128 results about "Time-varying network" patented technology

The invention discloses a millimeter wave radar human body tumble detection method based on multi-feature fusion, and the method comprises the steps: obtaining an original point cloud frame from each radar, and carrying out the timestamp calibration and space coordinate system conversion; performing noise filtering, ground segmentation and human body point cloud extraction on each frame of point cloud; dividing the preprocessed point cloud into a static cluster and a dynamic cluster; filtering false dynamic clusters; extracting a residual dynamic cluster set, and identifying and tracking the human body dynamic clusters in continuous frames by adopting a tracking algorithm; extracting a corrected time sequence feature from the tracked human body cluster; inputting the time sequence characteristics into a pre-trained deep time sequence network, learning a falling time sequence dependency relationship, and outputting an abnormal index; and performing multi-source fusion with the abnormal index to obtain a comprehensive index to judge whether to trigger an alarm. The method can adapt to a complex home environment, improves the detection accuracy and real-time performance, and reduces the false alarm and missing alarm.

The invention discloses a building health state detection system, and relates to the technical field of photogrammetry and image processing, the system firstly obtains four-dimensional space-time coordinate data of a plurality of monitoring points on a target building based on satellite image analysis, constructs a four-dimensional space-time network model containing spatial topology and time sequence evolution information according to the four-dimensional space-time coordinate data, and then obtains a four-dimensional space-time network model containing space topology and time sequence evolution information; the system maps network model nodes to a predefined standard reference network associated with a local gravity direction, then, the system analyzes dynamic relevance in the four-dimensional network model by adopting a graph self-attention network, and meanwhile, calculates relative relation changes of monitoring points relative to the predefined standard reference network through geometric analysis; the deformation interpretation module distinguishes and determines internal relative deformation, overall displacement and absolute posture change (including settlement and inclination) of the building according to the two types of analysis results and by referring to surrounding environment information, and the state evaluation module integrates the information to generate a health state report containing risk levels.

The invention provides a dynamic space-time air quality prediction method based on a physical constraint graph attention network, and the method comprises the steps: carrying out the collection and preprocessing of multi-source data, and obtaining the pollutant concentration and meteorological element historical sequence of a monitoring station; constructing a dynamic space-time diagram network fusing distance attenuation and a wind direction driving transmission path based on the longitude and latitude of the station and the real-time meteorological field; multi-cycle time sequence features are extracted in a self-adaptive mode through an enhanced time sequence network ETNet, and attention weight distribution is guided through physical priori; time sequence features are embedded into the dynamic graph, a physically constrained graph diffusion attention network PC-GDAN is input, and space-time diffusion modeling is achieved through a differential operator embedded into an atmospheric diffusion equation; time sequence and space features are fused, mass conservation and diffusion smoothness constraints are introduced in decoding prediction, and a concentration prediction result conforming to a physical rule is generated. The method effectively fuses physical mechanisms and data, has strong space-time modeling capability and physical interpretability, and significantly improves the precision and robustness of air quality prediction.

The invention discloses a multi-information fusion hydraulic cylinder internal leakage fault diagnosis method, and relates to the technical field of hydraulic cylinder fault diagnos.The hydraulic cylinder internal leakage fault diagnosis method comprises the steps that firstly, a hydraulic cylinder heavy load, time-varying and nonlinear motion curve and a test system based on a multi-source sensor are constructed, and internal leakage fault evolution under the actual working condition is simulated; and secondly, developing a hybrid deep learning semi-supervised feature fusion framework based on a convolutional auto-encoder, a multi-head attention mechanism, residual connection and a bidirectional sequential network, and enhancing the characterization learning ability of a multi-source signal in a complex and multi-noise environment through omnibearing deep feature decoupling. And finally, feature extraction and fusion are realized by using the displacement error and the pressure signal, and fault diagnosis is completed. Experiments show that compared with a comparison model, the diagnosis accuracy of the provided model and feature fusion strategy is improved by 3.95%; under-3 decibel noise, the accuracy is improved by 2.33%, and the scheme has high accuracy, reliability and generalization performance.

The invention discloses a multi-layer heterogeneous sequential network feature alignment and clustering method and device based on tensor self-representation. The method comprises the steps that sequential feature matrix sequences {},..., {} of data views in a multi-view heterogeneous network in time windows are obtained; based on the matrix sequence of the first time window, constructing a Laplacian regular term by introducing a self-representation learning mechanism, modeling multiple views by adopting a tensor structure and introducing a Schatten p-norm regularization mode to construct an optimization objective function for optimization, and representing shared feature representation of each data view in the first time window; and solving the function and carrying out clustering processing on the obtained function, wherein the clustering result of each time window is used for carrying out characteristic analysis on nodes in the network. According to the method, on the premise that the original structure and semantics of the network are guaranteed, unified modeling can be carried out on the incomplete multi-view heterogeneous network, and feature expression is aligned.

The invention discloses a video deblurring method based on 3D space-time grid perception, a medium and equipment. The method comprises the following steps: firstly, receiving a continuous N-frame video image sequence, and taking [N, H, W, 3] as an input dimension vector; the method comprises the following steps: calculating the shake intensity through the optical flow amplitude of adjacent frames, adaptively dividing each frame of image into S * S overlapped grid blocks, and extracting the grid blocks corresponding to pixel positions to form a [N, S, S, 3] sequence; inputting the grid block sequence into a 3D space-time network model for processing, and outputting deblurred grid blocks; and a fusion weight is calculated based on the distance between the pixel and the center of the adjacent grid block, the grid block is weighted and recombined, and a clear current frame image is output. According to the method, efficient and real-time deblurring is realized by utilizing a 3D space-time network and grid block processing in combination with scene shaking characteristics, the problem of blurring of port container monitoring pictures can be effectively solved, and the operation safety is improved.

The invention provides a low earth orbit satellite network task unloading and route planning combined scheduling method based on a time expansion graph and a related device, and belongs to the technical field of wireless communication. According to the method, a dynamic time-varying low earth orbit satellite network is constructed; expressing the constructed dynamic time-varying low earth orbit satellite network as a directed network topological graph based on a time sequence by using a time expansion graph; constructing a joint scheduling model by using the directed network topological graph based on the time sequence; setting constraint conditions of the combined scheduling model; based on the set constraint condition of the joint scheduling model, solving the objective function of the constructed joint scheduling model to obtain an optimal result; and performing task unloading and route planning combined scheduling on the low-orbit satellite network according to the obtained optimal result to obtain a task unloading and route planning combined scheduling result. The problem that the overall performance optimization effect is poor due to the fact that the coupling relation between task unloading and routing planning in the dynamic time-varying network scene is not comprehensively considered is solved.

The invention discloses a water level prediction method and device based on multi-modal fusion and a double-flow space-time network PredFormer, and the method comprises the steps: firstly, building a double-flow space-time network PredFormer model frame in a dynamic manner from a time sequence signal prediction problem as an entry point and considering that the water level and meteorological data have high nonlinearity and dynamic space-time dependence, and carrying out the prediction of a time sequence signal; obtaining spatio-temporal characteristics of rainfall data and water level data of a plurality of meteorological stations; thirdly, performing feature-level fusion on the obtained water level data time sequence features and multi-site rainfall data spatial-temporal features by using an efficient low-rank tensor fusion method; and finally, a multi-layer perceptron is used as an output prediction result. According to the method, the nonlinear time sequence relation problem of water level prediction is considered, time-space dynamic association of water level data and rainfall data can be fused, and meanwhile, the cross-site prediction problem of personality and generality of the water level data of the hydrometric station and the rainfall data of different meteorological stations is considered. Therefore, the method has strong generalization ability in water level prediction across multiple meteorological stations.

The invention provides an industrial internet intrusion detection method based on a diffusion model. The method comprises the following steps: cleaning and standardizing original industrial internet flow data; taking the standardized industrial internet traffic data as input, generating a synthetic sample for the minority class of samples by adopting a multi-layer diffusion model, and sequentially carrying out anti-standardization and format adjustment processing and data balance sampling on the synthetic sample to obtain balance sample data; performing feature extraction and dimension reduction on the balance sample; the local time sequence features are processed through a multi-head attention mechanism, and global dependency features of the weighted key features are obtained; taking the global dependency feature of the weighted key feature as input, and extracting a multi-dimensional time sequence feature containing a long-term time dependency feature through a multi-layer bidirectional SRU module; and performing intrusion detection classification according to the multi-dimensional time sequence characteristics. According to the method, attack recognition can be efficiently and accurately carried out when high-dimensional and high-time-sequence network flow data and data category imbalance are carried out.

The invention discloses a PHM-oriented equipment state monitoring and predicting method. The method comprises the following steps: acquiring a sound signal in equipment through a sensor; transmitting the sound signal data to a storage and analysis module, and storing the sound signal data in a database after the storage and analysis module receives the sound signal data; the storage analysis module inputs the sound data into the preprocessing module for preprocessing operation; inputting the preprocessed sound signal into a diffusion probability model for data balance processing; performing feature extraction and classification on the balanced data by adopting a parallel dynamic convolution-sequential network model; and judging the health state of the equipment according to the classification result. A wavelet denoising module, a diffusion probability model and a deep learning model are integrated in an equipment health state monitoring system, and acquired equipment sound data are processed and analyzed step by step. According to the method, the accuracy and reliability of fault prediction can be effectively improved, and more accurate decision support is provided for health management of equipment.

The invention discloses a sports market demand prediction method and device based on deep learning and a medium, and relates to the technical field of market demand prediction, and the method comprises the steps: collecting sports demand data, and carrying out the preprocessing; performing relation mining and graph structure learning on the preprocessed sports demand data through a graph attention space-time network to generate a macroscopic demand potential energy graph; performing potential area identification on the macroscopic demand potential energy diagram by adopting a pre-trained sports market simulation model, and outputting local demand prediction data; performing weighted fusion and error correction on the macroscopic demand potential energy map and the local demand prediction data, and outputting a sports demand prediction score; and making a sports market demand strategy according to the sports demand prediction score and the multi-granularity demand prediction report, and transmitting the sports market demand strategy to an enterprise manager through an enterprise decision support interface. According to the method, multi-level accurate prediction and decision support of sports market demands are realized through dual-mechanism cooperation of the graph attention space-time network and the space-time convolution.

The invention discloses a satellite secure transmission optimization method and system based on joint anomaly detection, mainly solving the problem of low end-to-end transmission reliability under the condition of network attack in the prior art, and the implementation scheme comprises the following steps: constructing a time-varying network topology according to constellation configuration; periodically updating a network topology and a satellite neighbor node set according to the law of satellite orbital motion; the network is monitored in real time based on an anomaly detection mechanism, whether the current network is subjected to interference attack and DoS attack is judged in time, and whether the current satellite can safely transmit data is judged; and according to indexes such as node states, link hops, detection precision, node load capacity and link transmission delay, and according to a network topology and a neighbor node set at the current moment, determining a routing control optimization problem, and guiding each satellite to select an optimal transfer satellite along a transmission direction according to an optimal solution to obtain an end-to-end optimal transmission path. The method can actively trigger a detection mechanism to quickly identify abnormal nodes, obtains a safe and reliable end-to-end transmission path, and can be used for a low earth orbit satellite network.

The invention discloses a hook load and torque prediction method based on machine learning, and the method comprises the following steps: S1, collecting original data, and carrying out the preprocessing of the original data, and obtaining a sample data set; s2, constructing a CNN-LSTM-Attention multi-dimensional sequential network model, and setting hyper-parameters of the model; s3, training the CNN-LSTM-Attention multi-dimensional time sequence network model by adopting the sample data set, so as to obtain a trained CNN-LSTM-Attention multi-dimensional time sequence network model; and S4, predicting the load and the torque of the hook by adopting the trained CNN-LSTM-Attention multi-dimensional sequential network model. According to the method, the hook load and the torque can be accurately predicted, the CNN-LSTM-Attention multi-dimensional sequential network model adopted for prediction can be explained through the SHAP, reference information can be provided in time before abnormal problems occur, decision making can be assisted, and therefore the torsional pendulum drilling operation efficiency is improved.

The invention provides an interest point recommendation method based on a dynamic hierarchical attention fusion space-time network, and the method comprises the following steps: 1, obtaining user historical sign-in and interest point feature data, and carrying out the preprocessing of the user data; 2, learning robust spatio-temporal information representation of the interest points through an adaptive spatio-temporal double-graph convolutional network module; 3, constructing a'interest point-functional unit-regional unit 'multi-level semantic hierarchical structure, and learning hierarchical semantic information representation of interest points by using a bidirectional hierarchical graph attention network module; 4, after fusing the two interest point representations, coding a user track and inputting the user track into an attention-frequency fusion Transform module, so as to obtain a multi-level semantic hierarchical structure; the module separates long and short term dynamic signals in a user behavior sequence through frequency analysis and dynamically fuses the long and short term dynamic signals with a self-attention mechanism to predict a next point of interest; and 5, training by adopting a combined loss function. According to the method, the complex characteristics of the user and the POI are effectively modeled, and the accuracy and the capability of capturing the dynamic preference of the user are improved.

The invention discloses a multi-agent conflict-free path planning method based on a directed space-time network and forward direction decision, which comprises the following steps: 1, acquiring environment information, constructing a grid network according to the environment information, and constructing a directed space-time road network for the grid network by using the space-time network, initializing starting and ending point information of the robot group according to the directed space-time road network; 2, establishing a forward direction decision model of the agent group, and avoiding conflicts by using constraint conditions of the model; and 3, solving a forward direction decision model of the agent group, and planning an overall optimal path. According to the method, the directed space-time road network can be constructed by utilizing the space-time network to accurately describe the environment, the agents and the tasks, the path of the agents is planned by utilizing the advancing direction decision model, and the possibility of multi-agent conflict is avoided, so that the feasibility of the path is ensured, and the efficiency and the safety of multi-agent path planning can be improved.

The invention discloses a counterfeit video detection method and system based on selective multi-scale space-time fusion module capture, and is suitable for the field of video counterfeit detection. In view of the problem of long-time memory loss of a forged video and the problem that the difference between adjacent frames is difficult to perceive due to the fact that the forged video is always subjected to smooth frame-by-frame processing in practical application, local time information of different scales and difference information between two adjacent frames are respectively captured by constructing a double-branch time network; and therefore, the capturing capability of the model on subtle differences in forged video samples is enhanced. In combination with a space inconsistency module and an information supplement module architecture, the method realizes video space-time multi-scale fine-grained modeling, and realizes space-time information interaction. According to the method, the problem that in the prior art, only a single time feature or a single space feature is depended on is solved, the defect that the generalization ability of a model is insufficient is overcome, and the accuracy and robustness of forged video detection are improved.

The invention discloses a subway network key station identification method considering cascade failure, an electronic device and a storage medium, and the method comprises the steps: simulating a cascade failure process after station failure based on a constructed subway weighted time-varying network and the initial state and degree value of each station in each time period, building a passenger flow distribution rule based on the consideration of origin and destination of passengers, and carrying out the identification of the subway network key station. And obtaining the metro weighted time-varying network after cascade failure caused by failure of each time period and each station. And based on the node failure rate, the network maximum connection rate and the network passenger flow intensity entropy after each station loses efficacy in each time period, calculating the comprehensive importance of each station in each time period, and identifying the key subway station in each time period based on the comprehensive importance. The method is used for assisting a public transport operator to identify key stations in each time period in a subway network, so that decision support is provided for emergency plan making under emergencies, and a scientific basis is provided for emergency operation scheduling and management of the subway network.

A semantic routing method for time varying networks, wherein the method is configured to provide a congestion-aware source-based routing for satellites, preferably LEO constellations, with a distributed name resolution system, comprising a multitude of routers, a multitude of border routers, and a service provider. Wherein the method comprises an announcement phase and a forwarding phase executed in a proactive mode or a reactive mode.

The invention discloses an electrocardiosignal noise reduction method based on deep convolution and a sequential network, which comprises the following steps of: firstly, obtaining a pure electrocardiosignal through a synthetic function or a public data set, and constructing a pure-noisy electrocardiosignal pair by adding myoelectricity noise, power frequency noise, baseline drift or a combination of the myoelectricity noise, the power frequency noise and the baseline drift; designing a time sequence network structure comprising an input layer, a multi-level residual layer, a channel pruning layer, a full connection layer and an output layer, extracting time sequence features in the residual layer by using causal convolution, improving the performance of the model in combination with normalization, an activation function and a residual connection and discarding mechanism, training the model in a supervised learning mode, and obtaining a time sequence network structure; a mean square error or a mean absolute error is adopted as a loss function, network parameters are optimized through back propagation, and finally, the trained model is deployed on electrocardiogram monitoring equipment, so that real-time noise reduction processing on actual electrocardiogram signals is realized. According to the method, the time sequence characteristics of the electrocardiosignals can be effectively reserved, the noise reduction precision and robustness are improved, and the method is suitable for various medical and health monitoring scenes.

The invention discloses a self-adaptive spatio-temporal network behavior pattern evolution analysis method and device and a storage medium, and the method comprises the steps: collecting agent multi-dimensional behavior data through machine learning, the agent multi-dimensional behavior data comprising financial data, social media data and news report data; performing data quality monitoring according to a user analysis target, and updating the multi-dimensional behavior data of the intelligent agent; performing tensor decomposition on the updated agent multi-dimensional behavior data to obtain target behavior data; according to the target behavior data, a self-adaptive space-time network model is constructed, and the type of the self-adaptive space-time network model comprises a directed graph, an undirected graph or a weighted graph; according to the self-adaptive space-time network model, behavior pattern evolution analysis is carried out, and a behavior pattern evolution analysis result is obtained. The behavior pattern evolution analysis is realized, and the analysis accuracy and the analysis depth are improved. The method can be widely applied to the technical field of behavior data analysis.

The invention relates to a wind power prediction method, system and device based on a multi-site dual-space-time network and a medium. The method comprises the following steps: acquiring historical data of a multi-site wind power plant, and preprocessing the historical data of the multi-site wind power plant to obtain preprocessed data; performing feature selection processing on the preprocessed data to obtain an optimized feature set; inputting the optimized feature set into a preset multi-site dual-time-space network to obtain a first future predicted value of each site and a second future predicted value of each site; the multi-site dual-time-space network comprises a long short-term memory network branch and a time convolution network branch; and carrying out weighted fusion on the first future predicted value and the second future predicted value to obtain a multi-station wind power predicted value. By adopting the method, the precision and robustness of multi-station wind power prediction can be improved, the clustering development requirement of the wind power plant can be better adapted, and powerful support is provided for power grid dispatching strategy optimization and safe and stable operation.

The invention discloses a satellite network time-varying resource scheduling method and device, equipment and a medium. The method comprises the following steps: identifying heterogeneous computing units carried by all nodes of a whole satellite network; constructing a heterogeneous computing resource matrix of satellite nodes, and constructing an on-satellite heterogeneous computing resource model; acquiring satellite orbit parameters, and generating a time-varying network link matrix based on the satellite orbit parameters; embedding the heterogeneous computing resource matrix into the time-varying network link matrix to form a network-computing fused time-varying resource matrix; dividing a satellite operation cycle into a time slot sequence, generating a full-cycle resource matrix sequence, and constructing a time-varying graph model of network calculation fusion; and satellite network time-varying resource scheduling is carried out based on the time-varying graph model. According to the time-varying network resource scheduling method and device, the technical difficulty of unified representation of network computing resources is overcome, time-varying graph theory connotation expansion under a network and computing fusion scene is achieved, the completeness and accuracy of a time-varying graph model are improved, and therefore efficient scheduling of time-varying network resources is achieved.

The invention discloses a homogenizing and cooling control method and system for an aluminum alloy extrusion part. The system comprises a multi-source data sensing module for constructing a sensor data set; the data preprocessing module performs space-time alignment, de-noising filtering and feature enhancement processing on the sensor data set, and outputs a standardized sensor data set; a cooling parameter generation module extracts spatio-temporal characteristics by using a deep spatio-temporal network model based on the standardized sensor data set, and generates an optimal cooling parameter combination; and the dynamic regulation and control module accurately controls the flow and pressure of a cooling medium by using a self-adaptive PID algorithm according to the optimal cooling parameter combination. Through intelligent regulation and control of modular collaboration and data driving, the problems of poor temperature uniformity and high energy consumption in a traditional cooling process are effectively solved, the consistency of mechanical properties of extruded parts is remarkably improved, and an efficient and reliable technical scheme is provided for aluminum alloy precision machining.

The invention discloses a trusted data synchronization system and method for a smart power grid, and belongs to the technical field of edge computing. In order to solve the problems of insufficient reliability, safety and adaptability in a data synchronization process in an edge weak network and time-varying network environment, a technical scheme of dynamically selecting an integrity check algorithm and an error correction algorithm based on node capability, network topology and load information is mainly adopted. According to the method, data synchronization path optimization, data integrity verification and error correction processing in an intelligent power grid environment can be realized, the data synchronization availability, safety and efficiency of the system are improved, meanwhile, the method dynamically adapts to network changes and node computing power differences, and the stability and robustness of the system are enhanced.

The invention relates to a computer-implemented method for transporting data (40) between an end-user device (20) and a server device (24) for providing a network service via a time-varying network (12), the time-varying network (12) including a plurality of nodes (14) that are interconnected intermittently in time.

This invention relates to the field of fertility data analysis technology and discloses an Internet-based regional fertility intelligent census system. By modeling data completion tasks through Markov decision processes, predicting completion value using deep Q-networks, achieving cross-period data calibration based on Bayesian temporal networks, and applying temporal smoothing constraints to ensure historical continuity, this system solves the technical problems of uneven distribution of missing historical fertility data, inconsistent detection technology standards, and limited supplementation costs. It achieves high-quality enhancement and spatiotemporal distribution analysis of regional fertility historical data.

The invention relates to the technical field of action recognition, and provides a hand action recognition method and system based on a lightweight cross-modal space-time network, and the method comprises the following steps: carrying out the preprocessing of a hand-washing video stream, and obtaining an RGB image sequence and a hand key point coordinate sequence; performing separation convolution operation of multi-layer space convolution and time convolution on the RGB image sequence, extracting motion-related features in combination with a channel attention mechanism, and generating a visual feature mark sequence; performing semantic extraction and time sequence compression on the key point coordinate sequence to obtain a structural feature mark sequence aligned with the X; and dynamically calculating a gating coefficient based on context information, fusing X and H to obtain fusion features, and outputting classification probability distribution of the hand washing steps. According to the method, a lightweight visual feature extraction structure is adopted, and a dynamic gating mechanism is introduced to realize adaptive collaborative fusion of visual and structural modes, so that high-precision and low-delay hand action recognition is realized on edge equipment.

The invention discloses a sea condition classification method and system based on a double-branch graph enhanced space-time network, and relates to the technical field of computers, and the method comprises the steps: receiving a multi-channel ship motion time sequence, carrying out the down-sampling of the multi-channel ship motion time sequence into an interlaced subsequence, and carrying out the down-sampling of the interlaced subsequence; performing feature extraction on each staggered subsequence based on a multi-layer perceptron, outputting compressed feature representation, and splicing all the compressed feature representations of the staggered subsequences to form time compression features; dynamically constructing an adjacent matrix between channels based on the time compression feature, and executing a graph convolution operation for information propagation to obtain a graph enhancement feature; and performing two-way fusion and classification based on the graph enhancement feature and the time compression feature, and outputting a sea condition classification result. According to the invention, through an efficient and lightweight network model, rapid classification of sea condition levels is completed on a resource-limited embedded device.

This invention discloses a low-latency speech enhancement method based on a microphone array. The method includes: setting a set of initial pole parameters; optimizing the initial pole parameters using an artificial neural network to obtain real poles; constructing orthogonal basis function models for each channel of the microphone array using the real poles and calculating the responses of filters of each order; performing frame segmentation and temporal feature extraction on the received signal from the microphone array, and estimating the adaptive beamformer weights constructed from the orthogonal basis function models using an improved temporal network; calculating the system response of each channel of the beamforming network based on the filter responses and beamformer weights to obtain the enhanced complete speech signal. This invention, by using an orthogonal basis structure beamforming network, can flexibly adjust the poles, increase the network's degrees of freedom, shorten the filter length, and reduce network latency; achieving better speech enhancement results with shorter filter lengths.

The present invention discloses a method and system for detecting forged videos based on capture by a selective multi-scale spatiotemporal fusion module, which is applicable to the field of video forgery detection. In view of the fact that forged videos usually have the problem of long-term memory loss, and in practical applications, forged videos are often processed smoothly frame by frame, making the differences between adjacent frames difficult to detect, the present invention constructs a dual-branch temporal network to capture local temporal information of different scales and the difference information between two adjacent frames, thereby enhancing the model's ability to capture subtle differences in forged video samples. Combined with the spatial inconsistency module and the information supplement module architecture, the present invention realizes fine-grained modeling of multi-scale spatiotemporal video and realizes the interaction of spatiotemporal information. The present invention overcomes the problem of relying only on a single temporal feature or spatial feature in the prior art, avoids the defect of insufficient model generalization ability, and improves the accuracy and robustness of forged video detection.