1869 results about "Feature mapping" patented technology

This invention relates to a gait emotion recognition method, system, storage medium, and computer device based on spatiotemporal graph convolution. The method includes the following steps: S1, data augmentation by reversing the temporal direction of gait; S2, obtaining deep emotion features and prior emotion features respectively through a spatiotemporal graph convolutional network and prior feature statistical methods; S3, performing nonlinear mapping on the prior emotion features using a feature mapping layer; S4, inputting the fused features of the deep emotion features and prior emotion features into an emotion classifier to obtain the emotion category. The feature mapping layer of this invention achieves more effective feature fusion by performing nonlinear mapping on prior features; it also introduces causal temporal convolution to replace general temporal convolution, effectively extracting fine-grained temporal features by enhancing temporal correlation and cross-period feature fusion. Furthermore, a walking direction recognition auxiliary task is designed to accelerate the training and convergence speed of the model, enhancing the ability to extract temporal-dependent features and the performance of emotion recognition.

The invention discloses a multi-modal fused intelligent question answering and knowledge retrieval method and system, and the method comprises the steps: building a multi-modal data index model oriented to a heterogeneous knowledge source, carrying out the feature mapping of text, image, table, chart, audio and video contents through a unified semantic embedding space, and generating a cross-modal index set; after a query request is received, performing semantic matching and structure matching on the cross-modal index set by using a multi-channel retriever to obtain candidate evidence fragments; and based on an evidence granularity decomposition strategy, performing minimum evidence unit division on text statements, table units, chart data points and multimedia frame contents in the candidate evidence fragments, and establishing a semantic consistency graph among the units. According to the method, high-credibility traceable generation of question and answer results is realized through multi-modal fusion and space-time consistency constraint, and the retrieval precision and interpretation transparency in a complex knowledge scene are remarkably improved.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as financial science and technology and medical health, and discloses a decision-making method and device guided by a multi-modal semantic map, equipment and a medium. Extracting a visual feature vector, a language feature vector and an action feature vector, splicing to generate a multi-modal initial feature, mapping the multi-modal initial feature to a shared semantic space, constructing a multi-modal semantic map, and inputting a map-guided attention mechanism to generate a cross-modal alignment feature; the cross-modal alignment features and task targets are input into a meta-learner to generate task adaptability features, the task adaptability features are input into a parallel reasoning network to execute subtasks in parallel, and a gating fusion network integrates output results to generate a global decision. According to the method, cross-modal semantic association and task adaptability are enhanced through the combination of shared semantic space mapping, map guiding attention and a meta learning device, and the accuracy and efficiency of multi-modal decision making are improved through the combination of parallel reasoning and gating fusion.

The invention relates to an underwater image enhancement method based on wavelet Mama, which aims at the problems of color shift, low contrast ratio and fuzzy details of an underwater image, extracts shallow layer features of an underwater low-quality image, inputs the shallow layer features into a multi-scale coding network, and performs joint enhancement on low-frequency color information and high-frequency detail features by using a wavelet Mama unit. Global semantic features are obtained by combining down-sampling layer-by-layer compression, and global modeling of color offset correction and contrast enhancement is completed; the spatial resolution is recovered through up-sampling, and reconstruction and enhancement of texture details and color information are realized in combination with jump connection and a wavelet Mama unit; and finally, the features are mapped to an image domain and fused with input image residual errors, and an enhanced underwater image with natural color, clear texture and balanced contrast is generated. According to the method, frequency domain-space domain joint feature modeling is realized through the wavelet Mama unit, and the color authenticity, the structural definition and the detail perceptibility of the underwater image are remarkably improved.

The invention relates to the technical field of nondestructive testing, and particularly discloses an automatic detection system and method based on a titanium plate welding part, and the method comprises the steps: obtaining an original physical field signal of a to-be-detected part under the excitation of a single energy field, and extracting a space energy attenuation gradient and time phase lag distribution through wavelet packet decomposition and Hilbert transform; constructing a dynamic propagation model for describing a signal propagation path evolution rule, and performing space-time registration and vector difference calculation with a preset ideal reference model to generate a difference evolution graph; high-dimensional topological feature mapping, density clustering and multi-scale persistence analysis are carried out on the atlas, and a stable abnormal mode caused by defects is identified and confirmed; and backtracking a dynamic evolution path of an abnormal mode, extracting defect core parameters, and completing three-dimensional positioning, type classification and security level evaluation in combination with process information.

The invention relates to the field of power data management, and discloses a power big data adaptive management method and system fused with spatial-temporal feature mapping, and the method comprises the steps: obtaining a real-time operation data flow from a multi-source power terminal, and constructing an original power data set with a time sequence label and a device identifier; the method comprises the following steps: dividing an original power data set into parallel processing units based on a storage-while-computing architecture, performing dynamic index updating by adopting an event-driven index mapping rule, and constructing a multi-dimensional data index cache system with real-time responsiveness; identifying a key abnormal trajectory through a time-varying feature nesting mechanism, and performing hierarchical measurement and entropy disturbance analysis on a data fluctuation degree in the key abnormal trajectory by using a streaming feature aggregation network; identifying potential security risk nodes in combination with the structure matching degree between the historical abnormal event evolution graph and the key abnormal trajectory; and generating a multi-level response instruction chain based on the risk assessment result. The method has the advantage of improving the operation safety of the power grid.

The invention discloses a three-dimensional wind field spatial-temporal feature reconstruction and efficient prediction method, and relates to the field of spatial-temporal feature reconstruction and efficient prediction. Constructing a three-dimensional terrain computational domain based on the digital elevation model of the target mountain region, performing multi-scene numerical simulation by adopting a fluid mechanics method or a mesoscale meteorological model, generating a wind field training data set, and constructing and training a wind field spatial feature mapping model; training a wind speed and wind direction short-time prediction model based on the actually measured data set; inputting the monitoring data obtained in real time into the wind speed and wind direction short-time prediction model to obtain a future wind speed and wind direction prediction value of each monitoring station; and inputting the wind speed and direction predicted values into the wind field spatial feature mapping model to obtain the mountain overall wind field distribution of the target mountain region at the future moment. By constructing an'actual measurement-simulation-modeling-prediction-reconstruction 'integrated technical framework, high-temporal-spatial-resolution short-time prediction from observation of local wind speed and wind direction to the overall three-dimensional wind field of the mountainous region is realized.

The invention belongs to the technical field of remote sensing super-resolution images, and particularly relates to a remote sensing image super-resolution system and method based on an adaptive Mamba-attention network. Comprising a feature extraction module used for carrying out shallow feature extraction on an input low-resolution image to obtain shallow features; the multiple cascaded adaptive state space blocks are used for processing the shallow layer features to obtain reconstruction features; and the reconstruction module maps the reconstruction features to a target resolution space through sub-pixel rearrangement operation to obtain a high-resolution remote sensing image. High-frequency details and a low-frequency structure are cooperatively processed in a feature space by using the remote sensing frequency sensing modulation module, and high-resolution output is generated by combining sub-pixel rearrangement up-sampling, so that high-quality reconstruction of a complex remote sensing scene is realized.

The invention discloses a multi-sensor fusion heat pump full life cycle AI maintenance early warning system, and relates to the technical field of new energy utilization, and the early warning system comprises a data collection module which obtains operation parameters in a heat pump full life cycle based on a sensor array, and constructs a data set after preprocessing the parameters; the operation parameters comprise temperature, pressure, flow and micro vibration; the data fusion module is used for extracting trend correlation characteristics and parameter coupling characteristics from temperature, pressure and flow parameters by adopting a dynamic sliding window adaptive to a working condition, and preserving core nonlinear information through KPCA dimension reduction; the micro-vibration signal extraction comprises frequency domain and time domain features. According to the method, features are extracted through a working condition adaptive dynamic sliding window, then through cross-space mapping and a life cycle-working condition double-attention mechanism, the analysis and early warning module depends on a core feature mapping library and a two-dimensional dynamic baseline, through instantaneous and accumulated deviation double judgment, abnormal accurate recognition and stage division are achieved, and early warning perspectiveness is high.

The invention provides a road marking quality intelligent evaluation system based on image analysis, and relates to the technical field of road facility monitoring, and the system comprises a detection unit, a marking quality evaluation unit, a marking wear prediction unit, a GPS positioning unit, a vehicle driving information unit, a control unit and a remote central control unit. The marking quality evaluation unit is constructed based on a differential geometry theory and comprises a curvature flow edge detection module, a marking geometric characteristic manifold representation module and a multi-scale differential invariant evaluation module, the system regards a marking as a two-dimensional manifold, and the marking quality is evaluated by calculating differential geometric quantities such as a Gaussian curvature, an average curvature and a shape index. And constructing geodesic distance measurement on the feature manifold, and evaluating the completeness, visibility and reflective performance of the marked line. And the marking wear prediction unit predicts the service life of the marking based on the traffic flow information and the environment characteristic mapping relation model, and generates a maintenance suggestion.

The invention belongs to the technical field of network security, and provides an abnormal traffic detection and attack recognition method and system based on deep learning, and the method comprises the steps: data preprocessing and feature extraction, cross-modal semantic alignment and knowledge graph construction, causal enhancement association reasoning, intelligent engine optimization, cloud edge collaborative resource scheduling, and result output. According to the method, statistical features and signature features are mapped to a unified semantic space through a cross-modal semantic alignment and knowledge graph construction module, a semantic barrier between heterogeneous features is broken through, time sequence causal discovery and transfer entropy calculation are introduced, a simple correlation and a reliable causal can be distinguished, and the method has a good application prospect. According to the method, the accuracy and credibility of attack chain reasoning are improved, the false alarm rate is reduced, online self-evolution of a detection model and dynamic optimal allocation of system resources are realized through intelligent engine optimization and cloud edge collaborative resource scheduling modules, and the overall adaptability, robustness and practicability of the system are enhanced.

The invention belongs to the technical field of power system load prediction, and relates to a long-range multivariable load prediction method and system based on time-frequency domain collaboration, and the system carries out the normalization and stabilization of a multivariate load time sequence through a data preprocessing module; the feature embedding module performs linear embedding on the block sequence to construct high-dimensional feature representation; the state space coding module extracts long-range dependency features and generates depth time sequence representation; the decoding prediction module maps the coding features into a preliminary prediction sequence; the time sequence alignment module identifies a leading-lagging relation among multiple variables and aligns a time sequence; the frequency domain optimization module realizes frequency domain component fusion based on adaptive filtering; and the model training optimization module is used for performing training and optimization through a signal attenuation loss function. The method can effectively improve the precision and robustness of long-range multivariable load prediction, and especially has obvious advantages in the aspects of processing complex dependency relationships and dynamic time delay.

The invention discloses an intelligent film and television scene synthesis method based on generative multi-mode script semantic mapping. The method comprises the following steps: firstly, performing semantic unit segmentation and dependency structure analysis on a movie and television play text, and establishing a script semantic mapping relationship; carrying out cross-modal feature mapping training by adopting a generative multi-modal semantic alignment method so as to generate an initial scene layout map; performing space-time consistency optimization on the initial scene layout map, and constraining a synchronization relation between a role action sequence and a camera visual angle according to a time sequence and space depth information; refining and reconstructing the scene layout map sequence to generate a rendering result sequence; and determining a lens switching rhythm and a focal length change track according to a rendering result sequence and a script semantic mapping relationship, and automatically generating a playable movie and television scene video with natural plot logic and uniform visual style. According to the method, the generative semantic self-adaptive synthesis from the script text to the video content is realized, and the automation and intelligence level of film and television production is improved.

The embodiment of the invention provides an elderly emotion recognition method and device based on multi-modal perception, and the method and device achieve the optimization and enhancement of the signal quality through innovatively constructing a multi-modal data preprocessing mechanism and integrating the facial expression, voice and posture features. And designing a personalized feature mapping model based on historical emotion expression data, and establishing an adaptive feature fusion strategy for intelligent matching in combination with a cross-modal attention network. A hierarchical time sequence classification mechanism is introduced, dynamic modeling of the emotional development trend is realized through a long-short term memory network, and accurate prediction of the emotional state is supported. According to the method, the defects of the traditional technology in the aspects of multi-modal processing, personalized modeling, time sequence analysis and the like are effectively overcome, and the accuracy and reliability of sentiment recognition of the old people are remarkably improved.

The invention provides a robot milling surface roughness prediction method and system based on parallel convolution and coordinate attention mechanism feature fusion. According to the method, firstly, multi-source heterogeneous high-frequency signals in the milling process of a robot are synchronously collected, empty slice segments are eliminated and environmental noise is filtered out by utilizing a self-adaptive threshold value and a robust statistical method, and a standardized time sequence sample set is constructed; secondly, establishing a multi-source heterogeneous feature fusion roughness prediction model, and independently extracting deep features of cutting force and vibration signals by adopting a double-branch parallel convolution architecture; a coordinate attention mechanism is introduced, heterogeneous modal features are mapped into a virtual two-dimensional topological structure, the dynamic relative contribution degree of cutting force and vibration signals changing along with the machining state is captured through pooling aggregation along the modal dimension, and an attention weight map is generated to conduct dynamic weighting on the features. And finally, outputting a surface roughness predicted value through a regression network. According to the method, the problem that an existing fusion method neglects the dynamic dependency relationship between physical quantities is effectively solved, the anti-interference capability of the robot in the weak rigidity machining process is enhanced, and the precision and robustness of surface roughness prediction are remarkably improved. The method can be widely applied to robot complex component machining quality analysis and intelligent optimization, and has high efficiency and reliability.

The invention relates to the technical field of machine vision, in particular to a precision part size automatic detection method and system based on machine vision, precision part images are collected through a high-precision industrial camera, part positioning is carried out, sub-pixel-level topological feature mapping is carried out on interested area images, and precision part size automatic detection is carried out. Comprising the steps of gray histogram equalization, gray morphological processing, edge detection and edge chain code tracking, construction of an edge point topological feature space, execution of sub-pixel subdivision, obtaining of an edge line through contour analysis of a feature distance and a feature angle, and double-constraint geometric reconstruction based on the edge line. The characteristic distance and the characteristic angle are used for rotation matrix conversion and geometric dimension calculation, a relation model of the geometric dimension and the actual dimension of the part is established, precise part dimension measurement is achieved through dynamic error analysis and compensation, the measurement precision is remarkably improved, the risk caused by unreliability of a single characteristic is effectively reduced, and the measurement accuracy is improved. And the measurement stability is improved.

The invention relates to the technical field of artificial intelligence, and discloses an artificial intelligence English general recognition large model training method and system, and the method comprises the steps: segmenting text data in a standardized multi-modal training set into semantic blocks, and constructing a semantic primitive-feature mapping table; associating the multi-modal English general recognition training corpus with the text semantic primitives in the semantic primitive-feature mapping table to obtain cross-modal alignment features; deep fusion features of the cross-modal alignment features after deep fusion are extracted; adjusting difficulty distribution of training samples in the training stage to obtain preliminary parameters; performing multiple rounds of knowledge distillation on the preliminary parameters based on entity relationship data in an English culture background knowledge graph to obtain culture enhancement model parameters; performing multi-dimensional evaluation on culture enhancement model parameters, and performing iterative optimization on the model according to an evaluation result to obtain a target artificial intelligence English general recognition large model; the training effect of the artificial intelligence English general recognition large model can be improved.

The invention provides an intelligent power grid energy optimization management method and system, and relates to the technical field of data processing, and the method comprises the steps: 1, synchronously collecting photovoltaic output fluctuation data, user load demand data and power grid frequency state data, so as to obtain an original monitoring data set; step 2, performing spatial-temporal feature mapping on the original monitoring data set, extracting a multi-dimensional feature vector, determining a dominant feature shaft system based on covariance analysis, and calculating a dynamic coupling degree between the feature shaft systems to divide a regulation and control domain; time sequence key characteristic quantities are selected in a regulation and control domain, a load response state transition path is constructed, a user behavior correction coefficient is deduced according to the load response state transition path, dynamic load balance is achieved through a multi-period rolling strategy, and a load distribution scheme is generated. According to the invention, through real-time dynamic regulation and control of power supply output, an excitation electricity price signal is generated and user demand response is linked, so that intelligent power grid energy supply and demand balance optimization is realized.

The invention discloses a feature fusion processing method for anesthesia depth multi-modal data, and belongs to the technical field of graphic data processing and pattern recognition, and the method comprises the steps: obtaining a multi-modal physiological signal and an electromyographic signal of a patient; performing time axis calibration on the physiological signal to generate an alignment signal; extracting a multi-modal feature vector and calculating an anesthesia depth index; performing deviation analysis on the basis of the electromyographic signal and the index to obtain an electromyographic response deviation index; performing graphical feature mapping on the real-time electroencephalogram signal to generate a real-time time-frequency map; when the deviation index exceeds a safety threshold value, performing graph pattern matching with a pattern template library to calculate a similarity score; and outputting the current anesthesia state mode in a classified manner. According to the method, a multi-modal signal graphical feature fusion technology is adopted, and a dynamic map generation and pattern matching mechanism is combined, so that the problem of complex pattern recognition of physiological signal graphic data can be solved, and the accuracy and timeliness of anesthesia state classification are improved.

The invention discloses a power distribution equipment health assessment method and system based on multi-source data, and the method comprises the steps: mapping each modal feature into a comparable measure, constructing a learnable cost containing power flow and heat consistency, outputting a modal weight through scene gating, and forming a fusion representation of physical consistency; driving a neural differential equation by fusing the stress force obtained through representation decoding, adopting monotone weight parameterization and introducing equipment-level damage budget, and obtaining damage and health indexes which are irreversible along with time; under the constraint of physical baseline life, combining a working condition input time-scene gating danger rate model, performing causal consistency correction through virtual intervention, and outputting an interval failure probability and residual life; and calibrating a dynamic threshold value in the working condition cluster, and triggering routing inspection, sampling and load shedding based on the risk sensitivity and the topological linkage risk priority. According to the method, multi-source physical consistent fusion, individualized health modeling and dynamic closed-loop optimization can be realized, and the accuracy and interpretability of health assessment of the power distribution equipment are improved.

A method of emotion recognition in cross-subject EEG signals, belonging to technical field of deep learning, includes the following steps: S1, constructing the extracted DE features into positive and negative samples by using a positive and negative sample generator; S2, sending the DE features of an anchor and the positive and negative samples into the encoder for coding, mapping the DE features to a latent space, performing regression prediction on the encoded anchor samples in the latent space by using an autoregressive model, training the encoder by using a probability supervision contrastive loss function; and S3, connecting the trained encoder to the classifier for fine tuning, and training the classifier through the cross entropy loss function; in this process, the encoder does not perform gradient propagation to complete cross-subject emotion recognition.

The invention discloses a pump set vibration monitoring method and system based on distributed sensing. The method comprises the steps that vibration collection signals and real-time working condition parameters of distributed sensing nodes are obtained, multi-node clock synchronization correction is executed, and synchronous vibration data frames are formed; time-frequency feature extraction is carried out on the multi-channel vibration data set, abnormal channels are identified through inter-channel consistency analysis, a local fault correlation frequency band is extracted, and a vibration feature vector is constructed; associating the vibration feature vector with a real-time working condition parameter to form a working condition-feature mapping table, identifying a feature mutation rotating speed point for the feature baseline, adaptively dividing a rotating speed interval to generate an adaptive working condition threshold value, and implementing out-of-limit detection to generate an abnormal trigger identifier; determining a traceability analysis window based on the abnormal trigger identifier, and executing multi-measuring-point coherence calculation and energy attenuation gradient analysis to construct an energy transfer link diagram; and finally, vibration source position positioning is performed to form fault source probability distribution, a fault mode label is matched, pump set vibration monitoring is completed, and quantitative positioning of a fault source is realized.

The invention provides a medical image segmentation method and system based on spatial detail enhanced vision, and relates to the technical field of image processing. The method comprises the following steps: performing initial feature mapping on an input and preprocessed medical image to obtain an embedded feature map; the embedded feature map is input into an encoder for feature extraction, and multi-scale features are obtained; the encoder comprises a plurality of encoding stages, and the number of channels is doubled and the spatial resolution is halved through down-sampling operation between the encoding stages; the multi-scale features are subjected to up-sampling and spatial enhancement reconstruction step by step through a decoder, and a high-precision segmentation result is generated; wherein the decoder comprises a plurality of decoding stages, the decoding stages correspond to the encoding stages, and feature fusion is carried out between the corresponding stages of the encoder and the decoder through jump connection. According to the method, the boundary description precision and the segmentation robustness of the low-contrast image can be improved without increasing the linear complexity, and the method is suitable for medical image segmentation scenes of skin lesions, gastrointestinal polyps and the like.

The invention provides a data security risk early warning method and system based on big data analysis, and the method comprises the steps: firstly constructing a data security risk feature map, collecting a heterogeneous data set in a mobile communication network in real time through a multi-source data access interface, generating a multi-source heterogeneous data fusion stream through distributed cleaning and standardization processing, and carrying out the data security risk early warning. Then inputting the data security risk feature into a preset distributed risk feature learning network, performing feature mapping and association enhancement processing based on a data security risk feature map to obtain a real-time risk feature vector, performing big data association analysis on the real-time risk feature vector, mining a risk feature conduction dependency relationship, generating a risk propagation path weight set, and performing big data association analysis on the real-time risk feature vector; and finally, determining a risk diffusion level and a key influence node, generating a security risk early warning instruction containing a risk diffusion path identifier, and pushing the security risk early warning instruction to a mobile communication security management platform, thereby realizing accurate early warning and quick response of the data security risk, and ensuring safe and stable operation of a mobile communication network.

The invention relates to the technical field of image pre-training, in particular to a training method and system of an image pre-training model and a storage medium. The method comprises the following steps: collecting an original training image, carrying out high-dimensional feature mapping and pyramid transformation to generate a multi-resolution feature body, carrying out texture definition judgment on the image based on the feature body, screening out a defective training image and a qualified training image, and for the defective training image, carrying out texture definition judgment on the qualified training image. A traceability correction technology is adopted to correct a defect correction training image, positive and negative contrast images are constructed to train a defect image pre-training model, meanwhile, weak disturbance transformation is performed on a qualified training image, a positive example pair before and after disturbance is generated to train a qualified image pre-training model, the two pre-training models are integrated into a collaborative sample stream, and the collaborative sample stream is extracted. And a high-performance integrated image pre-training model is obtained through integrated pre-training. According to the method, image quality grading and differential modeling are realized, the defect identification precision is improved, and the model generalization ability is enhanced.

The invention discloses a data and physical dual-drive crack propagation prediction method, which belongs to the field of petroleum engineering and comprises the following steps: step 1, constructing a real observation data set and a sampling data set; step 2, constructing a hybrid architecture fusing a Transform encoder and a graph attention network; 3, three independent and parallel decoders are constructed to map the shared features into the geometric dimensions and mechanical parameters of the cracks; 4, establishing a physical loss function based on linear elastic fracture mechanics and a material balance principle, and combining the physical loss function with a data loss function to construct a mixed loss function for model training; and 5, predicting the geometric dimension and mechanical parameters of the fracturing crack by using the trained model. According to the method, physical priori knowledge is embedded into a multi-task deep learning framework, and physical loss is embedded into a loss function, so that the precision and interpretability of model prediction are remarkably improved.

An industrial big data-driven vertical federated transfer-based anomaly detection method and system. The method comprises: acquiring a source domain data set and a target domain data set from an industrial scenario, the source domain data set being constructed on the basis of industrial data having known anomaly labels, and the target domain data set being constructed on the basis of industrial data without anomaly labels; and on the basis of a preset vertical federated transfer model: performing vertical federated feature extraction: mapping the source domain data set and the target domain data set into a common feature space to obtain potential features; performing domain adaptation: extracting features having domain invariance and discriminability from among the potential features; and performing joint domain alignment: aligning the distance between domains, and mapping the features having domain invariance and discriminability to obtain the anomaly labels.

The invention relates to the field of artificial intelligence cross-modal retrieval, and provides a video text cross-modal retrieval method and system based on spatio-temporal feature fusion. The method comprises the following steps: carrying out key frame sampling and time sequence partitioning on an input video, extracting static visual features through a spatial feature network, and extracting motion features through a time dynamic network; a self-adaptive gating fusion module is adopted to dynamically calculate spatial-temporal feature weights and perform weighted fusion; extracting text semantic features by using a pre-training language model; constructing a double-flow projection network to map video fusion features and text features to a unified measurement space, and optimizing a feature distance by adopting a contrast loss function containing difficult negative sample mining and intra-modal constraint; and outputting a retrieval result according to the cosine similarity sequence. The system comprises four units, wherein the gating fusion module is integrated with an FPGA acceleration circuit. According to the method, mAP (at) 10 is equal to 0.78 in a UCF-101 data set, the time sequence action retrieval accuracy rate is 92.8%, and the single video retrieval delay is 23 milliseconds.

The invention provides a prefabricated dish semi-finished product defect identification method and system with an AI algorithm, and the method comprises the steps: extracting semantic features from a refined defect candidate region set, carrying out the feature mapping of each region through a deep convolutional network according to the demands of atypical defect identification, and obtaining the defect description represented by a high-dimensional feature vector; marking original image data through a final defect identification result, and for inhibition of complex background interference, performing outward expansion from defect edge features by adopting a region growing algorithm to obtain complete defect region boundary information; after complete defect area boundary information is obtained, defect distribution changes of continuous batches of images are compared through a time sequence analysis method according to the monitoring requirement of production process fluctuation, and the quantitative basis of process adjustment is determined.

The invention relates to the technical field of electrical digital data processing, and discloses a train ice melting simulation optimization method of an electromagnetic thermal coupling model fused with a deep learning method, which comprises the following steps of: inputting a geometric representation tensor and a physical working condition parameter vector containing an electromagnetic excitation frequency and a reference environment temperature into a feature mapping neural network; outputting a dual-channel space source item tensor containing basic heat source power density and a heat source to temperature change sensitivity distribution matrix through nonlinear convolution operation; constructing a heat conduction discrete numerical value evolution operator configured with an active item linear correction interface; time stepping operation is executed according to the heat conduction time scale, a basic heat source is corrected in real time through the Hadamard product of a sensitivity distribution matrix and temperature deviation, and an operator is substituted for solution. On the premise that electromagnetic-thermal nonlinear coupling characteristics are reserved, decoupling of the time scale is achieved, and the calculation efficiency in high-frequency physical field simulation is effectively improved.