38 results about "Class separability" patented technology

The invention discloses an enterprise big data mining method and system based on artificial intelligence, and the method comprises the steps: carrying out the dynamic mode alignment through employing a multi-mode hypergraph neural network according to enterprise multi-source heterogeneous data, and generating a time-space consistent multi-mode joint embedded tensor; inputting the multi-modal joint embedding tensor into an orthogonal adversarial manifold learning module, and generating a low-dimensional compact semantic embedding vector with enhanced category separability; performing space-time causal association mining on the semantic embedding vector, and outputting a space-time causal meta-path map containing the recessive commercial logic; and inputting the space-time causal element path map into a dynamic game adversarial interpretation framework, and finally outputting an enterprise-level intelligent decision map with anti-factual robustness. By utilizing the embodiment of the invention, the multi-modal data can be efficiently integrated and intelligently analyzed, and the accuracy and effectiveness of the mining result are improved.

The invention relates to the technical field of information retrieval, in particular to a cross-modal hash retrieval model training method and device and a retrieval method and device. In the method, adaptive gradient triple loss can distribute adaptive gradients for triples with different difficulties by introducing a constraint term based on an included angle between an anchor point and a negative sample, so as to maintain the consistency of a neighborhood relationship in an original space, promote intra-class compactness and inter-class separability of a heterogeneous mode, and improve the robustness of the heterogeneous mode. Samples which are semantically similar to query samples can be retrieved; the step-by-step quantization loss decouples representation learning and binary code generation, the two-stage design retains a semantic structure in an embedding stage, and the subsequent quantization loss is minimized; therefore, the cross-modal hash retrieval model has good retrieval performance.

The invention discloses a rolling bearing fault classification method fusing adaptive distribution perception discrimination loss (ADADL), and belongs to the technical field of rolling bearing fault diagnosis. The rolling bearing fault classification method comprises the following steps of: obtaining a rolling bearing fault, and carrying out classification on the rolling bearing fault by using the ADADL as a fusion model, and carrying out classification on the rolling bearing fault by using the ADADL as a fusion model, and carrying out classification on the rolling bearing fault by using the ADADL as a fusion model. In a complex industrial environment, classification boundary fuzziness is often caused by noise interference and feature overlapping, and the accuracy of rolling bearing fault diagnosis is reduced. According to the method, an adaptive distribution perception discrimination loss function (ADADL) is provided, and intra-class compactness and inter-class separability are improved by adjusting intra-class distance through a dynamic threshold value and optimizing inter-class distribution through an adaptive boundary. And the cross entropy loss is combined with ADADL, so that the classification precision is further optimized, the model is helped to better process samples difficult to classify, and the robustness and the adaptive ability of the model are improved. The classification performance is remarkably improved on the CWRU data set, and particularly, excellent robustness and generalization ability are shown under the conditions of class imbalance and strong noise. Feature visualization results show that ADADL can optimize clustering boundaries of different fault categories, minimize overlapping regions, and relieve the problem of fuzzy classification boundaries.

The invention provides a deep compact feature representation network model based on a rank correlation induced dynamic mask mechanism, which is used for relieving confirmation deviation generated by distillation training due to excessive unlabeled data in semi-supervised learning and effectively compressing distribution of features in a potential space. According to the method, global and local connection between pixels is broken through a feature diffusion enhancement module, a cluster sensing neighborhood consistency module is introduced, and intra-class compactness and inter-class separability are enhanced by using adaptive metric learning to realize pixel-by-pixel feature representation; a rank correlation induction dynamic mask module is adopted, a category correlation mode is reserved in a rank matching mode, and high-probability errors are filtered out, so that the problem of rank relation loss caused by semantic consistency weakening probability alignment randomness is ensured. Evaluation on a plurality of data sets shows that knowledge is distilled to the student network through the teacher network, meanwhile, the correlation between the feature distribution geometric structure and the prediction space is optimized, and the method is obviously superior to an existing method, so that the generalization performance of the model is improved.

The invention discloses an unsupervised text-to-image pedestrian re-identification method and system. The method comprises the following steps: constructing an image-text matching relationship; clustering operation is carried out on the image features, and false identity tags are distributed to the pedestrian images and the corresponding text descriptions; constructing an image memory dictionary; obtaining a similarity set according to the image memory dictionary; removing pedestrian images with abnormal similarity and corresponding text descriptions according to the similarity set; respectively acquiring image features and text features of pedestrian images and text descriptions, and constructing a multi-stage ternary joint learning process; and obtaining text features to be re-recognized, and performing pedestrian image matching to obtain a pedestrian recognition result. According to the method, the intra-class compactness and inter-class separability of the samples are enhanced, and the retrieval precision and efficiency of the unsupervised text-to-image pedestrian re-recognition method are greatly improved.

The invention relates to the technical field of agricultural remote sensing and cultivated land monitoring, in particular to an accurate cultivated land identification method integrating multi-source remote sensing and phenological response, which comprises the following steps: S100, data acquisition: acquiring time sequence data of multi-source remote sensing in a research area; s200, the phenological period of the crops is extracted; s300, generating a training sample: based on phenological information, obtaining a positive sample by taking an adaptive flowering index as a threshold value; generating a negative sample by adopting a reverse rule of the positive sample rule, and purifying to generate a sample group; s400, feature fusion and optimization: evaluating the category separability of different time phase feature sets by adopting a J-M distance method; s500, classifier evidence fusion: using a D-S synthesis rule to realize classification result fusion; and S600, applications are migrated annually. According to the method, the multi-source remote sensing data are fused, the sample automatic generation strategy is constructed in combination with the phenological period characteristics of the crops, the evidence theory is introduced to fuse the results of the multiple classifiers, and high-resolution and high-stability accurate recognition of the cultivated land in the complex terrain and cloudy and rainy environment is achieved.

The invention discloses a prompt-driven two-stage multi-modal emotion representation learning method. The method comprises the following steps: respectively collecting original video data from a plurality of public multi-modal emotion analysis data sets; preprocessing and feature extraction are carried out to obtain vectorized multi-modal feature representations of vision, audio and texts, and multi-source emotion clues are obtained; in a training stage, a prompt-driven two-stage multi-modal emotion representation learning model is constructed, inter-class separability is enhanced and intra-class intensity features are reserved through an emotion anchor point comparison and alignment stage, including prompt-based emotion anchor point learning and comparison and alignment between joint representation and emotion anchor points; capturing the dynamic change of emotion expression through an emotion intensity offset estimation stage; in the reasoning stage, emotion category prediction and final emotion state prediction are carried out. According to the method, shared emotion features among multiple modes can be stably captured, interference caused by individual differences is effectively suppressed, and the accuracy and robustness of emotion prediction are remarkably improved.

The invention discloses a cross-domain underwater image recognition method, device and system, and a storage medium. The method comprises the steps of obtaining a to-be-recognized target domain image; inputting the target domain image to be recognized into an image recognition model for processing, and outputting a recognition result; wherein the image recognition model is trained through a training set, the intra-class aggregation and inter-class separability of a feature space are optimized by calculating comparison loss and dynamically distributing sample weights, a target domain model is optimized until convergence, and the training set is a source domain image data set with labels. According to the method, the recognition task of underwater image cross-domain migration is effectively solved, and a new thought is provided for the problems of limited data and cross-domain distribution difference in the target field.

The invention discloses a heterogeneous federated learning method based on attention guidance aggregation and prototype enhancement, and relates to a heterogeneous federated learning method. The invention aims to solve the problem that the existing heterogeneous federal learning method cannot improve the granularity of client feature representation learning and cannot solve class imbalance in global head training. The method comprises the following steps: step 1, constructing a feature aggregation mechanism based on attention guidance; 2, designing a contrast learning objective function, and explicitly enhancing inter-class separability and intra-class compactness of feature representation in a local training process; 3, designing a self-adaptive prototype enhancement strategy, and relieving a class imbalance problem; and 4, constructing a collaborative optimization framework of the global model and the local model. The invention belongs to the technical field of distributed collaborative learning.

This invention discloses a multimodal hash retrieval method, system, device, and medium based on a multi-viewpoint central structure, belonging to the fields of artificial intelligence and multimodal hash retrieval technology. The technical problem this invention aims to solve is achieving a balance between intra-class compactness and inter-class separability during multimodal hash retrieval. The technical solution adopted is as follows: constructing a multimodal dataset; constructing a multimodal hash retrieval model based on a multi-viewpoint central structure; and training the model. Specifically, constructing the multimodal hash retrieval model based on a multi-viewpoint central structure involves: modality-specific prototype learning: using an image modality deep multilayer perceptron and a text modality deep multilayer perceptron to extract refined features from the corresponding modalities, and calculating the average value of the refined features of the image modality and the text modality to obtain modality-specific prototypes, thereby obtaining unique features of the image modality and unique features of the text modality; multimodal ensemble class prototype learning; and multi-view semantic enhancement hash learning.

The application discloses an iris recognition method and device, system and storage medium, and performs preprocessing such as polar coordinate unfolding on an input original iris image in sequence; multi-level image features are extracted through an inverse residual module; subsequently, a global depth convolution GDConv operator is introduced to replace a global average pooling layer, so that spatial structure distribution features of iris textures are retained under the premise of extremely low calculation complexity; finally, an ArcFace loss function is used to optimize the distribution of features in a hyperspherical space, and the class separability of the features is enhanced by introducing an additive angle interval. By using the technical scheme of the application, the technical problems that spatial topological information is lost due to a global average pooling GAP operation in iris feature extraction by using an existing lightweight convolutional neural network, and the feature discrimination is reduced due to network lightweight are solved.

The application discloses a kind of key equipment intelligent diagnosis method and system for rotating machinery, it is related to the technical field of fault diagnosis and prediction, including, with physical inspiration as guidance, in feature learning stage, introduce bounded metric encapsulation mechanism, traditional unbounded feature space is mapped into bounded similarity manifold, to suppress high-dimensional space distance concentration effect and improve inter-class separability and intra-class compactness.In bounded metric space, realize multi-fault identification by combining near neighbor or support vector classifier.The application has significant improvement in weak fault (germination period) detection, composite / multiclass fault distinction and robustness under low signal-to-noise ratio condition, has good explainability and engineering deployment value, is suitable for the predictive maintenance and online health monitoring of wind power, rail transit, aerospace and process manufacturing and other scenes.

The present application relates to remote sensing image processing technical field, specifically a kind of tensor singular spectrum analysis method for hyperspectral image three-dimensional feature extraction, including based on spatial self-similarity adaptive embedding, decomposition and low rank representation based on t-SVD, feature image and classification, compared with prior art, through adaptive embedding and t-SVD process, a new adaptive embedding operation, utilize the spatial similarity feature of HSI, jointly utilize target pixel and non-local similar pixel, combine with re-projection operation, keep target inter-class difference while enhanced intra-class similarity, by designing a trajectory tensor, combined with t-SVD, jointly represent the global low rank feature of HSI, the arrangement of similar pixel in trajectory tensor makes it have low rank feature, and further extract low rank feature by truncated t-SVD, realize the feature extraction of three-dimensional hyperspectral image, and then improve the class separability in hyperspectral image classification.

This invention belongs to the interdisciplinary field of medical image analysis and deep learning, and discloses a method and apparatus for pathological slide image segmentation using multimodal supervised contrastive learning. The method generates biological prior feature maps through decoupling, extracts visual and semantic features by combining the master encoder and prior encoder, and achieves feature alignment using a cross-modal attention mechanism. Supervised contrastive learning is introduced to enhance intra-class compactness and inter-class separability. A multi-scale decoupled decoder is designed, with structural similarity loss and boundary-aware loss respectively supervising structural and detail branches. Finally, dynamic weighted fusion and pseudo-label calibration strategies are used to optimize segmentation performance, significantly improving the accuracy and robustness of pathological image segmentation.

The present invention discloses an intelligent target detection method based on variable-dimensional feature screening. Since the existence of feature redundancy in the process of constructing a multi-dimensional feature space will lead to an increase in unnecessary computing costs. In view of the problem that it is impossible to adaptively select feature combinations according to actual scenarios under the condition that the feature dimensions are not restricted during the feature screening process, a variable-dimensional feature screening method based on geometric metrics and variance inflation factors is proposed. While constraining the correlation between features, this method searches for feature combinations that maximize the degree of inter-class separability, removes redundant features while retaining sufficient useful information, and establishes a comprehensive and effective feature vector. Then, based on the measured data, an intelligent detector with a controllable false alarm rate is used to complete target detection.

This invention discloses an open-set specific radiation source identification method and system based on multi-granularity embedding guidance. First, selective attention complex convolution branches and Transformer branches are used to extract local fine-grained features and global temporal dependency features of radio frequency signals, respectively. Then, an attention-guided multi-scale feature fusion module performs cross-granularity alignment and semantic fusion on the dual-path features. A label-aware discriminative prototype embedding space is constructed, and joint loss is used to optimize the embedding space to enhance intra-class aggregation and inter-class separability. Unidentified unknown class samples are projected into the discriminative embedding space, and an angle-based prototype incremental update mechanism is used to further identify and subdivide the unknown categories. This invention can simultaneously achieve high-precision identification of known radiation sources and detection and subdivision of unknown radiation sources in open-set scenarios, significantly improving generalization ability and identification stability in complex real-world environments.

The invention provides an open set hyperspectral image accurate classification method based on difficult sample learning and dense spectral spatial feature enhancement, and belongs to the field of hyperspectral image processing. Firstly, a training sample is mapped into a dense spectral feature model, multi-layer information accumulation and fusion are achieved, spectral local changes are modeled finely, and the discrimination between adjacent categories is improved. Secondly, a spectral feature compression model is constructed, known class feature aggregation is promoted, meanwhile, a spatial modeling branch is introduced, and the sensing capacity for complex structures such as edges and textures is enhanced; and a class anchor point strategy is further adopted, so that intra-class polymerization and inter-class separability are improved, and a class center structure is stabilized. For negative samples which are difficult to distinguish, a difficult sample learning strategy is introduced, boundary learning is enhanced, and the unknown class recognition capability is improved. And finally, through the distance between the sample and the anchor point, the ground feature category is obtained, and an effective new thought is provided for open set classification of the hyperspectral image.

A semi-supervised medical image segmentation method based on diffusion-driven hard-soft prototype comparative learning belongs to the field of medical image processing and comprises the following steps: acquiring a medical image and dividing the medical image into a labeled data set and a non-labeled data set; generating a pixel-level confidence map by using a discriminator; using a confidence map to guide a diffusion model to carry out adaptive denoising and correction on the noise pseudo label, and introducing a local detail enhancement mechanism to generate a corrected pseudo label; a hard-soft collaborative prototype comparison module is constructed, a category prototype is constructed based on the corrected pseudo tag, a hard matching indication matrix is generated for high-confidence pixels, soft guidance constraint is applied to low-confidence pixels, and intra-class compactness and inter-class separability of a feature space are enhanced; performing iterative optimization on the model through a joint optimization objective function; inputting the medical image to be segmented into the trained segmentation network to obtain a segmentation result; according to the method, the problems of pseudo label noise accumulation and feature discrimination degradation in semi-supervised learning are solved, and the dependence on large-scale labeled data is reduced.

The invention belongs to the technical field of image recognition, and particularly relates to a CT image motion artifact classification model construction method and system based on feature prototype contrast learning. According to the classification method based on the prototype, the class center is used as feature storage, the inter-class separability and the intra-class consistency in the artifact classification task are enhanced, and the robustness and the classification precision of the model are remarkably improved. The method specifically adopts a Vision Transform (ViT) as a basic model for feature extraction, combines a strong global modeling capability, effectively captures long-range dependence and fine-grained features in artifact detection, and improves the capability of identifying the complexity of artifact types. And by introducing a prototype contrast learning strategy, the feature representation of the artifact image is optimized, and the overfitting problem of the model when training data is insufficient is relieved, so that the generalization ability and robustness of the model in practical application are improved, and an excellent classification result is obtained on a clinical data set.

This invention relates to a method for identifying individual communication radiation sources based on a multi-scale residual prototype learning network. The method comprises a signal preprocessing module, a network training and feature extraction module, a storage module, and an identification module. The signal preprocessing module is connected to the network training and feature extraction module, which is connected to the storage module, and the storage module is connected to the identification module. This invention extracts features from individual communication radiation source signals using a multi-scale residual prototype learning network and then uses a joint decision-making method to identify the attributes of the individual radiation source signals. In this method, MSRPLNet lays the foundation for identifying the features of individual communication radiation source signals by enhancing the inter-class separability and intra-class compactness of the learned features.

The present application belongs to the technical field of computer, and particularly to a face forgery detection system based on uncertain modeling.The system comprises a probabilistic Transformer module, an image block screening module and an uncertainty-aware single classification loss function module.The present application firstly models the dependency relationship between image blocks as Gaussian random variables, extends the Transformer model in a probabilistic manner, then introduces an image block selection module to identify areas with high uncertainty information for final classification, and finally quantifies the uncertainty of the entire image, uses the designed uncertainty-aware single classification loss function to make the model focus more on samples with high uncertainty and difficult to determine, and through only enhancing the internal compactness of real faces, improves the inter-class separability of real and false classes in the embedding space.

The invention discloses an open set specific radiation source identification method and system based on multi-granularity embedding guidance. The method comprises the following steps: firstly, respectively extracting local fine granularity features and global time sequence dependence features of radio frequency signals by using a selective attention complex convolution branch and a Transform branch; carrying out cross-granularity alignment and semantic fusion on the dual-path features through an attention-guided multi-scale feature fusion module; constructing a label perception discrimination prototype embedding space, and optimizing the embedding space by using joint loss to enhance intra-class aggregation and inter-class separability; and projecting the rejected unknown class sample to a discriminative embedding space, and realizing further identification and subdivision of the unknown class by using an angle-based prototype increment updating mechanism. According to the invention, in an open set scene, high-precision identification of a known radiation source and detection and subdivision of an unknown radiation source can be realized at the same time, and generalization ability and identification stability in a complex real environment are significantly improved.

The invention discloses a key equipment intelligent diagnosis method and system oriented to rotating machinery, and relates to the technical field of fault diagnosis and predication, and the method comprises the steps: taking physical inspiration as guidance, introducing a bounded measurement packaging mechanism in a feature learning stage, mapping a conventional unbounded feature space into a bounded similarity manifold, and carrying out the analysis of the bounded similarity manifold; therefore, the distance concentration effect of a high-dimensional space is inhibited, and the inter-class separability and the intra-class compactness are improved. And realizing multi-fault identification in combination with classifiers such as neighbors or support vectors and the like in a bounded metric space. According to the method, the robustness in weak fault (germination stage) detection, composite / multi-class fault distinguishing and low signal-to-noise ratio working conditions is remarkably improved, good interpretability and engineering deployment value are achieved, and the method is suitable for predictive maintenance and online health monitoring of scenes such as wind power, rail transit, aerospace and process manufacturing.

The present application relates to the technical field of remote sensing change detection, and discloses a remote sensing change detection method and system based on change decoupling enhancement, comprising: acquiring double-time-phase remote sensing images, constructing a change decoupling enhancement network model, the model comprising a double-branch encoder, a change representation decoupling module, a semantic consistency enhancement module and a segmentation head, the double-branch encoder extracts double-time-phase features of a double-time-phase image pair at different scales, the change representation decoupling module explicitly identifies and separates unchanged components by calculating the similarity between the double-time-phase features, the semantic consistency enhancement module enhances intra-class consistency and inter-class separability by dividing the change representation into multiple semantic groups and independently refining them, and the segmentation head performs remote sensing change detection according to semantic features. The present application can improve the semantic discrimination ability of change detection, thereby improving the accuracy of remote sensing change detection.

The present invention discloses an artificial intelligence-based enterprise big data mining method and system. The method includes: using a multimodal hypergraph neural network to perform dynamic modal alignment based on the enterprise's multi-source heterogeneous data to generate a spatiotemporally consistent multimodal joint embedding tensor; inputting the multimodal joint embedding tensor into an orthogonal adversarial manifold learning module to generate a low-dimensional, compact semantic embedding vector with enhanced class separability; performing spatiotemporal causal association mining on the semantic embedding vector to output a spatiotemporal causal metapath graph containing implicit business logic; inputting the spatiotemporal causal metapath graph into a dynamic game adversarial interpretation framework to ultimately output an enterprise-level intelligent decision-making graph with counterfactual robustness. Utilizing embodiments of the present invention, multimodal data can be efficiently integrated and intelligently analyzed, improving the accuracy and effectiveness of mining results.

The invention provides a retinopathy detection method fusing multi-focus segmentation and comparative learning, and designs a two-stage network by closely fusing focus segmentation and grading tasks. In the lesion segmentation task of the first stage, an improved MCA-UNet network is provided, the feature extraction capability is optimized by introducing a semi-wavelet collaborative attention module, and high-precision lesion spatial feature support is provided for the subsequent grading task. In the DR grading task of the second stage, a double-branch prototype comparison learning model is constructed, and a feature learning branch strengthens intra-class compactness and inter-class separability by constructing a class prototype, so that the problem of class imbalance ubiquitous in a DR data set is effectively relieved.

The present invention discloses a small sample target recognition method based on virtual feature enhancement and feature refinement. First, a feature extractor is used to extract a low-dimensional feature map from a limited number of SAR training samples. Then, an embedded feature enhancer is used to calculate the similarity and search for features far from the class center, obtaining intra-class dissimilar feature pairs and inter-class similar feature pairs. The feature pairs are used to construct and enhance the corresponding virtual features. Then, a dynamic hierarchical feature refiner is used to dynamically generate convolution kernels to integrate local features with different dimensions into global features, thereby improving the intra-class compactness and inter-class separability of the features. Finally, a loss function combining adaptive loss and recognition loss is used to optimize the network, improve recognition accuracy, and complete target recognition. The method of the present invention solves the problems of difficult data acquisition and high annotation cost, improves the deep network learning effect and overall recognition accuracy, and has higher recognition accuracy and robustness than existing small sample target recognition methods.

The application discloses a deep face recognition method based on a minimum margin loss function, and comprises the following steps: S1, preparing a data set, wherein no face image that may overlap with a reference data set is contained in the data set; S2, using a Tensorflow framework to build a network model based on an Inception-ResNet-v1 model; S3, inputting pictures of a training set in the data set into the network model to perform training; and S4, inputting pictures of a test set in the data set into the trained network model to recognize face images. The application combines the advantages of a Softmax loss function, a center (Center Loss) loss function and a minimum margin (Minimum Margin Loss) loss function, wherein the center loss is used for enhancing in-class compactness, and the Softmax loss and the minimum margin loss are used for improving inter-class separability. Experimental results show that the proposed minimum margin loss makes the technical performance of face recognition reach a new height and reduces the negative influence caused by margin preference.

The present application belongs to the field of fault diagnosis technology of high-speed train transmission system, and discloses a train bogie few sample cross-domain fault diagnosis method based on double path feature reconstruction. In view of the problems of sample scarcity, domain offset under cross-condition, cross-fault type and cross-equipment scene, firstly, one-dimensional vibration signal is converted into two-dimensional time-frequency graph through continuous wavelet transform, and deep features are extracted through convolution feature embedding network; then, a self-attention feature enhancement module is constructed, and fine-grained fault mode is extracted through position coding, self-attention weighting and nonlinear mapping; further, a bidirectional variation correction module is designed, the support set is used to reconstruct the query set to enhance the class separability, and the query set is used to reconstruct the support set to constrain the intra-class compactness; finally, the forward and backward reconstruction distances are fused through a double metric diagnosis module to obtain the fault category. The method can improve the cross-domain few sample fault diagnosis precision and robustness under the condition of very few target domain labeled samples.

The application provides a gas transient signal rapid detection method based on Mamba and ordinal metric learning, and relates to the technical field of gas detection and analysis. In view of the problems that the existing gas sensor relies on steady-state signals, resulting in long detection time, insufficient utilization of transient signals, and high intra-class variance and poor inter-class separability of transient signals, the application proposes a MOML framework based on Mamba encoder and triplet loss metric learning. The framework uses the Mamba encoder to extract long sequence features from the transient dynamic response signals of the gas sensor and map them to the embedding space; the embedding space is optimized through the triplet loss to enhance the intra-class compactness and inter-class separability; especially for the gas concentration regression task, a neighborhood negative sampling strategy is designed to realize ordinal metric learning. The experimental results show that the application can significantly improve the feature separability, improve the gas type recognition accuracy and concentration prediction accuracy, and realize low-delay and high-precision gas detection.