1123 results about "Black box" patented technology

The invention discloses a wide-area landslide rapid identification method based on an interpretable intelligent algorithm, and relates to the field of remote sensing science and technology, and the method comprises the steps: building a dual-channel feature extraction architecture through multi-source spatio-temporal data fusion and knowledge graph dynamic weighting: capturing image local textures through a lightweight CNN, modeling geological spatial correlation through a graph convolutional network, and carrying out the recognition of the landslide. Combining the SHAP value and causal reasoning to generate an interpretable contribution degree thermodynamic diagram and a rule chain; a knowledge graph bidirectional verification system is introduced, spatial logic contradictions are verified by using prior rules, and co-evolution of a model and a rule base is triggered based on misjudgment samples; outputting a multi-dimensional credibility report, quantifying uncertainty by Monte Carlo Dropout, and customizing interpretation granularity according to roles; a terrain-adaptive block-stream processing architecture is adopted, and edge lightweight deployment and federated learning are combined, so that wide-area real-time early warning and model dynamic updating are realized. According to the scheme, the limitation of a traditional black box model is broken through, and a disaster prevention closed loop with physical driving, transparent decision and second-level response is formed.

The invention relates to a multi-modal AI data fusion processing method, device and equipment and a medium, and the method comprises the steps: firstly extracting visual, auditory and text modal features through a pre-training encoder, executing dimension alignment, and generating a standard data feature set with unified dimensions; a cross-modal semantic graph is constructed based on a cosine similarity algorithm, and the problem of semantic mismatch of heterogeneous data is solved; residual enhancement is carried out on the map nodes, and noise interference is eliminated; fusing the optimized features and the semantic topology in combination with a graph convolutional network to generate aggregation graph representation; the fusion features are mapped to a low-dimensional semantic space through a variational auto-encoder, and cross-modal correlation essence is captured; the key dimension contribution degree is quantified, a visual report is generated, and semantic association rules among modals are disclosed, so that the dimension isomerism limitation of a traditional fusion technology is broken through, quantifiable cross-modal semantic mapping is established, the whole process traceability from feature fusion to decision interpretation is realized, and the method is suitable for popularization and application. And the multi-modal decision black box problem in the fields of medical diagnosis, automatic driving and the like is effectively solved.

The invention discloses an interpretability prediction analysis system for multivariable time series data. According to the method, the prediction precision and the decision support capability of the complex time series data are remarkably improved through multi-module cooperation. Firstly, an adaptive learning optimization module dynamically adjusts model parameters and a prediction strategy, so that the model can quickly adapt to time-varying characteristics of data distribution, for example, when a causal relationship between variables suddenly changes, the weight of latest data is automatically enhanced, and historical noise interference is reduced. The dynamic causal interpretation engine tracks the influence intensity and hysteresis effect of key variables in real time, converts traditional black box prediction into a traceable causal relationship chain, and helps a user to intuitively understand driving factors of a prediction result, for example, it is identified that prediction value sudden increase in a certain period is mainly derived from hysteresis effect accumulation of an upstream variable A.

The invention provides an intelligent landslide identification method and system based on hierarchical frame selection and boundary feature fusion, and the method comprises the steps: obtaining a target remote sensing image data set, carrying out the hierarchical frame selection of the target remote sensing image data set, obtaining a hierarchical frame selection region set, carrying out the boundary feature extraction of the hierarchical frame selection region set, and obtaining a boundary feature extraction result; obtaining a boundary feature set of each hierarchical framed area, inputting the boundary feature sets into a pre-trained boundary feature fusion model for feature fusion processing to generate a fused boundary feature set, performing feature comparison processing with a preset landslide morphological feature library based on the fused boundary feature set to generate a landslide identification feature set, and performing landslide identification on the landslide identification feature set. And performing region labeling processing on the target remote sensing image data set according to the landslide recognition feature set to generate a landslide region recognition result. According to the invention, through topology reconstruction of hierarchical frame selection and feature fusion, a landslide identification link with a self-interpretation capability is formed, and the problem of misjudgment accumulation of the model caused by feature black box transmission is avoided.

The invention discloses a battery degradation model construction method based on a Bayesian physical information neural network, and the method comprises the following steps: constructing a pseudo-two-dimensional battery data generation module based on a battery aging mechanism; designing a feature extraction network to extract IC feature parameters; constructing an aging parameter mapping network to describe the relationship between the IC characteristic parameters and the battery aging parameters; and constructing a Bayesian battery health state inference network in a parameter randomization mode. By adopting the battery degradation model construction method based on the Bayesian physical information neural network, while explicit mapping of IC features and aging parameters is realized, physical residual constraints are constructed by using an electrochemical equation in the battery, so that the battery degradation model has relatively high physical interpretability, and the risk of a black box model is avoided; through a Bayesian framework based on random parameter distribution modeling, uncertainty in a model modeling process is quantified, so that confidence quantification of model prediction is realized, and a risk sensitive decision is supported.

The invention discloses a two-dimensional wave spectrum intelligent forecasting method based on a physical information neural network, and belongs to the technical field of ocean information prediction.The two-dimensional wave spectrum intelligent forecasting method comprises the steps that basic data are collected, and the basic data comprise high-temporal-spatial-resolution sea surface wind field data and sea wave height data within the research range; historical sea wave data are obtained, ocean and meteorological data closely related to sea wave changes are integrated and processed, and a sea wave forecasting database is constructed; physical constraints are determined according to the sea wave forecasting database; an ANN neural network model is adopted as a basic model, physical constraints are added, the neural network meets the control equation of the mode, and a PINN physical information neural network is constructed; and carrying out interpretability analysis on the model forecasting process and result. According to the method, the sea wave dynamic process and the deep learning method are fused, interpretability is added for deep learning, and the black box problem of the deep learning method is solved.

Methods and systems for managing data processing systems are disclosed. A data processing system may include and depend on the operation of hardware and / or software components. Inference models may be implemented to predict future system infrastructure outcomes (e.g., component failures) using information recorded in logs that reflect the operation of the components. However, the models may be complex “black boxes” and may generate critical outcome predictions for downstream consumers without explanations of how the predictions are determined, resulting in downstream consumers having low confidence in the predictions. Therefore, hidden knowledge (e.g., structured knowledge attributes) of the models may be extracted and / or used to understand the underlying processes that the models use to predict the system infrastructure outcomes. The hidden knowledge may be provided for interactively managing data processing system(s) failures in order to increase the likelihood of preventing and / or mitigating future data processing system failures.

The invention discloses a big language model generation content security test system and method in a black box scene. The system comprises a jailbreak prompt word library module used for storing jailbreak prompt words for performing security test on a big language model; the violation question and answer pair module is used for storing violation question and answer pairs covering different types; the response acquisition module is used for obtaining a data request packet according to query content formed by the jailbreak prompt word and the query request; the security analysis module is used for calculating the similarity between response data corresponding to the query request and an expected violation answer, taking the similarity as a security score, and inputting the security score into the adaptive optimization module; and the adaptive optimization module is used for optimizing the jailbreak prompt words output by the jailbreak prompt word bank module by using a genetic algorithm according to the security score output by the security analysis module. According to the method and the device, the security of the large language model generation content can be effectively tested.

Methods and systems for managing data processing systems are disclosed. A data processing system may include and depend on the operation of hardware and / or software components. Inference models may be implemented to predict future system infrastructure outcomes (e.g., component failures) using information recorded in logs that reflect the operation of the components. However, the models may be complex “black boxes” and may generate critical outcome predictions for downstream consumers without explanations of how the predictions are determined, resulting in downstream consumers having low confidence in the predictions. Therefore, hidden knowledge (e.g., structured knowledge attributes) of the models may be extracted and / or used to understand the underlying processes that the models use to predict the system infrastructure outcomes. The hidden knowledge may be provided for interactively managing data processing system(s) failures in order to increase the likelihood of preventing and / or mitigating future data processing system failures.

Methods and systems for managing data processing systems are disclosed. A data processing system may include and depend on the operation of hardware and / or software components. Inference models may be implemented to predict future system infrastructure outcomes (e.g., component failures) using information recorded in logs that reflect the operation of the components. However, the models may be complex “black boxes” and may generate critical outcome predictions for downstream consumers without explanations of how the predictions are determined, resulting in downstream consumers having low confidence in the predictions. Therefore, hidden knowledge (e.g., structured knowledge attributes) of the models may be extracted and / or used to understand the underlying processes that the models use to predict the system infrastructure outcomes. The hidden knowledge may be stored in a repository and may be provided for downstream use in order to increase the likelihood of preventing and / or mitigating future data processing system failures.

The invention discloses an intelligent agent network access evaluation method and system based on automatic label generation and multi-dimensional evaluation, and the method comprises the steps: automatically generating an intelligent agent multi-dimensional label through a large-scale language model, constructing an intelligent agent portrait, and combining automatic test case generation and multi-dimensional evaluation index collection. The problems that in the prior art, agent evaluation standardization is lacked, evaluation efficiency is low, and'black box 'diagnosis is difficult are solved, standardized and automatic quality evaluation is carried out on various heterogeneous agents connected to a platform, and an agent portrait capable of being used for accurate retrieval and scheduling is generated. The method and the system can be widely applied to scenes such as a multi-agent collaboration platform, an agent store and an enterprise-level agent management system, and have the remarkable advantages of remarkably improving the ecological quality of the agent, greatly reducing the manual evaluation cost, accelerating the introduction and commercialization of a new agent and the like.

The invention belongs to the technical field of cross-border data security, and provides a cross-border personal data transmission security detection method and system based on differential privacy, and the method comprises the steps of multi-dimensional situation awareness, privacy budget allocation, interpretability and audit, and execution and response. By synchronously sensing technical environment risks, multi-national regulation constraints and multi-modal data features, comprehensive control of the cross-border transmission security situation is realized, and one-sidedness of single-dimensional sensing is avoided; privacy budget is dynamically allocated under multiple constraints based on a deep Q network algorithm, so that the protection intensity of high-sensitivity data is ensured, and excessive loss of effectiveness of low-sensitivity data is avoided; an interpretable report is generated in combination with causal analysis, full-process evidence storage is realized by relying on an alliance chain, the problems of decision black box and audit tracing are solved, cross-border supervision requirements are met, and the balance of cross-border transmission in the aspects of safety, compliance and availability is improved.

A data-driven autonomous driving decision optimization system and method, comprising: a data production module, a data screening module, a model encapsulation module, and a parameter tuning module; the data production module takes human driving data as input, annotation, preprocessing, format conversion, extracts key features and performs standardization, normalization, and encoding to generate raw data for the data-driven process that meets the algorithm input requirements; the data screening module screens corresponding data from training data by the decision-making module and performs effective classification; the model encapsulation module encapsulates C++ decision code and constructs a trajectory-pair evaluation cost map required for training using a ground-truth evaluation method based on trajectory pairs; the parameter tuning module, based on screened data under different scenarios, the encapsulated decision algorithm model, and the trajectory-pair evaluation cost map, employs black-box optimization to obtain decision parameters for the corresponding scenarios under the current decision algorithm.

The invention provides a port facility management and maintenance large model report review agent construction method and system, and the method comprises the steps: collecting a cross-modal original data set, constructing a multi-modal feature fusion perception layer, and generating facility damage feature alignment data; constructing a cognitive neural network four-level architecture, and generating an inference decision tree; constructing a root cause-path-result causal chain, and generating a fault attribution analysis report; constructing a prediction-intervention-verification active defense closed loop, and generating a Pareto optimal maintenance strategy set; and executing an intervention strategy and feeding back a verification result by using the digital twin verification platform and the block chain evidence storage system. According to the method, cross-modal data deep semantic alignment is realized through the multi-modal feature fusion perception layer, a data island is broken, and the damage feature extraction accuracy is improved; an interpretable causal chain is constructed based on related architecture and modules, the decision black box problem is solved, and a maintenance strategy has causal logic support; and real-time verification and credible tracing of a strategy effect are realized through an active defense closed loop.

The invention relates to the technical field of intelligent management of oilfield water injection systems, and particularly discloses a water injection pipe network operation situation analysis and energy consumption evaluation method and system. According to the method, real-time data and pipe network parameters of an oilfield water injection system are collected, and after anomaly detection and data cleaning treatment, a node pressure matrix and pipe section flow distribution are generated; performing residual analysis and time sequence modeling by combining real-time data and a simulation result to realize accurate positioning of operation abnormity; a fuzzy analytic hierarchy process for dynamic weight adjustment is adopted to construct a four-stage energy efficiency evaluation system of water injection station-pipe network-water injection well-overall system, and an energy consumption bottleneck thermodynamic diagram is generated; and finally outputting a strategy rehearsal report and a fusion cockpit. According to the method, the change of the water injection pipe network from'black box 'to'transparent' management is realized, the technical problems of single energy efficiency evaluation, scheduling dependence on experience and the like in a traditional method are solved, the system operation efficiency can be remarkably improved, and the energy consumption is reduced.

The invention discloses an urban drainage system multi-target prediction method based on an SWMM and a graph convolutional neural network, and the method comprises the steps: obtaining the actual measurement data of a target urban drainage system pipeline and an inspection well, constructing an SWMM model, and outputting the data; generating a space-time diagram sequence deep learning network GraphSAGE-GRU model on the basis of a graph convolutional neural network GraphSAGE and a gated cycle unit network GRU; and taking the preprocessed data as input, inputting real-time or predicted rainfall data by utilizing the trained model, and synchronously outputting node water head and pipeline load prediction results of the target urban drainage system. According to the method, the problem that the topological structure of an urban drainage system is not considered in a traditional method can be solved, and the hydraulic attributes of the inspection well and the pipeline can be comprehensively output. The problem that a traditional agent model is too black and lacks structural information is solved, and more simulation result output is provided by setting a space-time diagram structure.

The invention relates to the technical field of computer systems based on specific calculation models, and discloses a control method and system for an intelligent calculation center comprehensive energy supply device, and the method comprises the steps: collecting the operation data of energy supply equipment and an environment sensor in real time through an SCADA system, and carrying out the preprocessing; based on the collected data and simulation data generated by a simulation environment, performing offline mixed training on the reinforcement learning model to generate an energy scheduling strategy; inputting a real-time state into the trained reinforcement learning model to generate a preliminary scheduling instruction, performing security verification and interpretability analysis by using a large language model, and optimizing a strategy; and fusing the preliminary scheduling instruction with the suggestion of the large language model, generating a final scheduling command through the energy router control unit, and issuing the final scheduling command to the energy supply equipment for execution. The problems that in the prior art, black box decision making and simulation are not accurate, and experience is difficult to solidify are solved, and the purposes of decision making transparency, simulation high fidelity and experience structuring are achieved.

The invention discloses a heterogeneous intelligent computing power optimization management scheduling system for accelerating a large model reasoning task, and relates to the technical field of computing power optimization management scheduling. The video memory fragmentation problem in a long sequence scene is converted into a controllable block migration task, and the performance bottleneck of a traditional video memory exchange mechanism is broken through; based on an operator-level scheduling strategy of a hardware capability fingerprint database, position coding and other compute-intensive tasks are accurately matched with vector instruction set hardware, and resource mismatch loss caused by black-box scheduling is eliminated; an expert selection process is reconstructed by an integer routing and counting sorting algorithm, near-lossless reasoning is realized at a limited node of an instruction set, and the potential value of an old computing power pool is activated.

The invention relates to the field of man-machine interaction, and particularly discloses a multi-mode man-machine interaction application system for sky-eye insight based on a pathologist visual angle, which is characterized in that firstly, a candidate focus thermodynamic diagram is generated through rapid scanning of a full-slice image, low-power lens global browsing of a doctor is simulated, and the doctor is guided to lock a key area interactively through professional judgment; therefore, the processing efficiency of the oversized image is greatly improved. And then, the system only performs high-resolution deep analysis on the focus confirmed by the doctor, and performs multi-modal fusion on the extracted microscopic visual features and the patient text information to generate a preliminary report with an interpretable basis, so that the problem that the multi-modal function deviates from a clinical core task is solved. Finally, the doctor can check and finalize the report through visual interaction, and the dominant position and the final decision making right of the doctor in the diagnosis process are ensured, so that the bottlenecks of black box operation and low clinical acceptability of a traditional AI system are overcome.

The invention relates to the technical field of training data auditing, in particular to a semantic perception black box large language model training data auditing method and system.The method comprises the following steps that content is returned based on a text generation interface, lexical elements and candidate content are analyzed for multi-round sampling distribution, and a convergence index is determined; and calculating semantic paths and weights of the real content and the candidate content, and analyzing differences between the real content and the candidate content and the reference content to obtain an affiliation adaptation judgment result. According to the method, through multi-round sampling behavior fluctuation tracking, section feature dynamic extraction, stability change judgment and sequence-level weight aggregation, fine separation of training data attribution signals is achieved, a multi-level judgment system is constructed for complex expression and diversified output, judgment accuracy is optimized through a tension comparison signal group and an attribution adaptation judgment mechanism, and the judgment accuracy is improved. And a sensitive and adaptive training data member auditing strategy is formed, so that the risk hidden danger omission is effectively prevented, and the model data security boundary control is enhanced.

The invention discloses an urban flood prediction method based on a dual-drive urban flood model, and the method comprises the steps: carrying out the early-stage preparation of model construction, constructing an urban flood hydrological and hydrodynamic coupling model frame, constructing a deep learning model frame based on a GBDT algorithm, and carrying out the prediction of the urban flood. Assimilation of predicted values and measured data of a hydrological hydrodynamic model and a deep learning model is realized through a real-time data assimilation technology, then an output result of an urban flood hydrological hydrodynamic coupling model is used as an input feature of the deep learning model, and the input feature and parameters are dynamically adjusted according to the matching degree of a confusion matrix. The TP in the confusion matrix is maximum, the TN in the confusion matrix is minimum, finally, construction of the dual-drive urban flood model is completed, and prediction is conducted through the model. According to the method, the characteristic of high calculation efficiency of the deep learning model is exerted while calculation accuracy is considered, a layered coupling architecture is provided, and the problems that the calculation efficiency of a hydrological hydrodynamic model is low and a traditional deep learning model has a black box effect are solved.

The invention relates to a trajectory prediction model robustness enhancement method based on dynamic subspace projection decomposition. Comprising the following steps: firstly, extracting hidden layer semantic features containing historical tracks and map topology through a multi-modal feature encoder; secondly, constructing a dynamic routing mechanism based on scene self-adaption, and calculating projection weights of input features on a plurality of expert subspaces; then, executing truncation projection operation based on orthogonal decomposition, retaining core semantics located in a low-dimensional space, and filtering out adversarial disturbance located in an orthogonal complementary space; and finally, introducing a feature consistency constraint training mechanism, taking the reconstructed features of the clean sample as anchor points, and compulsively aligning the purified features of the confrontation sample with the anchor points. Compared with the prior art, the method has the advantages that the robustness of the model in white box gradient attack, black box query attack and physical semantic deception scenes is remarkably improved through feature purification of a physical level and structured consistency constraint, and the prediction reliability of the automatic driving system is ensured.

The invention is suitable for the technical field of logging lithology identification, and provides a logging lithology identification method based on physical information constraint, which comprises the following steps: step 1, data processing; 2, constructing a graph; step 3, constructing a graph attention neural network; 4, constructing a mixed loss function; and 5, lithology prediction. According to the method, through the synergistic effect of data driving and physical information, the performance is obviously superior to that of various pure data driving models. Benefited from the strong regularization effect of physical constraints, the method has stronger generalization ability and prediction stability on new data. The problem that a traditional black box model may generate a result violating physical common knowledge is fundamentally solved, and the geologic rationality of an output result is ensured.

The invention discloses a lightweight knowledge graph rapid construction method and system based on an NLP technology. The method comprises the steps of preprocessing an unstructured text and segmenting the unstructured text into semantic segments; unsupervised clustering is combined with the contour coefficient to determine the optimal clustering number, and a semantic association fragment cluster is obtained; performing word segmentation, part-of-speech tagging, NER and entity linking on the fragment cluster, extracting an entity and initial relationship, and fusing semantic similarity, TF-IDF word frequency collaboration degree and co-occurrence frequency to calculate a relationship edge weight; constructing a lightweight knowledge graph; in the question and answer stage, questions are disassembled through a process engine, related sub-graphs are retrieved, and answers with reasoning links are generated. The system correspondingly comprises a text preprocessing module, a clustering module, an entity relation processing module, a graph construction module, a question and answer reasoning module and a storage module. According to the method, the construction cost is reduced, the interpretability and the module coupling degree are improved, multiple scenes such as government and enterprise public opinions and medical assistance are adapted, and the problems of weak generalization, poor real-time performance and'black box 'in the traditional technology are solved.

The invention discloses a sensitive cue word generation method and system based on parameter sensitivity quantification, relates to the technical field of artificial intelligence security, and aims to solve the problem that sensitive test samples are difficult to generate in integrity verification of a black box large language model. The method comprises the steps of obtaining an original large language model; a comprehensive parameter sensitivity index is constructed, the microscopic sensitivity used for representing tiny parameter modification and the macroscopic sensitivity used for representing large modification such as parameter quantification and pruning are fused, and therefore the sensitivity of cue words is comprehensively quantified; in the continuous embedding space, a gradient optimization algorithm is adopted to maximize the comprehensive index as a target to carry out iterative optimization, and constraints such as semantic rationality are applied to ensure that the generated cue word is natural and smooth; and finally, mapping the optimized embedded vector back to the discrete lexical element sequence to obtain a final sensitive cue word. According to the method, the low-cost and automatic test sample generation is realized, and the accuracy, efficiency and concealment of cloud model integrity verification are remarkably improved.

The invention relates to the technical field of infrared light machine system thermal deformation prediction, in particular to a multi-material structure thermally induced stress deformation prediction method based on a graph neural network. The method comprises the steps of data set establishment, graph structure establishment, graph neural network model establishment and training and model and parameter optimization. Finite element nodes correspond to graph nodes, finite element edges correspond to graph edges, an encoder-message passing-decoder architecture model is established, and node states are updated through a three-layer physical symmetry message passing mechanism. Physical constraint loss including minimum displacement smoothness constraint and stress continuity constraint is innovatively added into a loss function. Compared with traditional finite element calculation, the method has the advantages that the speed is increased by more than 100 times, high hardware adaptability is achieved, the black box limitation of a data-driven neural network model is broken through, thermally induced stress deformation analysis caused by different material coefficients can be processed, the adaptability to geometric changes is high, and good engineering application value is achieved.

The invention discloses a mobile phone thermal simulation method and system based on multi-physics field coupling, relates to the field of mobile phone thermal simulation, and constructs hybrid simulation coupling machine learning and physical solution. A machine learning agent model and a throttling logic script are innovatively introduced. The machine learning agent model is used for quickly predicting the temperature according to the current power consumption so as to instantaneously respond to the power consumption change; and the throttling logic script is used for simulating a real temperature control frequency reduction strategy and dynamically adjusting the target power consumption at the next moment according to the predicted temperature. In this way, the original black box temperature control logic is explicit, and a power consumption-temperature closed-loop feedback path is constructed. And finally, taking the adjusted power consumption as a heat source, and performing physical field solving by a traditional CAE solver. Therefore, the problem of time scale difference between heat conduction and electrical change is effectively solved, and efficient and accurate simulation of true performance and temperature performance of the mobile phone in a long-time high-load scene is realized.

The invention relates to the technical field of battery health management and prediction, in particular to a lithium battery capacity fading trend prediction method and system, and the method comprises the steps: enabling data driving features and electrochemical features to generate deep fusion features through a dual calibration mechanism; inputting the feature sequence into an electrochemical process perception model, and encoding the feature sequence into a potential state vector representing personalized decline; and inputting the vector as an initial condition into a neural differential equation model, and finally generating a continuous health state attenuation trajectory and obtaining a prediction result by learning a decline dynamic law and performing integral solution. According to the method, deep fusion is carried out on a physical mechanism and a data driving model, the defect that a traditional black box model is weak in generalization ability is overcome, long-term, high-precision and continuous prediction of the full-life-cycle decline trend can be achieved only through early weak signals of a battery, and the reliability and the practical value are high.

The invention discloses an electric power meteorological prediction method and device based on adaptive enhancement, and relates to the technical field of electric power meteorological prediction, and the method comprises the following steps: collecting historical meteorological data and electric power data, and cleaning the collected data; according to the method, a model reasoning path tracing module is constructed based on Chain-of-Though logic, a prediction result is decomposed into subtasks such as meteorological factor contribution degree and power system response logic, interpretability of the prediction process is achieved, reliability of a generated result is evaluated by adopting sequence probability confidence, a visual interpretation report is dynamically generated, and the prediction efficiency is improved. According to the method, operation and maintenance personnel can visually acquire key information, sensitivity of the model to key meteorological events is optimized through adversarial training, prediction capability of the model to extreme meteorological conditions is improved, the trust problem of the black box model in power decision is finally effectively solved, a visual and reliable decision basis is provided for the operation and maintenance personnel, and the operation and maintenance efficiency is improved. And the operation safety and stability of the power system under the complex meteorological condition are improved.

The invention provides an electric power line defect detection method and system based on visual identification, and relates to the technical field of line detection.The method comprises the steps that firstly, a visible light and infrared image dual-light registration and differential operation technology is adopted, and a fusion feature map capable of reflecting component thermal anomaly and material difference at the same time is generated; secondly, a black box type target detection model is abandoned in a part positioning link, but line segment screening and reconstruction are carried out by combining probability Hough transform with specific prior geometric knowledge of a power line, so that dependence on a large amount of labeled data is reduced, interpretability of a positioning process is enhanced, and the positioning accuracy is improved; according to the method, accurate areas of key components such as wire insulators can be extracted in a complex background, the concept of a probability saliency map is introduced in a defect identification core link, so that a defect area is enhanced and highlighted, and then a complete defect contour is determined by adopting an adaptive threshold segmentation and area growing algorithm; accurate mapping from pixel-level features to object-level defects is realized.