—including environmental devices, perception models, symbolic analyzers, virtual actors, and software agents—interpret capability-requests independently through tool-side semantic interpreters and may act or decline to act. The system does not observe or evaluate tool behavior and receives only semantic evidence describing semantic meaning of any resulting effect. The governance engine evaluates each evidence fragment independently under identity, coherence, lineage, evidentiary sufficiency, deviation constraints, and contextual compatibility. Semantic-state transitions occur only when validated evidence satisfies governance constraints. The system performs no prediction, optimization, control computation, multimodal fusion, or supervisory coordination, remaining fully separated from device-level behavior.

The invention discloses a personalized communication training content generation method based on big data. The method comprises the following steps: S1, constructing a multi-modal feature sequence; s2, constructing a user expression brain map; s3, performing cognitive causal alignment on the time axis of the user expression brain map; s4, identifying ambiguous expressions in user input, expanding a plurality of semantic interpretation paths, forming a semantic trajectory tree, and generating a primary semantic path and a secondary semantic path; s5, generating conventional communication content based on the primary semantic path, constructing an evolution expression path based on the secondary semantic path, generating challenging expression content, and integrating and outputting a double-track fusion training corpus; s6, identifying a negative expression fragment in the double-track fusion training corpus, and generating an evolution training sequence based on an annealing control function; and S7, performing content reconstruction on the double-track fusion training corpus based on the evolution training sequence, and generating personalized communication training content. According to the method, multi-modal modeling and a semantic evolution control method are fused, and personalized communication training content generation is realized.

Device for the intelligent encapsulation of legacy data information systems with the integration of artificial intelligence, comprising: a collection unit for legacy interfaces, configured to intercept communication data from existing platforms; a semantic interpretation unit designed to analyze and map data structures and process semantics; a contextual abstraction layer designed to convert interpreted information into standardized, interoperable formats; a policy-driven management module that ensures secure access, Data transformation and regulatory compliance are ensured; and a learning and inference unit for adaptive optimization of data mappings and operational logic.

The invention discloses a data mining method and system based on artificial intelligence driving. The method comprises the steps that S1, an original data stream is received, and heterogeneous data feature signals are generated; s2, inputting the heterogeneous data characteristic signal into a dynamic quality sensing module, generating a quality evaluation feedback signal and a purification trigger signal, and outputting a cleaned optimized characteristic signal; s3, in response to the optimization feature signal, synchronously outputting a classification label signal, a clustering correlation signal and an abnormal scoring signal; s4, generating an interpretability enhancement signal based on the mode recognition control signal, and generating a semantic interpretation signal aligned with the business rule; s5, the semantic interpretation signal is received, a closed-loop optimization instruction signal is generated, the feature extraction strategy and the model hyper-parameters are dynamically adjusted through the reinforcement learning agent, and a self-adaptive iterative mining link is formed. According to the data mining method and system based on artificial intelligence driving, the problems of heterogeneous data fusion distortion, poor dynamic environment adaptability and untrusted model decision can be solved.

The invention discloses a geologic model dynamic construction method and system based on digital twinning, and relates to the technical field of geologic modeling, and the method comprises the steps: carrying out the semantic interpretation of multi-source geologic data based on a domain ontology library, and generating an initial geologic model; representing the initial geologic model as a semantic topological graph; in response to updating of the exploration data, performing semantic mapping on the exploration data and the domain ontology library, performing topology diffusion analysis and geological law constraint in a semantic topological graph based on a mapping result, and determining a semantic topological influence domain to be updated; creating a virtual branch for the semantic topology influence domain, and copying the geologic volume metadata in the influence domain into the virtual branch; in the virtual branch, according to the exploration data and the existing data, reconstructing the attribute of the geologic volume element to obtain an updated virtual branch; and combining the change content after conflict resolution to a semantic topological graph to obtain an updated geological model. The invention provides a convenient and reliable dynamic geological modeling method.

The invention provides an interpretable fault diagnosis method, system and device for a fan blade and a medium, and relates to the technical field of wind power plants. According to the method, multi-modal sensing data and priori knowledge of material science are integrated, and a pre-trained multi-modal big language model is utilized to perform data fusion and causal reasoning, so that accurate diagnosis of fault types of the fan blades, accurate positioning of damage positions and effective tracking of damage development paths are realized. The method not only improves the accuracy and robustness of fault diagnosis, but also provides an explanatory result which is easy to understand, thereby providing powerful support for engineering technicians, and enabling the fault diagnosis process to be more transparent and reliable. The method effectively overcomes the problems that a traditional single-mode diagnosis method is weak in generalization ability and lacks semantic interpretation.

This invention discloses a method for air-ground collaborative risk prediction and dynamic electronic fence construction based on a generative large-scale model, comprising: obtaining a standardized air-ground collaborative multimodal spatiotemporal dataset; generating a structured state representation for risk inference; outputting risk semantic interpretation, candidate risk links, and risk factor embedding representations; constructing a spatiotemporal risk field and performing forward prediction based on the risk factor embedding representation and the air-ground collaborative state map, obtaining a risk probability heatmap and uncertainty map within the future prediction time window; generating a three-dimensional dynamic electronic fence set, and assigning a confidence weight and an effective timestamp to each fence boundary; and performing event-triggered iterative updates to the air-ground collaborative state map, risk probability heatmap, and three-dimensional dynamic electronic fence set based on real-time new data. In dynamic scene simulation, this invention reduces the frequency of false fence triggering compared to periodic recalculation methods, while maintaining a significant decrease in risk omission rate.

The invention provides an automatic benchmarking method and system based on a large model and a storable medium, in response to a table name input by a user, the system transmits metadata and sample data of a resource table corresponding to the table name to the large model; the large model generates semantic interpretation corresponding to each field in the resource table to obtain a recognition result, and the system stores the recognition result into a database original table corresponding to the resource table as a newly added column and splits the resource table into field arrays according to the fields; traversing the field array by a system, and comparing elements in the array with data elements in a vector knowledge base by utilizing an RAG technology to obtain data elements similar to the field array; the large model selects similar data elements for benchmarking according to the metadata to obtain a benchmarking result of the current resource table, and meanwhile the benchmarking result serves as a newly-added column to be stored in the database original table. According to the method, a manually maintained hard coding rule is converted into a semantic matching model, the method has the advantages of self-adaptability, intelligence and lightweight operation and maintenance, and the data value can be released.

The invention discloses a construction fence structural change detection and multi-mode auxiliary disposal method and system, and belongs to the field of constructional engineering safety management, and the method comprises the following steps: collecting a current image of a fence region, and carrying out the registration and alignment of the current image with a benchmark reference image; respectively extracting feature maps of the two registered maps; generating a depth map based on the current feature map; fusing the reference feature map, the current feature map and the depth map to carry out change detection, and outputting a bounding box and a type label of a change region; estimating the physical size of the change area in combination with the depth map and the camera internal reference; constructing a structured result based on the bounding box, the type label, the physical size and the image; inputting the structured result into a multi-modal large language model, and outputting semantic interpretation and maintenance suggestions; and generating a maintenance work order. The intelligent level and response efficiency of construction safety management are greatly improved.

The invention discloses a business handling reservation state query method based on a knowledge graph, and particularly relates to the technical field of knowledge graphs. The method comprises the steps of collecting reservation original data, extracting a state keyword and constructing a state semantic vector; performing entity alignment and attribute mapping based on the state ontology library to generate a local knowledge sub-graph; integrating a plurality of sub-graphs to construct a unified reservation state knowledge graph, and generating a graph structure feature vector by adopting a graph embedding algorithm; reasoning the user nodes through a graph neural network, and identifying standard state entities associated with the user nodes at present; in combination with upstream and downstream relation chains in the atlas, outputting state feedback information including state semantic interpretation, a circulation path and a predicted handling time period; compared with the prior art, standardized expression and intelligent query of complex state semantics can be achieved, the accuracy and intelligent level of reservation state recognition are improved, and the method is particularly suitable for a state management scene in a multi-source heterogeneous service system.

The invention provides an intelligent structure design generation method and system, and is suitable for a generative design task in the field of complex structure design, the method introduces a semantic interpretation mechanism based on a sketch graph structure, combines multi-scale texture perception and a graph neural network embedding strategy, and can improve the structure design efficiency under the premise of not depending on a traditional topological optimization solution framework. And efficient generation and diversified control of the structural form are realized. According to the method, through gesture sketch input, fuzzy structure intention map construction, wavelet scale regulation and control and structure solution set evolution screening strategies oriented to man-machine collaboration, the controllability of understanding is enhanced, the diversity of understanding sets is remarkably improved, and therefore the requirements for multi-target comprehensive optimization of aesthetics, technology, mechanics and the like in actual design are met.

The invention provides a graph self-encoder community detection method based on semantic enhancement, and the method comprises the steps: constructing an attribute graph which comprises node attribute features and a network topology structure as input, and extracting the structure and semantic features through employing a graph convolutional neural network (GCN); a node structure relation is systematically expanded to a multi-order neighborhood from a first-order neighborhood by using a high-order adjacency matrix calculation module so as to capture a global connection mode of the network, and the stability and robustness of community detection can be further improved by reconstructing high-order information; a modularity maximization module is utilized to enable the model to face community detection, that is, a community view angle is introduced, so that community structure information is checked during iterative optimization of the model; on this basis, a community semantic matrix modeling module is introduced, explicit association between community division and attribute features is established through a structure-semantic alignment mechanism, and a semantic prototype of each community in a feature space is automatically learned; and finally constructing a joint optimization target containing graph structure reconstruction loss, modularity maximization loss and structure-semantic alignment loss, optimizing model parameters through an end-to-end training mode, and outputting a community division result with clear semantic interpretation. According to the method, deep collaborative optimization of the network structure and the node attributes in community detection is realized, and the accuracy and interpretability of community detection are effectively improved.

The invention relates to the technical field of computers, and discloses a subject cross analysis method and system, electronic equipment and a storage medium, and the method comprises the steps: obtaining paper metadata, calculating the subject affiliation degree of each paper, and dividing the papers into an institution paper set and a subject paper set according to the unit affiliation and the subject affiliation; preprocessing and vectorizing paper text contents in the organization paper set and the subject paper set to generate semantic vectors, and constructing a vector database; calculating a subject crossover degree based on the subject attribution degree of the organization paper set or the subject paper set; calculating the subject similarity based on the subject attribution degrees of different institution paper sets or different subject paper sets; keywords and co-occurrence keywords are extracted to form keyword clusters for semantic interpretation, or related papers are retrieved by using a vector database, and answers are generated in combination with a large language model. The method has the capability of integrating literature measurement quantitative analysis and large model semantic understanding, and intelligent analysis can be realized.

The invention discloses a rehabilitation semantic understanding and intelligent inquiry method based on large language model fine tuning, which comprises the following steps of: analyzing an input natural language into a standardized rehabilitation semantic structure, identifying semantic information and generating a semantic tag; and after semantic analysis is completed, dynamic and interpretable inquiry interaction is realized through a context memory unit, semantic state modeling and a strategy controller. According to the invention, an intelligent inquiry strategy controller and a semantic confidence mechanism are adopted, so that the model has self-review and dynamic inquiry capabilities. And when the system detects that the semantic meaning is fuzzy or the confidence coefficient is low, a follow-up question can be automatically generated, so that the model is not a passive responder any more and becomes an active information collector, and the naturalness and clinical practicability of man-machine conversation are improved. Meanwhile, an attention heat map and a semantic interpretation interface of the model provide a visual basis for doctors, the black box problem of the deep model is solved, and the supervision requirements of traceability, safety and controllability of medical artificial intelligence are met.

The invention discloses an operation and maintenance system data semantic automatic identification and entity construction method based on a large language model, and the method comprises the following specific steps: 1, compiling an automatic script, and traversing all menus, pages and function points of a system through recursion; 2, loading a system page by using a front-end automation tool, and extracting a UI text related to a data field; 3, a prompt word is constructed, and a structured Prompt template is designed; 4, according to the nested structure of the interface response data, the constituent parts of the entity are preliminarily determined, and the user entity comprises the attributes of user Id and user Name; step 5, visualizing and outputting a semantic recognition result; and outputting a final result including a field list, name and semantic interpretation, an entity list and an entity relationship in a graphical form. Original manual empirical work is converted into an automatic process, the system data understanding efficiency is greatly improved, and a large amount of manpower and time cost are saved.

The application discloses a kind of fast positioning combined semantic interpretation's semiconductor defect analysis method and system, belong to semiconductor defect analysis field.Wafer defect image diagnosis data set is obtained, including wafer defect SEM image and diagnosis report, diagnosis report includes sentence for multiple defect diagnosis tasks;Defect image diagnosis network is trained using wafer defect image diagnosis data set, and defect image diagnosis network includes FasterR-CNN encoder and the LSTM decoder based on attention mechanism for different defect diagnosis tasks;In the training process of defect image diagnosis network, the sentence of defect diagnosis task is regarded as the input of decoder, and the target sentence is generated by guiding decoder;Wafer defect SEM image is analyzed using the trained defect image diagnosis network, and generates diagnosis report.The application establishes a closer link between the latest progress of visual and language tasks and target detection, and the accuracy of output sentence is greatly improved.

This invention discloses an automatic mathematical formula conversion method and system integrating image recognition and semantic processing, relating to the field of data processing technology. The method includes: performing symbol-level recognition on mathematical formula images to obtain an original symbol data set; parsing the structured data to generate a data expression set containing symbol units, categories, positions, and interrelationships; constructing a candidate structure space for mathematical expressions representing multiple potential semantic interpretations; introducing a set of mathematical semantic constraints for consistency evaluation to determine the target mathematical expression structure; and performing structure-preserving formula conversion processing to generate the mathematical formula representation result. This invention solves the technical problems of insufficient structural ambiguity resolution, missing mathematical semantic information, and structural distortion during format conversion in existing technologies, leading to insufficient accuracy and reliability in mathematical formula recognition and conversion. It achieves the technical effect of accurate and unambiguous parsing of mathematical formulas and improves the accuracy and reliability of formula recognition and conversion.

This invention discloses a method for mobile application privacy policy compliance detection based on a large language model, specifically relating to the field of privacy protection. The method includes: collecting operational behaviors of a mobile application before and after invoking system permissions, constructing a time-sequential behavioral path chain, and inputting it into a large language model specializing in behavioral intent descriptions to generate semantic interpretations of nodes and concatenate them into a behavioral intent chain; determining behavioral jump relationships based on the direction of change between adjacent semantic segments, and constructing an evolution graph of system permission usage; extracting natural language paragraphs related to permission usage in the privacy policy, parsing and generating a set of system permission usage declaration scopes; comparing the semantic coverage of usage nodes in the evolution graph with the declaration scopes, identifying usage nodes not covered by the declarations, and tracing back their behavioral source structure based on the connection relationships in the graph to locate usage expressions not covered in the privacy policy content, and outputting structured privacy policy compliance analysis results.

A computer-implemented method is provided for converting a human intent of a user into an artificial intelligence (AI) prompt. The method includes receiving a textual statement from the user, the textual statement comprising the human intent; employing an interpretation generation engine to generate a plurality of semantic interpretations of the textual statement; receiving a selection of a first semantic interpretation of the plurality of semantic interpretations from the user, the first semantic interpretation comprising a clarified human intent; generating the AI prompt based at least in part on the selection such that the AI prompt is configured to elicit a response from an AI model based on the clarified human intent; and sending the AI prompt to the AI model in order to generate the response.

The invention discloses an API calling method and system based on a multi-modal large language model and a knowledge graph, and relates to the technical field of artificial intelligence, and the method comprises the steps: converting multi-modal data submitted by a user into multi-modal semantic representation; determining a unique identifier of the target API according to the multi-modal semantic representation and a pre-constructed domain knowledge graph by utilizing a multi-modal large language model; generating an API request of the target API according to the unique identifier; carrying out concurrent requests on API interfaces corresponding to the plurality of API requests to obtain API original data; semantic interpretation is carried out on the API original data, and then a corresponding multi-dimensional deterministic visualization result is output. According to the method, the multi-modal intention understanding is realized by converting the multi-modal data into the multi-modal semantic representation, the API dynamic arrangement is performed according to the unique identifier, and the multi-dimensional deterministic visualization result is displayed, so that the intelligent presentation of the result is realized, and the robustness and the adaptability of calling the API can be improved.

In order to prevent semantic data leakage, one or more semantic interpretations of an input text are obtained using a generative pre-trained translator (GPT), where the GPT is to interpret the input text with reference to a target context; and determining whether the input text includes hidden data leaks by checking the semantic interpretation of the input text using a data leak prevention (DLP) system, where the DLP system is used to detect data leaks based on a set of words and phrases of interest. Accordingly, it is possible to prevent hidden data leakage when sensitive information is hidden in text (e.g., by a high-level language technology).

The application relates to the technical field of automatic driving safety monitoring, in particular to a control method and system for explainable monitoring and remote takeover based on VLA decision, wherein when a VLA model generates a decision, semantic information of the current decision and semantic information of a target region with the highest weight related to the current decision are extracted, and a semantic explanation of the current decision is constructed based on the semantic information of the current decision and the semantic information of the target region, so that the problem that an existing VLE model lacks explainability is solved. In addition, a rule library is constructed in advance, the semantic explanation is matched with the rule library, it is judged whether there is a conflict with the traffic rules in the park, if there is a conflict, the conflict probability is quantified further, and the remote takeover mode is decided based on the conflict probability, so that the takeover speed is improved, and the small car can be prevented from continuously running in the case of violating the traffic rules in the park.