53results about How to "Rich feature representation" patented technology

The invention relates to the technical field of artificial intelligence and the field of financial science and technology, and discloses a speech recognition error correction method, which comprises the following steps of: training and generating a target field Chinese character text understanding model by collecting dialogue Chinese character texts of a target field; collecting pinyin data, and training to generate a pinyin understanding model; receiving to-be-detected voice data and converting the to-be-detected voice data into to-be-detected Chinese character texts and pinyin texts Inputting the Chinese character text and the pinyin text to corresponding understanding models to generate semantic features and pinyin features; combining the semantic features and the pinyin features to form joint features; inputting the joint features into a speech recognition error detection model, generating error probability distribution of each word, and outputting an error tag; and correcting the Chinese character text through a template error correction module based on the error tag. According to the method, more comprehensive feature representation is formed by combining semantic features and pinyin features, and the accuracy of speech recognition error detection is improved; and a template error correction module is utilized to effectively correct identification errors.

The invention discloses a micro-expression recognition method and system based on a state space model and double-flow fusion, and relates to the technical field of expression recognition, and the method comprises the steps: extracting a starting frame image and a top frame image of a micro-expression continuous video frame sequence; inputting the initial frame image and the top frame image into a micro-expression recognition model based on a double-flow fusion network, extracting an optical flow image, and performing face key point detection and clipping operation on the top frame image and the optical flow image respectively; carrying out image block embedding operation to obtain texture features of the top frame image after cutting and motion features of the optical flow image; the two features learn the global dependency relationship and the local correlation between the image blocks from the positive direction and the negative direction through a parallel multi-layer bidirectional Mama fusion module, and deep image features and motion features are extracted after feature fusion; and the deep features of the two modes are interacted through a double-flow cross attention classification module, comprehensive features are learned, and a plurality of classifiers are combined to identify and output the micro-expression type more accurately.

The invention provides a building material surface defect detection method based on infrared polarization imaging, which belongs to the field of image processing, and constructs a building material surface defect detection model through input, image preprocessing, a feature extraction module, a feature fusion module, a building material polarization enhancement image module, a building material surface defect detection module and output. Information of a polarization image, an infrared image and a depth image is fused, more comprehensive feature representation is provided, the model has higher recognition capability on fine defects, a self-attention mechanism further optimizes the focusing effect of the features, interference of background noise is avoided, it is ensured that a defect area is fully recognized, and the recognition efficiency is improved. The method has better adaptability and performance in the field of defect detection, can cope with more complex and variable building material detection tasks, and particularly shows stronger advantages when detecting tiny defects and complex surfaces.

The invention provides a multi-modal protein function prediction method and system, and the method comprises the following steps: carrying out the processing of a one-dimensional convolution layer and a pooling layer of protein sequence data, extracting the local dependence features in a protein sequence, and capturing the mode and key sequence features in the protein; using protein sequence data as input, and using a large-scale pre-trained protein language model to extract residue-level embedded information in a protein sequence as node features to be applied to the contact graph; the method comprises the following steps: taking protein sequence data as input, predicting the probability of contact between protein residues based on a Cb-Cb distance, and constructing a contact graph; and performing joint processing on features of different modes by using a multi-head attention mechanism and a residual structure, integrating data from different modes through feature fusion, and fusing the extracted key sequence features, the residue level embedded information and the contact graph. The prediction method is accurate in prediction and good in performance.

The invention discloses an insulator defect detection method and system based on RT-DETR improvement. The method comprises the following steps: collecting a disclosed insulator defect data set and obtaining an image as a data set through a video interception frame; the method comprises the following steps: carrying out improvement based on a target detection model RT-DETR, changing a backbone network into a StarNet star network to carry out feature extraction, carrying out further improvement, removing a P5 detection head, adding a P2 detection head, and improving a star convolution module in the StarNet star network; a multi-head attention mechanism is changed into a partial self-attention mechanism PSA, and further improvement is carried out to obtain an S-PSA attention mechanism module; a RepC3 convolution module is improved into a RepMob convolution module; combining the loss function by using Inner-IoU and CIoU to obtain an Inner-CIoU loss function, and carrying out loss calculation; and training the improved RT-DETR, verifying the improved RT-DETR to obtain an optimal weight, and loading the weight into the improved RT-DETR to carry out insulator defect detection. According to the invention, the defect detection is carried out on the insulator by using the improvement based on RT-DETR, so that the detection is more accurate and efficient.

The invention discloses a GNSS interference identification method based on spatial transformation, multi-scale feature fusion and KAN. The method comprises the steps of forming an interference signal data set; obtaining a time-frequency image training set and a time-frequency image test set; constructing a GNSS interference identification classifier composed of a spatial transformation module, a multi-scale feature fusion module and a KAN module; a trained GNSS interference identification classifier is obtained; and obtaining a final GNSS interference identification classifier. And finally, outputting the category of the GNSS interference signal acquired in real time by the GNSS interference identification classifier. According to the invention, the spatial transformation module is used to carry out translation, scaling and other operations on the interference signal time-frequency image, and the number of training samples is increased. And the multi-scale feature fusion module adopts parallel convolution kernels of different scales to extract multi-scale spatial features of the time-frequency graph, so that the richness of feature representation is increased, and the recognition accuracy is further improved. The KAN module uses a KAN to replace a traditional multi-layer perceptron to execute a final classification task, and a more complex nonlinear function can be fitted by using fewer parameters.

The invention relates to an underwater acoustic target recognition system and method based on multi-modal depth feature fusion, and belongs to the field of underwater acoustic target recognition. The method comprises the following steps: firstly, performing preprocessing on an obtained underwater acoustic target original audio and associated metadata, and constructing a multi-modal data set; and then the constructed multi-modal sample is input into an identification model, the model extracts deep representation of each modal through a multi-branch feature coding network, depth alignment and complementary aggregation of different modal features are realized by using a cross-modal cross attention mechanism guided by potential query, and a target identification result is output based on a decision network of mixed experts. According to the method, experimental verification is carried out on two disclosed underwater acoustic data sets, the experimental result verifies the effectiveness of the multi-modal deep fusion and hybrid expert adaptive decision strategy adopted by the method, and through mining the complementary advantages of acoustic features and semantic priori, the multi-modal deep fusion and hybrid expert adaptive decision strategy is obtained. And the robustness and generalization ability of the underwater acoustic target recognition system in the strong-noise and multi-working-condition environment are remarkably improved.

The invention discloses a lightweight fabric surface defect detection method and device, and relates to the technical field of target detection. The method comprises the following steps of: performing data preprocessing and division on a pre-acquired fabric image data set; secondly, the model is improved, down-sampling Conv in a backbone and a check is replaced with ADown, StarBlock is introduced into a backbone part C2f to form a C2fStar module, and PCDeect is introduced into a head part of a detection head; the fabric defect picture data set is trained through the built YOLOv8 + Aown + C2fStart + PCDeect detection network, and a model pre-training weight is obtained; inputting a to-be-detected fabric picture into the optimal weight model for fabric defect detection, and outputting a detection result; the device comprises a processor, and when the processor runs the computer program, the steps of the lightweight fabric defect detection method are executed. According to the method, the parameter quantity and the calculation quantity are effectively reduced in a fabric defect detection task, meanwhile, the industrial defect detection requirement can be met, and the model precision is improved.

The invention discloses an Alzheimer's disease deep joint learning diagnosis method based on multi-modal feature fusion, and provides a new method for constructing a 3D fusion image through multi-modal feature fusion of an MRI-p value under the condition that a 3D image structure is not damaged. End-to-end deep joint learning diagnosis based on a residual neural network and an attention mechanism is planned to be constructed, and feature information and position information related to AD are automatically learned to obtain a classification basis of diagnosis. Experiments include classification of three levels, respectively, disease diagnosis based on AD and HC, AD and MCI, and MCI and HC. By introducing gene data, feature information of disease diagnosis can be expanded, and more comprehensive feature representation is formed, so that a diagnosis model is promoted to better capture complex features of diseases. And inputting the single-mode MRI into the deep joint learning diagnosis model to carry out an experiment so as to verify the effectiveness of the multi-mode fusion image information.

The invention discloses a tomato maturity detection method based on deep learning, and is applied to the field of tomato maturity detection in a complex scene. Comprising the following steps: acquiring tomato picture data sets at different maturity stages in a complex scene; manually marking the position and the type of the tomato in the image, and dividing the data set; the method comprises the following steps: introducing a Swin Transform into a backbone network; dSConv is introduced into the neck part; constructing a loss function of the improved YOLOv8 model; an improved YOLOv8 model is trained; according to the method, improvement is carried out based on YOLOv8, richer global context information can be provided, the model is more excellent in the aspect of processing complex scenes and details, and therefore the accuracy of the model for detecting the tomato maturity in the complex scenes is improved.

The invention discloses a physical field prediction method based on a deep fuzzy physical information neural network, and the method comprises the following steps: S1, collecting physical field data, carrying out the preprocessing of the data with noise, and obtaining a processed training data set Dtrain and a test data set Dtest; s2, constructing a deep fuzzy physical information neural network model; s3, inputting a physical field training data set with Gaussian noise and a corresponding partial differential equation, and training the neural network model to obtain a trained neural network model; s4, physical quantities in different time-space coordinates in the test data set are calculated through the trained neural network model, and physical field prediction is achieved; the method solves the problems that the existing physical information neural network method does not consider the condition that noise exists in a training data set, the training effect of a deep neural network model is poor under the interference of a noise data set, and physical information in a partial differential equation cannot be accurately learned, so that the prediction performance of a physical field is seriously reduced.

The present application relates to the technical field of apple tree pest image processing, and specifically to a pest segmentation detection method for apple trees, the method comprising: obtaining an apple tree pest image; obtaining feature points in the apple tree pest image at various scales; obtaining auxiliary directions according to the gradient changes of each feature point in different directions, screening suspected pest feature points according to the number of auxiliary directions of all feature points; obtaining the pest probability of the suspected pest feature point according to the grayscale change difference and local density of each suspected pest feature point in different auxiliary directions; screening pest feature points according to the pest probability of the suspected pest feature point in the apple tree pest image at all scales, and identifying the location of the pest in the apple tree pest image to be detected according to all pest feature points. The present application can improve the accuracy of pest segmentation detection.

The invention relates to a hard disk health degree assessment and early warning method and device based on multi-model integration and a medium, and the method comprises the following steps: obtaining SMART log files running in real time in each hard disk node in a service cluster, and carrying out the preprocessing; the preprocessed SMART log file is input into a large model to be processed, the health degree score of each hard disk node is output, and the large model is integrated through multiple models; and judging whether the health degree score of each hard disk node exceeds a dynamic health threshold, if so, outputting an early warning signal, and if not, not outputting the early warning signal. Compared with the prior art, the method has the advantages that the risk that backup cannot be carried out due to the fact that the hard disk breaks down is reduced, and the purposes of cost reduction and efficiency improvement are achieved.

The invention provides an intelligent driving closed-loop control method based on a pre-training large model. The intelligent driving closed-loop control method comprises the following steps: S1, obtaining full-quantity heterogeneous input data in a vehicle driving process; s2, performing feature coding on the total heterogeneous input data to generate feature data suitable for pre-training large model processing; s3, inputting the encoded feature data into a pre-trained large model for feature extraction, and outputting a key value pair; s4, constructing a plurality of task head sharing key value pairs in a cascade form, and generating intention-level and track-level outputs; s5, outputting and generating a vehicle control instruction according to the intention level and the track level to control the driving state of the vehicle; according to the method, full-quantity heterogeneous input features in a closed-loop form are constructed, so that the use efficiency of data is remarkably improved, and the most comprehensive feature representation is extracted from multi-source input, so that an output result with higher quality is generated; according to the method, the problems of information loss and accumulative errors in a traditional modular system are effectively solved, and the accuracy and reliability of decision planning are improved.

The invention discloses a personnel abnormal behavior analysis method based on deep learning, and belongs to the technical field of image data processing, and the method comprises the steps: obtaining a personnel abnormal behavior detection data set D; constructing an improved YOLOv5 network based on the spatial feature enhancement module; training the improved YOLOv5 network to obtain a target detection model; based on the loss functions of the improved YOLOv5 network, the DeepSORT algorithm and the improved SlowFast network, training the SlowFast network to obtain a SlowFast model; and carrying out abnormal behavior identification on a target in the to-be-identified video sample based on the target detection model, a DeepSORT algorithm and a SlowFast model. According to the method, the problem of class imbalance is solved by using adaptive weight, and the accuracy of video action recognition is improved by using time information, so that the robustness and accuracy of a model to a complex scene are improved, the computing resource demand is reduced through lightweight design, and greater flexibility and efficiency are provided for practical application.

The invention relates to a subway passenger flow prediction method and device based on a multi-scale space-time diagram attention network. The method comprises the steps of collecting passenger flow data of subway OD pairs, subway line network structure data and land use type data; constructing an OD-level urban rail transit knowledge graph based on the collected data, and calculating node features of the knowledge graph, including calculating historical passenger flow features, calculating land use information, calculating accessibility of OD pairs and calculating correlation of the OD pairs; inputting node features of the knowledge graph into a multi-scale space-time diagram attention network, performing diagram filtering on the node features, and updating the node features of the knowledge graph in combination with a result of a space-time attention mechanism; and inputting the updated node features into the multi-scale LSTM model, and outputting a multi-scale prediction result of the subway passenger flow. Compared with the prior art, the method has the advantages of being more accurate in prediction, beneficial to intelligent scheduling and the like.

The invention relates to an artificial intelligence image processing method. The method comprises the following steps: S1, data reading: reading historical image data and a to-be-processed image from a storage medium; s2, data screening: according to statistical analysis of historical image data, determining an image feature needing to be screened as a first image feature, obtaining an interval corresponding to the first image feature, if a first feature value of the preprocessed to-be-processed image data accords with the interval, carrying out screening, and otherwise, transmitting to the next step; s3, data enhancement and synthesis: performing data enhancement operation on the remaining images, and synthesizing new image data by using a generative adversarial network; and S4, brightness level division and label mapping: converting the color image into a gray level image, calculating an average gray level value of the gray level image, and dividing the image into low brightness, medium brightness and high brightness levels according to comparison between the gray level value of each pixel point and the average gray level value.

The invention provides a cross-domain transfer learning laser ultrasonic layering defect detection method suitable for a composite material and application thereof.The method comprises the steps that firstly, multi-channel input containing a frequency domain, a time-frequency domain and Koopman operator characteristics is constructed for the problem of data distribution deviation caused by the batch difference of the composite material; secondly, extracting and fusing high-dimensional spatial-temporal features through a multi-channel attention mechanism and a graph convolutional network; and finally, migrating the defect features of the source domain to the target domain in combination with dynamic distribution self-adaption and domain adversarial training strategies. The method can effectively overcome model performance reduction caused by material attribute change, realizes high-precision layering defect identification and imaging under the condition of no target domain label, and is suitable for composite material detection in a complex industrial scene.

The invention provides a sequential access prediction method based on a multi-task fusion access interval and related equipment, and is applied to the technical field of data processing. The method comprises the following steps: processing access information of a target user in a preset time period, and generating historical access time information of the target user, historical access record information of the target user and real-time access time information of the target user; processing the historical access time information of the target user, the historical access record information of the target user, the real-time access time information of the target user and the physiological feature information of the target user based on the target access result prediction model, and generating an interval prediction sub-task and a sequential access prediction sub-task; processing the interval prediction subtask and the sequential access prediction subtask to generate an initial prediction feature vector; and processing the initial prediction feature vector based on the access purpose information of the target user to generate target prediction information of the target user.

The invention relates to the technical field of data processing, provides a hybrid internal threat detection method based on DACGAN-Transform, and solves the problems of data imbalance and lack of fine-grained analysis of a detection method in internal threat detection. According to the method, samples which are similar to normal data in distribution but have abnormal characteristics are generated through a generative adversarial network (GAN), a data set is enhanced, abnormity is preliminarily judged, hierarchical characteristic extraction is conducted on log data through a Transform model, abnormity detection comprises single log abnormity detection and context abnormity detection, and the fine granularity and accuracy of detection are improved. And finally, in combination with the overall abnormal score of the GAN and the layered abnormal score of the Transform, comprehensive evaluation is performed through a multi-layer perceptron, and whether the log entry is abnormal is judged, so that the internal threat detection precision and the system security are effectively improved.

The invention relates to the technical field of target detection, in particular to a multi-scale and multi-stage open semantic target detection method and device and electronic equipment, and the method comprises the steps: obtaining a training image data set; performing multi-scale feature alignment on the training image data set to obtain a vision and language alignment result; generating a new class of pseudo labels according to the alignment result of the vision and the language, training a vision and language pre-training model according to the new class of pseudo labels and the real labels of the classes, and performing multi-stage iterative optimization on the pseudo labels by the vision and language pre-training model in the training process; and performing open semantic target detection on the target image based on the trained vision and language pre-training model. Therefore, the problems of how to better understand the local features and improve the object detection precision are solved.

The invention discloses a method for detecting small defects on the surface of lightweight steel, and belongs to the technical field of industrial machine vision. The method is based on an improved YOLOv8n architecture and cooperatively works through three core technical means: firstly, a Swin Transform module is introduced into a backbone network to model long-distance spatial dependence and suppress complex background interference; secondly, a high-resolution P2 detection branch is constructed, shallow details and up-sampling semantic features are fused, and microdefect characterization is enhanced through bidirectional refining circulation; finally, P2 exclusive adaptive threshold focus loss ATFL is adopted, threshold update is only limited to P2 detection branch samples, and precise optimization of difficult and tiny defects is achieved in cooperation with a gradient directional return mechanism. According to the scheme, the detection recall rate and the positioning precision of the tiny defects are effectively improved while the model parameter quantity is remarkably reduced, and high-precision and light-weight industrial deployment is realized.

The invention provides a dual-channel-based multi-modal feature fusion remote sensing image classification method and device, and relates to the technical field of remote sensing image processing and analysis. Comprising the following steps: preprocessing remote sensing images of different modes and combining the images to obtain a plurality of image groups; a preset classification model is constructed, the preset classification model comprises a backbone network, a fusion module, a gating mechanism layer and a decoding layer which are connected in sequence, and the structure type of the backbone network comprises a twin backbone network, an isomorphic backbone network and a heterogeneous backbone network; according to the plurality of image groups, selecting one of a twin backbone network, an isomorphic backbone network and a heterogeneous backbone network from a preset classification model as a backbone network to obtain a corresponding target classification model; training the target classification model through the training set to obtain a trained target classification model; and inputting the verification set into the trained target classification model, and outputting a classification result. In this way, the accuracy of remote sensing image classification is high.

The present application relates to the technical field of machine vision, in particular to a kind of robust visual detection system of ton bag goods based on multi-cluster feature fusion and posture correction drive, the system includes non-local mean enhancement module, multi-feature clustering fusion module, geometric feature extraction module, dominant posture clustering module, posture correction transformation module and goods segmentation determination module, non-local mean filtering is carried out to the original image of ton bag goods to obtain enhanced image, the present application is equipped with non-local mean enhancement module and multi-feature clustering fusion module, can realize the efficient enhancement of ton bag goods image and multi-dimensional feature fusion, the space alignment weighted superposition of color cluster and texture cluster is generated Fusion segmentation chart, let the feature representation of ton bag goods homogeneous region be more comprehensive, accurate, effectively improve the integrity and effectiveness of geometric feature extraction, so that the feature analysis ability of system to ton bag goods image is significantly improved.

The invention discloses a target detection method and device suitable for an infrared imaging scene and a storage medium, and the method comprises the steps: collecting an infrared scene picture through an infrared camera, marking the shot picture, inputting the marked picture into a network, and extracting a multi-layer feature map through a newly constructed global feature extraction network; inputting the extracted feature map into a feature fusion network constructed by a feature pyramid to carry out multi-scale feature fusion; afterwards, in order to retain tiny target information in the shallow feature map, a jump connection mode is adopted, and the original feature map and the fused feature map are fused again to obtain final input. And an output mode of three detection heads is adopted, so that the detection of a large target is ensured, and the detection of a small target is also ensured. Under the infrared scene, the feature extraction of the target is ensured, and the detection accuracy is ensured at the same time.

According to the multi-input hierarchical infrared image super-resolution reconstruction method based on the collaborative CNN and Transform, through ingenious combination of the advantages of the CNN and the Transform, clearer detail information and edge structures can be reconstructed in a reconstruction task of an infrared image super-resolution image, the contrast ratio and definition are improved to a certain degree, and the reconstruction efficiency of the infrared image super-resolution image is improved. Unnecessary artifacts and noise are reduced, a clearer appearance is kept while high-precision textures and structures are recovered, and a good visual effect is achieved.

The present invention discloses a multimodal three-dimensional target detection method and device for autonomous driving, comprising the following steps: extracting image features from image data; extracting point cloud features from point cloud data; bidirectionally enhancing and fusing image features and point cloud features to obtain point cloud enhanced image features and image enhanced point cloud features; respectively decoding the point cloud enhanced image features and the image enhanced point cloud features to obtain high-resolution image features and decoded point cloud features, fusing the high-resolution image features with the decoded point cloud features to obtain point cloud features of fused image features; sending the point cloud features of fused image features to a three-dimensional region proposal network and a refinement module for three-dimensional target detection. The present invention adopts an attention-based fusion method, in which the fusion module only focuses on useful features and suppresses noise, fully utilizing the complementarity of the two modalities for efficient fusion, and providing an effective solution for environmental perception of autonomous driving systems.

The invention discloses a driver distraction detection method based on a fusion attention mechanism and Bi-LSTM, and the method comprises the steps: collecting an original image of a driver, carrying out the data preprocessing of the original image of the driver, and obtaining a target data set; constructing a driver distraction detection network model based on a fusion attention mechanism and Bi-LSTM, and inputting the target data set into the driver distraction detection network model for model training to obtain a trained model; evaluating the overall performance of the trained model according to the performance index, and obtaining a target detection model with the driver distraction detection capability; and the target detection model is deployed and migrated to an actual vehicle-mounted environment of the new energy vehicle, driver behaviors are detected in real time, and a real-time detection feedback result is obtained. According to the invention, the accuracy of distraction behavior detection is improved.

This invention relates to the field of visual inspection and discloses an outdoor frost detection method based on high- and low-frequency signal fusion. The method first performs high-frequency component enhancement preprocessing on the input image to highlight surface details. Then, a dual-branch detection network is constructed, where the main branch extracts global semantic features, and the high-frequency branch enhances high-frequency information through wavelet transform and learnable scaling. Next, the features from the two branches are fused across layers using learnable scaling, and a channel-space joint attention mechanism is introduced for calibration. Finally, the frost category, confidence level, and region location are output based on the fused features. This invention, by explicitly enhancing and utilizing high-frequency details in the image, effectively improves the detection accuracy and robustness of low-contrast, fine-textured frost regions in complex outdoor environments, enabling real-time, automated frost monitoring and early warning.

The application relates to the technical field of eye movement data processing, in particular to an eye movement feature map acquisition method and device, a terminal and a storage medium. A time sequence matrix is constructed through each eye movement data sequence; an attention weight matrix is determined according to the time sequence matrix and an attention mechanism, so as to distinguish the importance of different parts in the time sequence matrix; an eye movement space-time graph is constructed according to the time sequence matrix, and feature extraction is performed on the eye movement space-time graph, so as to obtain a multi-dimensional space-time feature graph; finally, a weighted eye movement feature graph is obtained by combining the attention weight matrix and the space-time feature graph, so that more comprehensive and effective feature representations of eye movement data can be acquired.