15 results about "Evidence synthesis" patented technology

The invention discloses a special vehicle welding quality detection system and detection method based on hybrid deep learning and an improved evidence theory, and the system comprises a data collection and preprocessing module, a feature extraction module, an improved GA-XGBoost weld defect classification prediction module, and an improved evidence theory decision-making layer fusion module. Wherein the data collecting and preprocessing module is used for collecting and preprocessing numerical data signals and image signals; the feature extraction module extracts feature vectors of the image through an improved ResNet18 model, and performs feature extraction on the preprocessed numerical data; four XGBoost classifiers are constructed on the basis of an improved GA-XGBoost weld defect classification prediction module, image features and numerical features serve as input, probability distribution of different welding states is output, and an improved genetic algorithm is applied to hyper-parameter optimization of an XGBoost model; defining an identification framework theta based on an improved evidence theory decision-making layer fusion module, taking four classifiers as four evidences of an evidence theory, outputting probability distribution of a welding seam state by the four classifiers as a basic probability distribution value of a corresponding basic credibility distribution function, performing preprocessing by adopting a conflict evidence fusion algorithm in an improved D-S evidence theory, and obtaining a fusion result; performing multi-source information fusion by applying a D-S evidence synthesis rule; according to the method, the welding defects can be efficiently and accurately detected and classified.

The present application relates to the technical field of personnel behavior early warning, and discloses a multi-modal AI real scene three-dimensional personnel abnormal behavior identification early warning system, comprising: a rule parameter module for generating a region rule table and a link parameter table; an evidence archiving module for generating an evidence package by time window archiving the three-dimensional space observation channel, access control state, equipment operation state and region occupancy state; a space gating module for determining a space main condition gating quantity; an evidence synthesis module for generating an alarm evidence and calculating a multi-modal mutual evidence consistency score; a hierarchical judgment module for determining an alarm state; a report linkage module for generating and reporting a structured alarm message and triggering a linkage action; and a log closed loop module for updating an alarm threshold based on an event log. The present application realizes stable hierarchical early warning of personnel abnormal behavior and closed loop self-adaptive adjustment of alarm parameters.

The invention discloses a spacecraft complex service damage high-precision classification method based on multi-modal fusion, and the method comprises the steps: collecting a vibration response signal under ultrasonic excitation through a sensor, and constructing a vibration data set Dvib after the vibration response signal is processed; the method comprises the following steps: acquiring a temperature distribution sequence under photo-thermal excitation through an infrared thermal imager, and processing by adopting a distance interval division and partition enhancement algorithm to construct an infrared data set Dir; based on the multi-modal data Dmul, constructing and training a deep residual network classification model of a vibration mode and an infrared mode; and further constructing a multi-modal fusion test set Dmultest, performing evidence synthesis and decision fusion on an output result of the bimodal classification model by adopting a D-S evidence theory, and finally outputting a high-confidence damage type and grade classification result. Through a targeted data enhancement strategy and a deep fusion mechanism, the limitation of single-mode detection is effectively overcome, and the accuracy and robustness of spacecraft complex service damage classification are remarkably improved.

This application provides a map generation method and system based on multi-source heterogeneous geographic information processing. It acquires three types of heterogeneous data: a geographic element attribute database, a geographic coordinate semantic database, and a geographic reference image database. Spatial spectrum reconstruction processing is then performed to generate a spatial spectrum cube. Belief propagation fusion processing is then used to construct a belief network from the spatial spectrum data. Combined with spatially constrained probability field optimization and D-S evidence synthesis, a geographic evidence chain is generated. Next, spatiotemporal field reconstruction processing is performed to separate the static and dynamic states of the geographic evidence chain and couple them into fields, constructing a hyperdimensional spatiotemporal map. Finally, holographic projection conversion processing is used to perform phase field compression and coherent diffraction reconstruction on the hyperdimensional spatiotemporal map, generating a holographic geographic projection map that supports 3D interaction. This method achieves the automatic generation of a structurally unified, highly reliable geographic map that supports dynamic behavioral reasoning and true 3D interaction from multi-source heterogeneous geographic data, significantly improving the decision-making value and cognitive efficiency of geographic information.

The application provides a kind of array group target data association method and system based on adaptive clustering, comprising: accumulating same load data;Each load data is clustered, and initial array group division is obtained;Single cluster target set is screened;According to array group determination condition, remove non-array group target;The characteristics of array group target of two time points before and after are described respectively, and the characteristics of formation and formation are modeled;According to the array group target association matching model of the characteristics of formation and formation, the characteristics of formation and formation, the corresponding basic probability distribution function is calculated respectively;Evidence synthesis is carried out, and the comprehensive basic probability distribution function is established;Based on two-dimensional distribution, the comprehensive basic probability distribution function is associated with decision, and the array group target matching result is obtained.The application can effectively reduce the influence of array group target motion state change, sensor's missed detection and observation error, and significantly improve the correct association rate of target.

The application discloses a social network information credibility evaluation method based on D-S evidence theory, which comprises the following steps: 1, feature extraction, using BERT pre-training model, ResNet50 neural network, TF-IDF algorithm and cosine similarity method to extract text features, image-text consistency features and title-text consistency features; 2, evidence generation, using the basic probability assignment method and related formula to convert the original evidence into the corresponding BPA; 3, evidence fusion, calculating the evidence clarity, evidence credibility and evidence weight, and combining the evidence synthesis rule to realize the evidence fusion. The method firstly respectively utilizes the text features, image-text consistency features and title-text consistency features to evaluate the information credibility; then converts the credibility evaluation results based on the three single features into evidence, and uses the improved D-S evidence theory fusion rule to make fusion and decision, and finally outputs the final result of the comprehensive credibility evaluation.

The invention relates to the technical field of mine intelligent operation and maintenance, in particular to a mine production operation and maintenance management system. The technical problems of state sensing lag, insufficient diagnosis confidence and unreliable man-machine interaction caused by multi-physics field coupling, strong electromagnetic interference and equipment isomerism in a deep mine roadway when a mine production operation and maintenance management system in the prior art is applied are solved. According to the technical scheme, the mine production operation and maintenance management system comprises an augmented reality terminal, an edge computing module, a digital twin modeling module, a decision optimization module and a data fusion gateway; through the multi-source evidence dynamic fusion and electromagnetic environment adaptive compensation technology, the robustness of the system is improved, the multi-modal evidence synthesis process is optimized based on the variational Bayesian framework, the communication link anti-interference logic is reconstructed in combination with the interference-confidence transfer function, the high-confidence diagnosis of the equipment degradation path and the stable projection of the AR instruction are realized, and the reliability of the system is improved. And the mine safety operation and maintenance efficiency and the equipment reliability are improved.

The application discloses a kind of based on deep reinforcement learning's vehicle networking trusted data sharing method and system thereof.The application includes trust information modeling module, trust evaluation module, trust policy updating module, threshold comparison module.The application first collects vehicle networking real-time context information, then input context information into the trust evaluation module based on evidence theory, using the comprehensive weighted average of direct trust degree and indirect trust degree to calculate global trust value.In order to improve the effectiveness of conflict evidence source synthesis, the application proposes a conflict evidence synthesis method based on Pignistic distance to ensure that false messages do not affect driving decisions.Second, in order to improve the accuracy and robustness of malicious vehicle detection, the application designs a trust policy updating model based on deep reinforcement learning, which can learn and optimize the decision strategy of the vehicle detection system through reinforcement learning algorithm.

This invention discloses a method and system for detecting hate content in videos based on multi-agent collaboration. The detection method includes: S1 Data preprocessing: performing multimodal data processing on the input video content; S2 Opposite perspective agent debate: setting up supporter and opponent agents, both of which are used to capture semantic cues in the video content and generate explanatory evidence; S3 Evidence synthesis: setting up a synthesis agent to integrate the explanatory evidence and generate a comprehensive explanation; S4 Reason enhancement prediction: fusing multimodal features and reason features, and performing prediction through a two-layer multilayer perceptron; S5 Hate score calculation: calculating a hate score based on the predicted content, and determining whether the video content contains hate content by comparing it with a preset threshold. This invention, through multi-agent collaboration and reason enhancement prediction mechanism, can effectively identify subtle hate content in video content, while providing a credible and comprehensive explanation.

The invention discloses an arbitration evidence construction method and device, an arbitration evidence processing method and device and a storage medium, and relates to the technical field of computers.The arbitration evidence construction method comprises the steps that original transaction data of multiple data sources is acquired, and federated learning features of the original transaction data are extracted through a federated learning method; generating a corresponding transaction data hash value based on the original transaction data; determining a plurality of ownership credibility evaluation indexes, and determining a dynamic quantitative score of the current transaction according to the ownership credibility evaluation indexes; and constructing an arbitration evidence packet according to the federal learning features, the transaction data hash value and the dynamic quantitative score. According to the method, the arbitration evidence is constructed by combining the multi-dimensional transaction data features, the dynamic quantitative score and the transaction data hash value, and multiple factors influencing arbitration are comprehensively considered, so that the arbitration evidence construction reliability and arbitration reliability can be effectively improved.

The invention relates to a storage tank life evaluation method based on test and flight data fusion. The method comprises the following steps: (1) respectively carrying out denoising and standardization processing on collected test and flight data; (2) extracting life correlation characteristics by adopting a principal component analysis method; (3) constructing a double-source data fusion model based on an improved D-S evidence theory, and fusing the extracted core features to obtain a comprehensive feature vector; and (4) constructing a life evaluation model based on fusion features, inputting the comprehensive feature vector to obtain a single machine residual life evaluation result, and optimizing model parameters through cross validation in the process. According to the method, data processing and reliability evaluation theories can be combined, a double-source data fusion theoretical method is formulated, evidence synthesis and decision rules are established, existing data are fully utilized to realize damage accumulation quantification and accurate life evaluation, and technical support is provided for ensuring task reliability, reducing operation cost and realizing high-frequency emission.

The invention relates to the technical field of personnel behavior early warning, and discloses a multi-modal AI live-action three-dimensional personnel abnormal behavior recognition early warning system, which comprises a rule parameter module, a data processing module, a data processing module, a data processing module and a data processing module, wherein the rule parameter module is used for generating a region rule table and a link parameter table; the evidence archiving module is used for performing time window archiving on the three-dimensional space observation channel and access control state, the equipment operation state and the area occupation state to generate an evidence packet; the space gating module is used for determining a space main condition gating quantity; the evidence synthesis module is used for generating alarm evidence and calculating a multi-mode mutual identification consistency score; the grading judgment module is used for determining an alarm state; the reporting linkage module is used for generating and reporting a structured alarm message and triggering a linkage action; and the log closed-loop module is used for updating the alarm threshold value based on the event log. According to the invention, stable grading early warning of abnormal behaviors of personnel and closed-loop adaptive adjustment of alarm parameters are realized.

A brain-like auxiliary diagnosis and treatment method for rare diseases is realized based on a GPU or NPU server, a system review agent is used as a sensing front end of the method, is connected with an external medical database through a standard API interface, executes automatic retrieval and quality preliminary screening of literatures, and outputs an original literature set; the document processing agent is used for receiving an original document set of the system review agent, performing structured extraction on an unstructured PDF document by utilizing a deep analysis tool, and generating a structured knowledge base containing a core clinical field; and the NHSC brain-like agent adopts a brain-like architecture based on a dual processing theory, receives the structured knowledge base and the patient data of the document processing agent, and outputs a final dynamic diagnosis and treatment suggestion through fast and slow dual-channel cooperative processing. Algorithm evaluation of evidence quality is realized, an evidence screening mechanism is enhanced from the source, risks caused by illusion of a large language model are effectively reduced, and stability and reliability of an evidence synthesis result are guaranteed.

The application introduces a kind of tobacco monopoly administrative punishment case file intelligent evaluation system based on multi-modal evidence chain analysis, aiming at solving the deficiencies of the prior art in the aspects of evidence chain time sequence analysis, multi-modal evidence comprehensive judgment and legal application accuracy. The system integrates optical character recognition (OCR), natural language processing (NLP) technology to automatically extract key information in the case file materials, and constructs a time sequence evidence graph to detect evidence missing and logical conflicts. The legal knowledge graph module dynamically stores and updates tobacco monopoly laws and regulations clauses, ensuring that the case facts can be accurately mapped to specific legal clauses. By adopting the three-layer architecture design of "evidence graph + rule engine + knowledge graph", the application realizes the comprehensive judgment of multi-modal evidence, greatly improves the efficiency and accuracy of case file evaluation, reduces the subjectivity and error rate of manual evaluation, and provides strong technical support for tobacco monopoly administrative punishment.