148 results about "Semantic annotation" patented technology

The invention provides a mine abnormal event real-time identification method and system based on time sequence characteristics, and relates to the technical field of mode identification, and the method comprises the steps: carrying out the time-space alignment and semantic annotation of multi-modal monitoring data, and constructing a time sequence knowledge graph; calculating a dynamic association weight between entities, and analyzing a risk propagation path; predicting a risk situation based on a historical evolution rule; and dynamically generating a differential early warning strategy and establishing a closed-loop tracking system. According to the invention, early identification, accurate prediction and efficient disposal of mine safety risks can be realized, and the mine safety management level is improved.

The invention relates to the field of artificial intelligence, and discloses an intelligent AI semantic annotation method and system based on multi-modal analysis, images, audios, texts and multi-modal data of a sensor are synchronously collected through AI, deep semantic features of all modals are extracted by using a deep learning model after preprocessing, vectors are generated, and the multi-modal data are subjected to semantic annotation; fusing into a semantic scene vector by means of a cross-modal attention mechanism, constructing a cross-modal association graph, and learning a causal logic path by using a graph neural network; constructing a spatiotemporal causal diagram by combining spatiotemporal information, and integrating an event time sequence and spatial dependence by using a time sequence diagram neural network to form a venation path; according to the method, the two knowledge maps are fused into a comprehensive knowledge map, preliminary annotations are generated through rule base reasoning, a high-precision result is output after optimization, the problems of semantic loss and annotation fragmentation caused by modal splitting and context missing in a traditional method can be effectively solved, meanwhile, the manual verification cost is reduced, and the integrity and efficiency of semantic annotation in a complex scene are improved.

The present disclosure provides a method for generating annotation data for robotic training using a hierarchical transformer-based model with multiple layers. The transformer-based model includes Alpha models generating low-level control outputs and Beta models generating high-level control outputs. The method receives multimodal input data comprising visual sensor data and natural language instructions, processes this data through the hierarchical transformer-based model to generate annotations at different abstraction levels, wherein Beta models create semantic annotations describing task objectives and Alpha models generate motor command annotations specifying robotic actions, and stores these annotations with the input data to create annotated training data for robotic control systems.

The invention relates to a smart city ESG multi-source data visualization analysis method, device and equipment and a medium, and the method comprises the steps: carrying out the standardized semantic annotation and space-time alignment of city multi-source heterogeneous data, and generating a semantic triple set with a space-time label; the real-time intensity of the causal relationship in the ESG dynamic knowledge graph is constructed and quantified; semantic binding is carried out on the map and the city information model, and a visual descriptor is generated; discovering and rendering a causal relationship chain in the three-dimensional scene based on user interaction; and deducing simulation intervention and optimizing model parameters. According to the method, dynamic and integrated presentation and interactive deduction of complex causal relationships of environment, society and governance in a multi-dimensional space-time scene are realized, and the problems of surface analysis, disjunction of data and space and insufficient decision support in the prior art are effectively solved.

The invention provides a semantic aerial view visual relocation method and device in a non-exposed scene, electronic equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a multi-view image sequence under a non-exposed scene (such as a tunnel, an underground pipe gallery or an underground parking lot); semantic recognition is carried out based on a pre-trained semantic target detection model, and spatial consistency semantic features are extracted through a semantic-geometric dual-channel fusion mechanism combining a semantic mask and geometric constraints; the method comprises the following steps of: realizing three-dimensional reconstruction by using a voxel micro-renderable modeling method (VGGT), and generating a dense three-dimensional semantic point cloud fusing semantics and a geometric structure; two-dimensional semantics are mapped to a three-dimensional space through a projection and back projection relation, and point cloud semantics are endowed; main structure planes such as the ground, the left wall surface and the right wall surface are extracted, and a two-dimensional semantic aerial view with semantic annotation is generated; and pose estimation is carried out based on a reciprocal matching strategy guided by a semantic mask, so that visual repositioning with high precision, high robustness and semantic interpretability is realized. The method breaks through the problems of low precision, sparse features and poor semantic consistency of traditional visual repositioning in a non-exposed environment, and can be widely applied to the fields of intelligent transportation, underground inspection and unmanned system positioning.

The invention relates to the technical field of water network intelligent management and control, and discloses a water network management and control multi-source space digital intelligent fusion base construction and decision optimization method, which comprises the following steps: cooperatively obtaining multi-source space data; format conversion, coordinate system unification and systematic semantic annotation are carried out through a data preprocessing platform, and a normalized spatial data set is formed; constructing a data fusion model, quantitatively evaluating the quality of a data source by adopting a fuzzy comprehensive evaluation method so as to determine a fusion weight, and performing weighted fusion by adopting a specific algorithm aiming at different data types such as vectors, grids and three-dimensional models so as to generate a unified spatial data model; and finally, a decision optimization system performs decision scheme generation and iterative optimization based on the model in combination with a hydrodynamic model and a genetic optimization algorithm. By establishing a standardized data processing and targeted weighted fusion process, the problems that multi-source heterogeneous data is difficult to integrate and the fusion precision is not high are effectively solved, and a reliable data base is constructed.

The invention discloses a human body gesture generation method and related equipment, and relates to the technical field of computer vision, and the method comprises the steps: obtaining multi-modal input information, the multi-modal input information comprises a voice audio signal, text transcription information and a reference gesture sequence, the text transcription information comprises a semantic annotation, and the reference gesture sequence comprises a reference gesture sequence; the reference gesture sequence is a basic gesture template in the target scene; generating a multi-modal feature based on the multi-modal input information; performing time alignment processing operation on the multi-modal features to obtain alignment condition features; performing space-time decoupling modeling operation on the alignment condition features to obtain optimized gesture potential features; inputting the optimized gesture potential features into a diffusion model for iterative denoising to obtain denoised gesture potential features; and generating a target human body gesture sequence based on the de-noised gesture potential features.

The invention provides an efficient annotation-driven hierarchical fault positioning method, which comprises the following steps of: guiding a large language model to intelligently analyze and generate semantic annotations of a text based on a positioning algorithm of large model annotations, and performing fine-grained software fault positioning from a file level to a function level and then to a position level by using the annotation-based hierarchical positioning method. And an efficient hierarchical progressive sorting and screening algorithm is used for ensuring the software fault positioning efficiency. According to the method, the defect positioning efficiency can be ensured while the positioning accuracy is ensured. The positioning algorithm based on large model annotation strategically balances the information density, provides enough context clues, and improves the model understanding ability; according to the annotation-based hierarchical positioning method, positioning is divided into file / function / position hierarchies, and annotations are added by using a large model cue word project in sequence, so that the positioning accuracy is improved; and an efficient hierarchical progressive sorting and screening algorithm is provided, so that the optimal balance between the performance and the cost is realized.

The invention discloses a three-dimensional visualization head and neck tumor preoperative imaging system based on deep learning. The method comprises the following steps: image analysis and preprocessing; carrying out image denoising and spatial-temporal feature extraction based on a deep learning convolutional neural network architecture, and introducing a u-Net model to realize image segmentation and semantic annotation; performing three-dimensional reconstruction and surface rendering by adopting a Marking Cubes algorithm, and realizing visual interaction image presentation and personalized operation by combining volume rendering and surface rendering; and data is output through multiple ports, and data viewing and interactive operation of a web terminal and a mobile terminal are supported. The method has the advantages that according to the technical scheme, through fusion of a three-dimensional attention mechanism and multi-phase data, the tumor segmentation Dice coefficient reaches 0.92 + / -0.03, and the method is superior to a traditional threshold segmentation method.

The invention relates to the technical field of medical report generation, and discloses a medical image automatic annotation and report generation system which comprises a preprocessing module, a hierarchical feature extraction module, a semantic annotation module, a guide generation module, a diagnosis reasoning module and a man-machine cooperation interface. According to the method, through standardized preprocessing, hierarchical feature extraction and automatic generation of the annotation and the report, the processing efficiency is greatly improved, and the manual load is reduced; secondly, on the basis of hierarchical feature extraction and semantic and visual embedding alignment, feature support and semantic mapping precision of annotation and reporting are enhanced, and result accuracy and cross-scene consistency are improved; meanwhile, multi-modal alignment processing and multi-dimensional information fusion enhance the adaptability of the system to different image types, generate a structured report conforming to the clinical process and improve the clinical applicability; in addition, a man-machine cooperation feedback mechanism forms a parameter optimization closed loop, so that the system dynamically adapts to new scenes and specifications, and continuous improvement of generalization ability is realized.

The invention relates to the technical field of integrated circuit design automation, and discloses an integrated circuit design whole-process collaborative verification and consistency guarantee method, which comprises the following steps of: constructing a unified design data model, and representing design data from function specifications to physical layout by using a multi-level attribute graph structure; establishing a design intention keeping mechanism, and tracking the transmission of the design intention in the hierarchical refinement process through intention semantic annotation, cross-level entity mapping and intention keeping degree evaluation; executing a hybrid verification strategy, combining formalized verification with dynamic simulation, and guiding verification resource allocation through a coverage rate feedback mechanism; performing constraint consistency check, performing constraint comparison at key nodes of the design process, and automatically identifying and positioning constraint conflicts; according to the method, the capture time of design errors can be brought forward by 80%, the number of design iterations is reduced by 40%, and the design quality and efficiency are remarkably improved.

The invention relates to the technical field of computers, and discloses a visual operation and maintenance supervision platform based on a digital twinning technology, and the platform comprises a data preprocessing module which is used for preprocessing a multi-source heterogeneous monitoring data flow; the data fusion module is used for constructing an equipment topological relation graph, extracting an initial feature vector set of equipment nodes and carrying out space-time fusion; the digital twin modeling module is used for performing parametric modeling and semantic annotation; the real-time synchronization module is used for extracting a model updating instruction set and carrying out asynchronous message transmission and parallel updating; the anomaly detection module is used for training an anomaly detection model and analyzing anomaly types and anomaly reasons; the root cause tracing module is used for constructing an associated sub-graph of abnormal equipment, predicting a root cause candidate equipment list and carrying out hybrid reasoning and evidence fusion; the visual presentation module is used for visualizing a fault influence range, dynamically displaying network traffic and performing multi-view roaming and information drilling; according to the invention, the fault positioning efficiency and the operation and maintenance management level are improved.

The invention provides an intelligent construction site safety monitoring and early warning method and system, and the method comprises the steps: fusing the multi-source data, such as personnel positioning, equipment state and environment parameters, and building a structured time sequence feature set with spatial semantic annotation; constructing a symbolized rule base in combination with a field safety specification; applying a dynamic graph attention network, dynamically modeling personnel, equipment, environment nodes and an interaction relationship thereof, and deducing potential causal association strength between spatial entities; through path semantic distillation and anti-fact intervention, a compliant and accurate multi-stage causal reasoning path is formed, and key risk nodes and propagation channels are visualized in a three-dimensional BIM scene; and finally, dynamic early warning event generation and natural language interpretation are realized, model closed-loop optimization based on field feedback is supported, and the timeliness, interpretation and adaptive capability of construction site safety risk discovery are improved.

The invention discloses an electronic component intelligent classification method and system based on design data, and the method comprises the steps: obtaining a design file of a printed circuit board and a corresponding board image; analyzing based on the design file to obtain bit numbers and layout pose information of the components, determining image areas of the components in the board surface image according to the bit numbers and the layout pose information, and generating standardized component images corresponding to the components one by one; extracting feature vectors from the standardized component image to form a component feature set; executing clustering based on the component feature set to obtain a plurality of category clusters; and displaying representative samples of the category clusters, receiving semantic annotations of users on the category clusters, establishing a mapping relationship between the clusters and the categories, and outputting classification results of the components based on the mapping relationship. According to the technical scheme, the electronic component classification efficiency is improved.

The invention belongs to the field of navigation and positioning, and particularly discloses a robot navigation method and system for enhancing a semantic map. According to the method, the geometric information and the semantic information are extracted based on the pre-stored GIS map file, and the GIS data are generated, so that a structured basis with global geographic information and semantic annotation is provided for subsequent processing; the GIS data and the SLAM map are aligned through hierarchical mapping and feature matching, unified association of local geometric information and global geographic information is achieved, and the limitation of separation of the local geometric information and the global geographic information in a traditional navigation map is overcome; on the basis, a fused semantic map is generated by dynamically associating semantic information in the navigation process, so that the robot can perform perception and reasoning by utilizing environmental semantics in real time; and finally, path planning and decision control are performed based on the fused semantic map, so that the burden of repeated mapping of the robot depending on a single SLAM is remarkably relieved, and the navigation efficiency and the environmental adaptability are improved.

The invention provides a dynamic knowledge graph construction and diagnosis reasoning method oriented to intelligent inquiry, and relates to the technical field of knowledge graphs, comprising the following steps: acquiring multi-source medical data, performing entity recognition and semantic annotation, establishing a causal probability graph based on a structured entity set, and establishing a dynamic knowledge graph; and selecting an optimal questioning problem according to the information gain in the inquiry process, dynamically updating the causal probability by using the Bayesian rule, and finally propagating the conditional probability along the causal path to generate a diagnosis conclusion. According to the invention, personalized inquiry decision and accurate diagnosis reasoning are realized, and the diagnosis accuracy and efficiency of the intelligent inquiry system are improved.

The invention provides an eBPF program automatic generation method, system and device based on retrieval enhancement and thinking chain reasoning, and the method comprises the steps: constructing an eBPF semantic knowledge base, and extracting and structuring a plurality of program examples with semantic annotations; after a natural language task description input by a user is received, the natural language task description is coded into a semantic vector, and accurate retrieval of related contexts is achieved in combination with semantic indexes; based on a retrieval result, constructing a structured prompt, and guiding a language model to gradually generate an eBPF program according to stages; an eBPF verifier is used for carrying out legality check on the generated program, and behavior testing is carried out in combination with task input; when structural errors or semantic deviations are found, the model is guided to be automatically repaired based on error information. The method can be operated on a multi-language model platform, supports the output of two styles of BCC and BPFtrace, and has the advantages of controllable structure, accurate semantics, stable deployment and the like. Experimental results show that the automatic generation efficiency and accuracy of the eBPF program can be remarkably improved.

The invention discloses a multi-table fine-grained retrieval method and device based on large model decomposition reasoning. The method comprises the following steps: firstly, prompting a table cell value and field description of each field of a table to a large model to obtain a field-level semantic annotation; then, relevant table fields are positioned through an analysis-mapping-filling strategy, a large model is used for understanding and decomposing problems, and key information such as entities and conditions is extracted; a field set required by the question is screened out in combination with table structure information; then, based on a field retrieval result, analyzing a value range of the field and further positioning related cells; and finally, integrating the retrieved fields and cells to generate a fine-grained target sub-table which is used for answering user questions. Compared with whole-table-level retrieval, the method only captures required information like human reading, significantly reduces redundant interference, improves precision ratio and response speed, and can be widely applied to medical treatment, finance, complex database management and other scenes with strict requirements for accurate structured data retrieval.

The invention relates to the technical field of medical data processing, and discloses an automatic generation system and method of an electronic medical record editor based on AI. The system comprises a voice acquisition module, a feature extraction module, a structure analysis module, a standard mapping module, a quality control evaluation module, a dynamic optimization module and a closed-loop management module. The voice acquisition module captures a doctor-patient conversation audio stream in real time; the feature extraction module preprocesses audio, separates audio features and extracts key medical information; the structure analysis module identifies medical entities and association relationships through semantic annotation and dependency analysis; the standard mapping module matches the data with a preset medical record template to generate standard medical record paragraphs; the quality control evaluation module analyzes the integrity, consistency and compliance of the medical records according to a multi-dimensional rule base; the dynamic optimization module adjusts a generation strategy weight according to an analysis result; and the closed-loop management module generates a quality control score and triggers automatic defect revision. The system can automatically generate standard medical records and realize dynamic optimization of medical record quality.

The invention discloses a power distribution terminal automatic test method and system based on a distribution network semantic knowledge graph. The method comprises the following steps: acquiring operation data of a power distribution terminal and reading of an environment sensor, performing entity alignment and relation mapping, and generating a multi-modal data set with semantic annotation; extracting association features of semantic entities by using a random forest algorithm in combination with semantic constraints, and constructing a semantic relationship matrix with weight attributes; determining the interference degree of the environmental factors to the entity association, and adjusting the elements of the relation matrix according to the interference degree; and generating a test scene behavior sequence in combination with topological constraints in the knowledge graph, performing fault point prediction by adopting a graph neural network algorithm, positioning a coordinate set based on a problem of probability labeling, identifying an uncovered interaction relation path, and generating a multi-dimensional test coverage evaluation report containing semantic description. According to the scheme, the cooperative work mechanism is evaluated from the system level, the accuracy of the test result in the actual operation environment is improved, and the test coverage range is increased.

The invention discloses a cognitive link diagram modeling and visual expression method, device and system, and belongs to the field of user behavior data modeling visualization, and the method comprises the steps: building a diagram expression and a traceable path of a causal structure based on a user behavior data modeling result; wherein the logic relevance and evolution directivity between entities in the graph are enhanced through a semantic annotation mechanism of nodes and edges of the graph, and the user cognition evolution process in the user behavior data is presented. According to the method, path-traceable graph structure visual expression is supported, and the cognitive evolution process of the user can be presented more finely.

The invention relates to the technical field of artificial intelligence, and discloses a small language webpage adaptive collection method and device based on a large language model.The method comprises the steps that a target small language webpage is loaded and rendered through a headless browser, and a document object model tree structure is obtained; carrying out analysis and semantic annotation on the document object model tree structure based on a large language model, identifying dynamic content nodes and generating a self-adaptive content positioning rule; based on a self-adaptive content positioning rule, performing semantic vectorization on the at least two continuously collected webpage contents, and calculating semantic similarity; querying a preset frequency mapping rule according to the semantic similarity, dynamically adjusting the initiation frequency of a subsequent acquisition request, and generating an acquisition strategy; and executing the acquisition strategy to access the target webpage, and injecting an operation sequence for simulating human interaction behaviors in the acquisition process. Complete collection of the dynamic content of the small language webpage can be achieved, and continuity and stability of the collection process are guaranteed.

The invention relates to the field of remote sensing image processing, and particularly discloses a natural resource normalized monitoring-oriented remote sensing image rapid selection method, which comprises the following steps of S1, acquiring multi-source remote sensing image data, and performing preprocessing to obtain preprocessed image data containing semantic annotation; s2, carrying out coordinate conversion on the preprocessed image data containing the semantic annotation, obtaining feature parameters of the preprocessed image data containing the semantic annotation by adopting a coordinate mapping model based on deep learning, inputting the feature parameters into the coordinate mapping model for targeted correction, inputting an image data path, and outputting the image data path; reading a field name of a data attribute table in the image data path; s3, for the image data after coordinate conversion, firstly inputting data paths of a front batch of falling images and a rear batch of falling images, and reading the front batch of falling image data and the rear batch of falling image data; by adopting the technical scheme of the invention, the image data quality can be improved, the multi-source image coordinate mapping error is reduced, and the result reliability is ensured.

This invention discloses a method for generating personalized sports highlights based on user preferences, specifically relating to the fields of video processing and intelligent recommendation technology. It involves acquiring user viewing and interaction behavior data and locating corresponding video segments, performing event-level segmentation and semantic annotation on the video segments to construct user behavior-event semantic association data. A causal relationship model is introduced, and by intervening in event semantic variables and implementing counterfactual inference, the strength of counterfactual behavior is calculated to exclude the influence of negative semantics. Furthermore, by combining the difference between the original behavior strength and the counterfactual behavior strength, segmented weight calculations are performed, and user behavior features are semantically corrected and reconstructed to generate user interest feature vectors. Finally, these vectors are input into a user profile model to achieve personalized sports highlights generation or content recommendation. This invention effectively removes the interference of negative event semantics on user behavior, improves the accuracy of user interest modeling, and enhances the quality of highlights content and the effectiveness of personalized recommendations.

This invention discloses an AI-interactive AR (Augmented Reality) cultural wall system for cultural tourism, belonging to the field of AR system technology. The surface of the cultural wall is equipped with a pre-embedded feature set, including visual markers, differentiated texture features, 3D depth variation areas, and specifically designed interactive hotspots. These features provide the system with stable visual benchmarks, semantic distinction criteria, depth cues, and precise positioning anchors. The system includes: a feature point acquisition module, used to collect the feature set on the cultural wall through a multimodal feature extraction algorithm to construct a 3D spatial map with semantic annotations; a user positioning and pose estimation module; an AI content generation engine; and a spatial adaptation and rendering module. Through the cultural wall feature set and the multi-module collaborative system, precise spatial integration of virtual content and the physical wall surface is achieved, along with intelligent dynamic content generation based on user behavior, effectively enhancing the immersion, interactivity, and personalization of the AR cultural experience.

The invention relates to the technical field of image processing, and discloses a data annotation method and device, a storage medium and data annotation device.The data annotation method comprises the steps that to-be-annotated data and a trained target model are obtained; preprocessing the to-be-labeled data through a 3D point cloud cross-mode and a preset model to obtain structured data of the to-be-labeled data; inputting the structured data into the target model to determine a semantic mask of the to-be-labeled data; obtaining an uncertainty score of the semantic mask; and when the uncertainty score is greater than a preset score, updating the target model, and performing data annotation by using the updated target model. Therefore, the precision and integrity of semantic annotation in a complex scene are remarkably improved, the model can be dynamically optimized, and manual intervention can be reduced.

The invention relates to the technical field of live broadcast management, in particular to an AI-based field control management system for multiple live broadcast rooms. According to the AI-based field control management system for the multiple live broadcast rooms, interaction data are acquired from the live broadcast rooms through data acquisition, and standardized interaction records are formed through data processing analysis and format conversion; semantic annotation calls a preset language model to carry out semantic recognition and emotion recognition on the interaction text, a user tag is generated in combination with a rule base and is associated with a user identifier, and an annotated interaction data set is obtained; according to instruction generation, a field control instruction is generated based on the annotated interaction data set and the user portrait, scheduling and optimization determine the priority of the live broadcast room and generate cross-live broadcast room resource allocation configuration according to the field control instruction and the running state of the live broadcast room, and the field control management effect of multiple live broadcast rooms is improved.

The invention relates to the technical field of power distribution operation simulation, and discloses a power distribution operation three-dimensional virtual model library construction method based on a domain knowledge graph. According to the method, real-time interactive feedback is realized by constructing a power distribution field knowledge graph and performing semantic enhancement, performing semantic annotation and feature extraction on a three-dimensional model, constructing an intelligent model library and maintaining consistency, intelligently generating an operation scene and performing compliance optimization, and entity and model semantic mapping is established by adopting a semantic topology energy quantification algorithm; evaluating scene quality by using a process constraint optimization function; according to the method, the problems of semantic association deficiency, low scene generation efficiency and poor compliance of a traditional model library are solved, the model retrieval accuracy, the scene generation efficiency and the virtual training effectiveness are improved, and the method is suitable for power distribution operation virtual simulation and personnel training.

This invention discloses an intelligent analysis and early warning cloud database system for big data in water conservancy project operation and maintenance, specifically relating to the field of water conservancy project operation and maintenance technology. The system is based on a cloud-native architecture and includes an intelligent data lake warehouse, a time-series graph fusion storage engine, an embedded model computation module, a collaborative early warning and analysis module, and a self-learning optimization module. The intelligent data lake warehouse performs semantic annotation preprocessing on multi-source heterogeneous operation and maintenance data; the time-series graph fusion storage engine deeply integrates domain knowledge graphs with semantically annotated time-series data sequences, logically binding the time-series data sequences to graph nodes through globally unique identifiers to form a time-series-graph hybrid data model; the embedded model computation module deploys dedicated analysis models; the collaborative early warning and analysis module aggregates multimodal evidence chains, schedules model collaborative computation, and performs fusion analysis to generate early warnings; the self-learning optimization module optimizes the model online based on feedback. This invention achieves data fusion storage and intelligent early warning, improving operation and maintenance efficiency and reliability.

The invention discloses a structured layer module combination system and method for video generation, and relates to the technical field of video shooting and generation, and the method comprises the steps: S1, obtaining N continuous layers to generate a layer image stack, where N is a natural number and is greater than 2, and each layer comprises image data and focal length data of image data imaging of the layer; s2, through image segmentation, target detection and semantic annotation, constructing a layer mark with a spatial level, a shielding logic and a semantic mark, and providing a structural constraint for subsequent generation; s3, processing at least one image layer to generate an animation image layer; and S4, according to the layer mark, synthesizing the animation layer with other layers to generate a video sequence frame. The method has the beneficial effects that each layer has spatial depth, semantic label and shielding logic, and is suitable for structural modeling and expression of a complex scene; the animation can be visually smooth and semantically coherent on the basis of an inter-layer structure and action prediction model.