513 results about "Sequence modeling" patented technology

The invention relates to a remote education data processing system which comprises the following steps: under a remote teaching task, pre-defining a task intention and a data expectation point; a semantic timestamp and a task binding label are printed on each data fragment; mapping the collected confusion data including silence, eye movement drift and prediction into a unified learning semantic vector; a micro-expression + interactive behavior + time sequence decision path ternary modeling mode is introduced, and a potential cognitive intention corresponding to the feature combination is recognized; teaching context information is fused; constructing a cognitive state mapping model; reasoning a current cognitive state label from multi-modal sensing data; searching intervention track VS effect feedback data in a historical database; generating a predicted intervention behavior sequence by using a sequence modeling algorithm; a dynamic combination suggestion chain including light prompt, content reconstruction, personalized practice and tutoring invitation is adopted; and superposing the cognitive state sequences of all students into a group cognitive trajectory map.

The invention relates to an automatic modeling agent system based on an MCP protocol and a large model, and the system comprises a multi-modal input interface module which is used for receiving, analyzing and standardizing modeling instructions or intention descriptions inputted by a user through different media to form structured feature representations; the semantic understanding central module is used for performing semantic analysis on the structured feature representation by using a pre-trained large language model and outputting a user modeling intention; the MCP instruction compiler module is used for converting a user modeling intention into an MCP instruction sequence following an MCP protocol; and the modeling execution engine module is used for analyzing and automatically executing the MCP instruction sequence to complete modeling of the target three-dimensional model. According to the method, three-dimensional modeling can be completed without knowing professional modeling terms and concepts by a user, and the modeling automation and intelligence level is greatly improved.

The invention discloses a feature attention and bilinear gating fused speech emotion recognition method and device, and the method comprises the following steps: 1, collecting an audio file, obtaining corresponding label information, generating audio waveform and time frequency representation data through preprocessing, and constructing an audio waveform mask and a time frequency mask to mark an effective information region; 2, constructing a dual-path feature extraction module which comprises a time-frequency feature extraction module and a pre-training acoustic feature coding module; wherein the time-frequency feature extraction module models emotion correlation through local convolution and a multi-dimensional attention mechanism, and performs global time sequence modeling based on a bidirectional gated loop network; the pre-training acoustic feature coding module extracts high-level speech representation with high expression ability for emotion distinguishing by using a pre-training model; and step 3, constructing a feature fusion module and an emotion classification module, and combining with a dual-path feature extraction module to form a speech emotion recognition model.

The invention discloses a composite structure damage form monitoring method and system based on deep learning, and the method comprises the following steps: collecting multi-source monitoring data of a composite structure in a loaded state, and carrying out the preprocessing; reconstructing a damage evolution trajectory in a high-dimensional phase space by adopting a delay coordinate embedding method, and executing dimension reduction to generate a chaotic dynamics low-dimensional trajectory; extracting singular attractor features, and generating a singular attractor feature set; carrying out sequence modeling through an improved Linformer damage identification network, and generating a prediction vector; training an improved Linformer damage identification network based on the prediction vector, and introducing nonlinear dynamic constraints to generate a damage identification network of the nonlinear dynamic constraints; and performing damage form classification and damage evolution prediction. According to the method, dynamics and deep learning are fused, composite structure damage monitoring is achieved, and the method has the advantages of being high in accuracy, high in stability and reliable in early warning.

The invention relates to the field of recommendation systems, and particularly discloses a personalized recommendation method based on multi-modal behavior sequence modeling, which comprises the following steps of: acquiring multi-modal behavior data such as user text, image, time and place, preprocessing, and realizing dynamic fusion of the data by utilizing a multi-modal self-attention mechanism (MMSA) to obtain a multi-modal behavior sequence model; and the interest evolution of the user is accurately captured. An independent RNN module is adopted to model long-term and short-term interests of a user, and the long-term and short-term interests are combined through a self-learning weight coefficient, so that the change of the user interests is reflected more accurately. In addition, by introducing an online learning and incremental learning mechanism, model parameters are dynamically adjusted according to real-time feedback of the user, and it is ensured that a recommendation result can respond to user interest changes in time. According to the method, the defects of an existing recommendation system in the aspects of data fusion, time sequence modeling and real-time adaptability are effectively overcome, recommendation individuation and accuracy are improved, and the real-time updating capacity and scene adaptability of the system are enhanced.

The invention discloses a multi-source intelligence-driven end-network cooperative attack defense method, which comprises the following steps of: S1, respectively acquiring terminal log data, network flow data, threat intelligence data and equipment state data, and carrying out standardization processing on the terminal log data, the network flow data, the threat intelligence data and the equipment state data; s2, constructing a feature data hypergraph by taking the standardized data as nodes; s3, inputting the hypergraph into a depth hypergraph anomaly detection model, extracting high-order correlation features and detecting an anomaly mode; s4, inputting an anomaly detection result into a hypergraph Transform sequence modeling mechanism, performing multi-head attention calculation and dynamic interaction, and generating a dynamic fusion feature; s5, judging an attack behavior in real time according to the dynamic fusion features, and generating an end-network cooperative linkage defense strategy; and S6, issuing a defense instruction in real time according to the defense strategy, and executing end-network cooperative attack defense. According to the invention, the real-time performance and accuracy of end-network cooperative attack defense are improved, and the overall efficiency of network security defense is enhanced.

The invention discloses a writing brush calligraphy practice correction system based on real-time handwriting analysis. The system comprises a sensing layer used for collecting handwriting tracks, physiological signals, environmental parameters and ink mark characteristics; the edge calculation layer is used for executing noise filtering, coordinate system normalization, multi-modal data alignment and feature primary extraction through an Apache Kafka data pipeline; and the algorithm analysis layer is used for carrying out super-long calligraphy stroke sequence modeling by adopting an S4 architecture time sequence model, introducing a Neural ODE module for modeling, capturing dynamic characteristics in a continuous pen wielding process, constructing a multi-physics-field coupled PINN framework, constraining neural network prediction through a physical loss function, deploying a dual-stage characteristic extractor to extract high-order characteristics, and extracting the high-order characteristics. Book style features are extracted, and the current practicing book of the user is classified in real time; the intelligent correction layer is used for generating a dynamic correction suggestion, constructing a personalized learning path, realizing self-adaptive scoring and providing aesthetic dimension feedback at the same time; and the user interaction layer is used for providing an AR correction interface.

The invention discloses an electric power system safety early warning method and system based on multi-modal cooperation, and relates to the technical field of electric power system safety early warning, and the method comprises the steps: collecting multi-source operation data, carrying out the preprocessing, carrying out the multi-modal feature extraction and fusion based on the preprocessed data, and carrying out the multi-modal feature extraction and fusion. Inputting an edge detection model and outputting an abnormal confidence score in combination with an attention mechanism; and performing alarm grading according to the abnormal confidence score, constructing a causal diagram for alarms with high risk levels in combination with associated security events, and performing future attack path prediction by adopting a time sequence diagram neural network. According to the method, multi-scale convolution and a channel attention mechanism are fused, the extraction capability of the multi-source data time sequence features of the power system is enhanced, the anomaly detection precision is improved, dynamic attack path prediction is realized in combination with RMTPP and causal atlas topological constraints, sequence modeling is enhanced through self-attention and position coding, and the detection accuracy is improved. And the perspectiveness and the reliability of the safety early warning of the power system are obviously enhanced.

The invention discloses a refrigerated cabinet multi-sensor heterogeneous data fusion method based on artificial intelligence, and the method comprises the following steps: collecting multi-source heterogeneous data in the operation process of a refrigerated cabinet, and constructing a unified event time axis; preprocessing the multi-modal data, and generating a physical enhanced multi-modal tensor in combination with a heat-airflow-electricity model; constructing a disturbance activation vector and a modal mask pattern, and embedding the disturbance activation vector and the modal mask pattern into a multi-modal tensor; inputting the disturbance enhanced input data into an improved Tide model to perform long sequence modeling, and outputting a state prediction sequence; performing prediction residual analysis, constructing a dynamic anomaly scoring function, and fusing a disturbance activation vector and a modal mask graph to generate an anomaly type mark and confidence fusion state representation; sparse gradient uploading, parameter aggregation and structure synchronization are carried out among multiple devices by adopting a federal training framework. According to the method, the multi-modal data fusion precision and the anomaly recognition reliability of the refrigerated cabinet can be improved, and cross-equipment collaborative intelligent optimization is realized.

A system for the real-time delivery of personalized motivational content based on biometric information; the system includes: a biometric acquisition module configured to capture a variety of physiological signals from a user, wherein the physiological signals include at least heart rate variability, electrodermal activity, facial expressions and electroencephalographic (EEG) signals; a preprocessing module that is operationally coupled with the biometric acquisition module, wherein the preprocessing module is configured to remove noise, normalize and extract signal features from the physiological signals in real time; a multimodal biometric fusion engine configured to temporally align and synchronize the extracted features across signal modalities using dynamic time distortion and confidence-weighted interpolation; a motivational state inference model with a hybrid neural architecture comprising a Convolutional Neural Network (CNN) for spatial pattern recognition and a Recurrent Neural Network (RNN) for temporal sequence modeling, wherein the inference model is configured to output a motivational input score and an affective state classification; an engine for recommending motivational content, configured to select and prioritize content from a content repository based on motivational uptake score, user profile metadata, contextual signals including time of day and geolocation, and historical content effectiveness profiles; and a content delivery subsystem comprising one or more output modalities selected from an acoustic actuator, a visual display, a haptic actuator or an environmental controller, wherein the content delivery subsystem is capable of presenting the selected motivational content in a modality that is dynamically adapted to the user's current psychophysiological state.

The invention discloses an underwater sonar target recognition system based on multi-domain feature fusion and lightweight modeling, and the system comprises a Trifusion block, a novel lightweight attention residual network, a long and short time attention LSTM and a Mamba module which are connected in sequence, and achieves target recognition through multi-domain parallel extraction of fusion features, lightweight convolution and attention optimization, long and short time dependence capture and long sequence modeling. The method has the advantages that complex noise is comprehensively represented, the model efficiency and stability are improved, the bottleneck of time sequence processing is broken through, and the method has high performance, high robustness and wide adaptability.

The invention discloses a control system and method for an automatic beer production line, relates to the technical field of automatic control, is used for solving the problems that state judgment is lagged and abnormal trend is difficult to identify in time, identifies electronic tags of key equipment through an RFID reader-writer, imports models and serial numbers into a database, and automatically associates operation parameters. The distributed sensors collect temperature, pressure, flow and vibration data in real time, the analysis unit calculates the operation deviation and the health degree, a multi-condition control strategy is configured through the programmable logic controller based on the health degree, a time window and a task priority algorithm are combined, real-time data and historical data are compared to recognize abnormity, and the real-time data and the historical data are analyzed. And a fuzzy control algorithm is fused to generate an adjustment instruction, the equipment state is automatically adjusted, time sequence modeling is utilized to analyze whole-process data, a start-stop strategy is formulated, energy distribution and process optimization are performed, and intelligent scheduling and self-adaptive optimization of the beer production line are realized.

The invention discloses a power grid photovoltaic output and load sequence modeling method, system and device and a storage medium, and the method comprises the steps: comprehensively utilizing the multi-scale feature extraction capability of a time-frequency decomposition technology, the time sequence dependence modeling capability of a long and short-term memory network, and the global hyper-parameter optimization capability of a Bayesian optimization algorithm; and carrying out collaborative modeling and prediction on the photovoltaic output and the power load under a unified framework. By introducing a source load time-delay correlation analysis and probability interval construction mechanism, point prediction results and uncertainty intervals of photovoltaic, load and net load can be output at the same time, and a set of source load integrated prediction system with high prediction precision, strong robustness and reliable interval characterization capability is constructed. The method can improve the precision and reliability of photovoltaic power and load prediction, also can reduce the risk in power system scheduling, optimizes the energy storage configuration strategy, and especially has wide popularization potential and application prospects in the scenes of new energy grid-connected operation, intelligent micro-grid and virtual power plant management and the like.

The invention relates to the technical field of artificial intelligence and computer vision, discloses a subway key component fault detection method and system based on an AI visual large model, and aims to solve the problems of low detection precision, weak generalization ability, insufficient multi-mode understanding, poor real-time performance and lack of state evolution modeling in the prior art. The method comprises the following steps: acquiring images of key components through a multi-view industrial camera array, and performing distortion correction, illumination normalization and noise suppression; a pre-trained visual large model is utilized to extract deep space features, and modeling is carried out on a continuous frame feature sequence through bidirectional LSTM to capture a time sequence change trend. By introducing the large-scale visual large model and spatio-temporal joint modeling, the identification capability of tiny defects is improved, the discrimination stability is enhanced, high-precision and low-delay automatic detection is realized, the false alarm rate and the omission ratio are remarkably reduced, and the detection efficiency and the system maintainability are improved.

The invention provides an industrial equipment residual service life prediction method based on a DMM-JA model, and relates to the technical field of industrial equipment predictive maintenance, and the DMM-JA model comprises a dynamic bimodal fusion and multi-scale feature extraction module DBF-MSFEModule, an LSTM-Mama mixed sequence modeling module, a jump perception attention module and an output layer. The method comprises the following steps: preprocessing bimodal sensing data of industrial equipment; the preprocessed bimodal sensing data is processed through a dynamic bimodal fusion and multi-scale feature extraction module DBF-MSFEModule, and multi-scale fusion features are obtained; the multi-scale fusion features are input into an LSTM-Mamba mixed sequence modeling module, and joint time sequence features are obtained; the joint time sequence features are corrected through a jump perception attention module, and robustness features are obtained; and inputting the robustness characteristics into an output layer to obtain an RUL prediction result of the industrial equipment. Key features in the equipment degradation process can be accurately captured, and the accuracy of residual service life prediction is improved.

The invention discloses a mixed Mamb-Attention air quality prediction model based on multi-scale decomposition and a construction method of the mixed Mamb-Attention air quality prediction model. The method comprises the following steps: firstly, constructing a multi-scale decomposition module (MSD), decomposing an input time sequence into a trend term, a season term and a residual term through a parallel sliding window group, and realizing cross-scale feature fusion by utilizing group normalization and convolution; then designing a periodic pyramid module, extracting multi-level periodic features based on fast Fourier transform (FFT), and enhancing the perception ability of the model to different time scale periodic laws; the Mama branch is used for capturing long-range dependence, the self-attention branch is used for extracting a local dynamic mode, and the output of the Mama branch and the output of the self-attention branch are fused through residual connection and layer normalization; and finally, the prediction head module completes feature aggregation and result output. The model gives consideration to long sequence modeling capability and calculation efficiency, can accurately capture multi-scale dynamic change and non-stationary features in air quality data, improves the precision and stability of air quality prediction, and has good practical value and popularization prospect.

The invention discloses a location social network service recommendation method based on trajectory collaborative filtering and Mama. The method comprises the following steps: firstly, collecting service interaction data of a user in a location social network; carrying out representation modeling on the service interaction behavior based on static and dynamic joint representation learning, and generating a point-of-interest static representation embedding vector and a user interest signal embedding vector; constructing a service interaction behavior sequence modeling network based on a state space machine model Mamba, embedding and inputting a user interest signal into the network, modeling a user long and short term interest state transition mode field through the state space model, and outputting a prediction interest point matching embedding vector; and finally, optimizing the model through a minimized cross entropy loss function, generating a user interaction service recommendation list based on the normalized score matrix, and taking the first K interest points with the highest score as recommendation results. According to the method, a unified user long-term interest transfer mode field and a short-term interest response mode are constructed, and the problem of popularity bias of a recommendation system is relieved.

The invention discloses a multi-intersection traffic signal cooperative control method driven by a cross attention neural network, and the method employs a local cooperative Transform architecture, integrates a decision converter and a shared memory mechanism, and achieves the efficient modeling of a space-time dependence relation of a multi-intersection traffic state. The method comprises the following steps: firstly, through a memory head module, extracting a hidden state of each agent in a sequence modeling process, and updating global shared memory for supporting information interaction and strategy collaboration among multiple agents; and then, a cross attention module is adopted to carry out cross calculation on the local representation and the shared memory of each agent, so that dynamic perception and efficient modeling of the global state of the traffic system are realized. A backbone network of the model is based on Transform, and the understanding ability of time and space traffic characteristics is enhanced through position coding, self-attention and cross attention mechanisms. In the fine tuning stage, only the inserted Adapter module and the output layer are subjected to parameter updating.

The embodiment of the invention discloses a traffic state prediction method based on multi-modal data and adaptive topology modeling. The method comprises the following steps: dynamically determining an adjacent matrix between roads according to the type and historical traffic flow of each road, and respectively inputting a GCN model and a GAT model according to an adjacent matrix dynamic traffic network diagram; extracting global spatial features by the GCN model according to the weight of each edge and the current feature representation of each node; strengthening the effect of a key node by the GAT model to obtain a local attention space feature; performing short-time timing sequence modeling and long-time timing sequence modeling on the time sequence of the multi-modal traffic data of each node in the latest period of time to obtain a short-time feature and a long-time feature respectively; and predicting a future traffic state according to the global spatial features, the local attention spatial features, the short-time features and the long-time features. According to the embodiment, the traffic state prediction accuracy is improved.

The invention discloses a sleep classification method and system based on time-frequency combination, and belongs to the technical field of biomedical signal processing and artificial intelligence. In order to solve the problems that an existing method is high in manual dependence, insufficient in time-frequency feature fusion and low in long-time-sequence modeling efficiency, sleep stage classification is carried out mainly through automatic time-frequency feature extraction, a dynamic attention mechanism with memory enhancement and a lightweight multi-branch neural network structure. According to the method, efficient modeling and accurate classification of the multi-scale electroencephalogram signals can be achieved, the deep sleep recognition capability and the overall classification accuracy are improved, meanwhile, the model parameter quantity and the reasoning delay are remarkably reduced, and good real-time performance and clinical applicability are achieved.

The invention discloses an Internet of Things equipment long protocol text sequence modeling method and device, equipment and a medium, and relates to the cross technical field of the Internet of Things communication technology, deep learning model optimization and the natural language processing technology, and the method comprises the steps: collecting Internet of Things equipment protocol texts of a plurality of manufacturers; performing data cleaning, word segmentation and semantic marking operation on the collected protocol text to generate a structured sequence; designing a hierarchical residual position correction module based on dynamic residual position coding, constructing a protocol model, and training the protocol model through the structured sequence; and analyzing the to-be-analyzed protocol text based on the trained protocol model, and generating a mapping rule and type conversion logic between the to-be-analyzed protocol text and the target protocol. The problems of position information attenuation, gradient instability, poor dynamic adaptability and the like in long protocol analysis can be effectively solved.

The invention discloses a single-sample time sequence knowledge graph reasoning method based on long sequence modeling, and belongs to the field of time sequence knowledge graph reasoning. According to the method, an attention mechanism and co-occurrence analysis are combined, features are aggregated from adjacent entities, and therefore tracking of the historical evolution trend is achieved, specifically, the attention mechanism is adopted to integrate relation information in entity neighborhoods, and the relevance between the adjacent entities and the importance of the neighborhoods of the adjacent entities are clearly explored in combination with the co-occurrence analysis. And furthermore, an encoder based on a state space model is adopted to encode the time sequence interaction of the entity, so that the method not only can efficiently model a long sequence, but also can capture long-term dependency. And finally, introducing a measurement network, applying the learned representation to the measurement network, evaluating the similarity between the support entity pair and the query entity pair from multiple perspectives, and ensuring that the possibility of reflecting future events is comprehensively evaluated. According to the invention, a comprehensive experiment on a widely used time sequence knowledge graph data set shows that the method is obviously superior to a baseline model.

The invention belongs to the technical field of intelligent traffic, and discloses a multi-scale space-time fusion image feature extraction method based on traffic flow, which comprises the following steps of: 1, constructing a dynamic image generation module; a self-adaptive adjacency matrix is generated in combination with historical traffic data, spatial embedding and time embedding, the matrix is used for spatial-temporal feature extraction of a GCN layer, and the generated adjacency matrix can flexibly capture static and dynamic relationships; 2, constructing a learnable weighting module; according to the module, the weights of different features are adaptively adjusted, so that various feature information is effectively fused; 3, constructing a sequence feature mapper; through a recurrent neural network (RNN) and a sequence compression-expansion mechanism, dynamic features of time sequence signals are effectively extracted and enhanced. According to the method, adaptive adjustment of an adjacent matrix is realized through a dynamic graph generation module, the multi-feature fusion capability is improved in combination with a learnable weighting mechanism, and the long sequence modeling capability of an RNN layer is optimized by adopting a sequence compression and expansion strategy.

The invention discloses a coal rock instability precursor identification method and system based on fracture topological parameters, and the method comprises the steps: obtaining sound wave data through micro-seismic monitoring, and extracting coal rock fracture core parameters in combination with seismic source positioning and seismic source mechanism inversion; building a fracture network topology model based on a graph theory method by considering the geometric dimension and direction characteristics of the fracture, and obtaining fracture space-time topology evolution parameters to build a multivariable time sequence topology characteristic data set; introducing a multivariable time sequence graph model MTAD-GAT, dynamically capturing a structural dependency relationship among topological parameters by using a graph attention mechanism, mining collaborative anomaly characteristics of the structural dependency relationship, and capturing an evolution trend of the structural dependency relationship in combination with a time sequence modeling module; and fusing the prediction error and the reconstruction error to form an abnormal score as an instability precursor index, extracting covariant features and evolution trends of a plurality of key topological indexes in the precursor stage, and assisting in distinguishing an instability critical state, thereby realizing unsupervised automatic identification and intelligent early warning of the topological precursor.

The invention provides an intelligent medical multi-round dialogue diagnosis reasoning method and system based on deep learning, and relates to the technical field of deep learning, and the method comprises the steps: constructing time sequence features through employing an attention mechanism for time sequence symptom description, and carrying out bidirectional sequence modeling to generate comprehensive symptom features; using the medical knowledge graph to detect logic contradictions and information loss to generate standardized features; calculating information gain of an inquiry direction based on a deep neural network to generate optimal inquiry content; and iteratively updating according to user feedback until the diagnosis information entropy is lower than a threshold value, and outputting a diagnosis result. According to the invention, the accuracy and efficiency of medical diagnosis are improved.

The invention discloses a unified information extraction method based on Transform, belongs to the technical field of natural language processing, and realizes unification of information extraction tasks through a sequence-to-sequence modeling mode. According to the method, parameter redundancy among the independent models is effectively reduced, and accurate multi-dimensional information extraction of entities, relationships, classification and the like can still be realized in a low-resource scene through the pre-training language model. Meanwhile, the finite-state machine is adopted to restrain the generation process, accurate structuring of the output result under the complex text condition is further ensured, and the overall extraction effect and the flexibility of system application are remarkably improved.

The invention provides a user behavior analysis and interest recommendation method which fuses time-assisted features, introduces comparative learning and uses meta-learning optimization. According to the method, a user interaction sequence is enhanced by using a time homogenization strategy, and optimization is performed by using contrast learning, so that a next interested article is recommended for a user. The specific process comprises the following steps: preprocessing original interaction data, and performing time-level data enhancement on a user interaction behavior sequence, so that the original interaction sequence becomes a more uniform sequence; and then constructing a comparative learning framework, performing comparative training on the original sequence and the enhanced sequence, and optimizing a comparative learning result by constructing a positive and negative sample pair, combining the original loss and the comparative loss and adopting a meta-learning mechanism. Moreover, a multi-head self-attention mechanism is introduced to model a context relationship between nodes, user behavior features in time evolution are captured, and the discrimination capability of feature representation is improved. According to the method, on the basis of fusing time information, contrast learning and meta learning, the sequence modeling capability in a data sparse environment is effectively improved, and accurate description of user interest dynamics is realized.

According to the Mama network solenoid valve fault diagnosis method based on the frequency domain characteristics, the problem of solenoid valve system fault diagnosis can be solved. According to the method, a pneumatic solenoid valve fault data set is constructed by collecting operation signals such as voltage and current, and multi-scale frequency domain features of the signals are extracted by adopting Wavelet Packet Transform (WPT) and Discrete Fourier Transform (DFT). WPT-DFT preprocessing features are embedded into an improved channel-space joint attention mechanism module, and the improved channel-space joint attention mechanism module is combined with a Mamba network with extremely high sequence modeling capability to construct an end-to-end fault diagnosis model. Compared with a traditional convolutional neural network, the method can more effectively capture deep dynamic features in a time sequence and highlight a key frequency region, and experimental results show that the method has higher diagnosis precision and generalization ability and is suitable for intelligent detection of the electromagnetic valve under complex working conditions.

The invention discloses a texture perception state space modeling method for an image restoration task, and the method comprises the steps: 1, constructing a region selection mechanism based on texture complexity, and enabling the region selection mechanism to be used for distinguishing a flat region and a high-texture region in an image; 2, introducing a texture modulation mechanism, and performing explicit adjustment on a state transition matrix in the state space model; 3, enhancing the context modeling capability of the model through a multi-direction sensing module; and 4, by combining position embedding and a sequence modeling structure, the capability of the model in the aspects of image structure understanding and spatial information maintenance is improved. The method can effectively alleviate the problem of information loss when a traditional image restoration method processes texture details, improves the structure restoration capability of a complex region, gives consideration to the restoration quality and the calculation efficiency, is suitable for multiple image restoration scenes such as image super-resolution, image rain removal, low-light image enhancement and the like, and improves the image restoration efficiency. And the method has good engineering adaptability and actual deployment value.

The invention relates to the technical field of intelligent agriculture and computer vision, in particular to a small-sample high-density chicken counting framework based on a deep learning Mamba structure, which comprises a feature extraction network, a support-query enhancement module and a decoder. According to the method, a multi-scale feature extraction network based on a residual block (ResNet Block) is introduced, so that local detail information is effectively reserved; and then, a support-query enhancement module is constructed by introducing a Mamba structure, and context interaction between support features and query features is effectively enhanced by utilizing the long sequence modeling capability of linear complexity of the support-query enhancement module, so that the problems of individual overlapping and boundary fuzziness in a high-density chicken flock scene are solved. Experimental results on a PoultryCount real breeding data set show that the mean absolute error (MAE) and the root-mean-square error (RMSE) of the method are reduced compared with those of an existing method, and the chicken counting precision and generalization ability under the condition of a small number of labeled samples are remarkably improved.