234 results about "Spatiotemporal encoding" patented technology

The invention relates to the technical field of data processing, and discloses a comprehensive management method and system for a smart park, and a storage medium. The method comprises the steps that environment, energy consumption and people flow data are collected through the Internet of Things, and a basic data set is obtained through edge node processing and semantic alignment; constructing a digital twin system by using a spatio-temporal data cube and a graphical mode; constructing a land digital identity system by adopting space-time coding and block chain technologies; performing anomaly identification and risk analysis on the activity data to obtain supervision early warning data; performing multi-dimensional analysis on the enterprise operation data to form an enterprise portrait; and presenting resource distribution through mixed reality, and analyzing cross-department data to obtain operation efficiency data. According to the invention, holographic perception, intelligent supervision and cooperative service of park elements are realized.

The invention discloses a cross-source data three-dimensional reconstruction method and system based on improved Gaussian sputtering, and the method comprises the steps: collecting an unmanned plane inclined image and a ground panoramic image of a target region, and constructing a time-space correlation data set; based on multi-view geometric constraints, space-time coding matching point pairs are established through an adaptive feature pyramid, intelligent incremental cross-source data sparse reconstruction is carried out, and point cloud and camera parameters are output; adopting improved Gaussian sputtering, compressing a three-dimensional Gaussian kernel into a two-dimensional Gaussian primitive through double tangent vector constraint, and fitting surface geometry to realize multi-scale reconstruction; and optimizing primitive parameters by using a differentiatable renderer, completing multi-scale fine reconstruction through gradient back propagation, and generating a high-precision three-dimensional model. According to the method, multi-scale accurate geometric prior input and accurate camera poses are provided for three-dimensional reconstruction, the dependence on professional manual operation in a traditional three-dimensional reconstruction method is greatly reduced, and meanwhile, the geometric accuracy and visual fidelity of a reconstruction result are remarkably improved.

The invention provides a cellular network flow space-time prediction method and system based on a hybrid neural network, and belongs to the technical field of intelligent communication. The method adopts a layered deep neural network architecture, and comprises a data embedding layer, a space-time coding layer, a feature fusion layer and an output layer. The data embedding layer maps a historical traffic sequence, cross-domain external data and metadata into high-dimensional features; the space-time coding layer is used for respectively fusing one-dimensional causal convolution and a Mama neural network to extract multi-scale time features and densely connecting convolution and a multi-head attention mechanism to capture multi-scale space features through time and space modeling branches; the feature fusion layer realizes adaptive weighted fusion of spatial-temporal features, cross-domain features and metadata features by using a gating fusion mechanism; and the output layer performs linear transformation on the fusion features to generate a final prediction result. According to the method, the spatial-temporal dynamic capture of the service traffic is accurate, the prediction curve is highly fit with the true value, and the accurate prediction of the multi-service traffic of the cellular network is realized.

The invention discloses a multi-modal traffic flow prediction method based on multi-source data feature fusion. The method comprises the following steps: acquiring traffic data; dividing a training set and a test set according to the traffic data; constructing a traffic prediction model, and training the traffic prediction model by adopting the training set; verifying the traffic prediction model by using the test set; the traffic prediction model comprises a cloud image encoder used for carrying out feature extraction on cloud image data to obtain cloud image features; the space-time encoder is used for performing feature extraction on the space-time data to obtain space-time features; the fusion module is used for fusing the cloud image features and the spatial-temporal features to obtain fusion features; and carrying out addition operation on the cloud picture features, the spatial-temporal features and the fusion features, and inputting into a spatial-temporal decoder for prediction to obtain a final prediction result. According to the method, multi-azimuth and multi-dimensional information is fused, redundant information in the spatial-temporal feature extraction process is simplified, efficient and high-precision prediction is achieved, and the method can be widely applied to the technical field of traffic flow prediction.

The invention relates to the technical field of artificial intelligence processing, in particular to a manufacturing and processing system based on artificial intelligence simulation control parameters. According to the manufacturing and processing system, three-dimensional space-time coding is carried out on vibration, temperature and deformation time-varying signals through a multi-mode sensing module; generating a composite feature body including an equipment rigidity distribution matrix, a thermal deformation gradient vector and a material residual stress tensor, and solving the problems of spatial-temporal asynchronization and physical field feature splitting of multi-source data; the parameter resolving module constructs a parameter incidence matrix based on a fractal neural network, combines dynamic constraint conditions of thermal deformation gradient vectors, performs multi-objective optimization on a rigidity distribution matrix and residual stress tensor, generates an initial parameter solution set for eliminating time-lag deviation, and breaks through the nonlinear coupling optimization bottleneck of a traditional static model; and the state dissociation module calls a historical wear feature library to construct a performance attenuation reference surface, and tool accumulated wear and instantaneous thermal deformation disturbance are synchronously inhibited through convolution operation and projection correction of a dynamic compensation factor.

The invention relates to a traffic flow prediction method based on multi-channel space-time interaction and fine-grained graph convolution, and the method comprises the steps: carrying out the processing of traffic flow data, a time label and a self-adaptive space-time feature through a space-time feature embedding module, and carrying out the extraction of a time sequence feature and a space feature from a high-dimensional embedded feature through a coding module; the encoding module comprises a multi-channel space-time encoder; performing spatio-temporal feature fusion on the time sequence features and the spatial features by adopting a bidirectional spatio-temporal interaction fusion module; and performing feature extraction on the space-time fusion feature, the traffic flow data, the time label and the adjacent matrix by adopting a fine-grained space-time diagram convolution module, and then performing prediction by adopting a prediction module to obtain a traffic flow prediction result. According to the method, multi-scale time and space features are extracted by using a multi-channel space-time encoder, global space-time dependence is constructed by fusing attention through bidirectional space-time interaction, a fine-grained dynamic space-time relationship is captured by using a dynamic and static graph combined graph convolutional network, and the accuracy of traffic flow prediction is improved.

The invention discloses a coordinated optimization multi-step prediction method based on target correlation space-time coding-task decoding, and the method comprises the steps: constructing a space-time encoder formed by stacking a plurality of space attention enhanced time sequence convolution modules, and carrying out the complete adaptive noise set empirical mode decomposition of a target value historical sequence; constructing a parameter sharing mechanism, and training a multi-step prediction model formed by a space-time encoder and a task specific decoder by using a multi-task learning normal form; training input is preprocessed historical industrial process production data, and gradient conflicts among multiple tasks are reduced by using a gradient adjustment mechanism and a dynamic weight distribution method; and preprocessing real-time industrial process production data, and inputting the data into the trained prediction model to obtain a prediction result. According to the method, the relationship among different prediction tasks can be effectively coordinated, the overall model performance is optimized, and challenges encountered in long-term prediction are more effectively processed by dynamically adjusting weight distribution and gradient flow among multiple tasks.

The invention discloses a layer chain constraint three-dimensional human body posture estimation method based on a monocular video stream, and relates to the field of three-dimensional human body posture estimation. According to the method, a layer chain constraint three-dimensional human body posture estimation model is adopted, the model comprises a space-time coding basic block, and the space-time coding basic block comprises an SFM module, an IFM module, an EFM module, a space-time attention module and a space-time diagram convolution module; the SFM module extracts inherent space information from the attitude features from a single frame angle; the IFM module carries out frame distribution limiting operation on the attitude features from an inter-frame angle and carries out modeling on inter-frame time and space dependency relationships. According to the method, the 2D attitude is modeled from three perspectives of single frame, inter-frame and whole, the physical topological structure of the human body can be explicitly expressed, and the task precision can be better improved through mutual fusion and conversion of the global level and the local level.

The invention discloses a multi-modal content intelligent auditing and violation detection method and system, and relates to the technical field of information processing. The method comprises the following steps: carrying out audio-picture separation on a video stream, and carrying out parallel processing on audio-to-text and visual key frame extraction; a space-time encoder is constructed to record the corresponding relation of the time stamps of all the modes; constructing a multi-modal resource target dictionary, and forming an inter-entity knowledge graph by using relation categories; calculating a confidence coefficient difference index between modals by comparing, learning and training the shared semantic space; and carrying out violation judgment, triggering a sensitive characteristic threshold value for any mode, and starting multi-mode evidence cross validation. According to the method, audio and picture separation is carried out on the video stream, parallel processing is carried out on audio-to-text and visual key frame extraction, a multi-modal resource target dictionary is constructed, violation judgment is carried out according to confidence coefficient difference indexes among modals, and false information auditing efficiency and detection efficiency are improved.

The invention relates to the technical field of data analysis, in particular to a city physical examination accurate analysis method and system based on AI multi-modal data collaboration, and the method comprises the following steps: constructing a self-adaptive fractal space-time grid, and mapping a city multi-modal data source into a space-time coding vector with a fractal dimension; establishing a physical field driven cooperative resonance network, and extracting an abnormal cooperative mode exceeding a normal resonance threshold; calculating a modal entropy chain value according to the real-time metabolic rate of the urban system, and generating an optimal weight matrix through a non-equilibrium thermodynamic model; and fusing the abnormal cooperation mode and the dynamic weight matrix, and generating a space-time causal map for displaying a fault source and a propagation path by using a causal discovery algorithm. According to the method, the multi-modal data collaboration and causal analysis technology is utilized, the emergency response capability of an urban system in the case of sudden failures is improved, the stability of urban operation is guaranteed, and the space-time causal atlas provides comprehensive visual information for the failure propagation process.

The invention belongs to the technical field of power system application, and particularly relates to an intelligent power flow data filling method based on space-time coupling. The invention provides an intelligent power flow data filling method based on space-time coupling, which comprises the following steps: firstly, carrying out data preprocessing on original data through Hampel filtering, phase alignment and normalization, constructing a space-time encoder through joint coding of adaptive graph convolution (Adaptive GCN) and attention LSTM, capturing space-time coupling characteristics, and carrying out data processing on the space-time coupling characteristics; a multi-physical constraint generative adversarial network (Pry-GAN) is designed through generator physical constraints and a multi-scale discriminator to realize data filling, and it is ensured that a filling result has both numerical reasonability and physical feasibility.

The invention belongs to the technical field of three-dimensional oil reservoir seepage, and particularly relates to a coupled neural operator fusing Fourier space-time coding and multi-field attention. The neural operator realizes joint prediction and efficient simulation of key physical quantities such as pressure, oil phase saturation and water phase saturation. The problems of calculation bottleneck, weak generalization ability and the like of a traditional method in the aspects of processing high-dimensional space input, time sequence evolution and physical coupling field modeling are solved. According to the coupling neural operator, a coupling correlation mechanism between a physical field and time and space is introduced through a multi-representation structure and attention enhancement, so that the cooperation efficiency between operator tasks, the space-time information utilization rate and the expression ability of a model to complex flow field interaction behaviors are effectively improved; the method has the characteristics of fast overall reasoning, low persistence cost, fast and stable optimization process convergence and the like. Fourier high-dimensional spatial-temporal feature mapping is fused, so that the model has higher spatial-temporal change sensing ability, and the modeling stability of the time evolution process is improved.

The invention discloses a method and system for constructing a digital twinborn body based on space-time coding and a computing device, and relates to the technical field of digital twinborn, and the method comprises the steps: obtaining the original real-time data of each physical entity of each target region collected by each data collection device in real time; performing data cleaning and formatting processing on the original real-time data of the physical entity to obtain target real-time data; acquiring geographical location information and time information of the physical entity, encoding the geographical location information and the time information to generate a space-time code, and associating the target real-time data with the space-time code to obtain space-time coded real-time data of the physical entity; and constructing a data twinborn body corresponding to the physical entity based on the time-space coding real-time data so as to monitor the state of the physical entity in real time based on the data twinborn body corresponding to the physical entity. According to the invention, high-precision mapping and real-time monitoring can be carried out on the physical entity, high-precision space-time synchronization of data from different sources can be realized, and the expandability is improved.

The invention discloses a traffic flow prediction method based on feature embedding and a space-time multi-head self-attention mechanism, and relates to the technical field of intelligent traffic and flow prediction, and the method comprises the steps: obtaining historical traffic flow sequences of N sensors at the first T moments in a traffic network; taking historical traffic flow sequences of N sensors in a traffic network at the first T moments as input, and generating features including time and space through a feature embedding layer; the generated features including time and space are input into a stacked space-time coding layer, calculation is carried out through a decoupled time multi-head self-attention mechanism and a space multi-head self-attention mechanism, new time features and space features are obtained, the time features and the space features are fused, and space-time features are obtained; inputting the spatial-temporal features into an MLP module, and further extracting high-level features through multi-layer MLP; and outputting the high-level features through a regression layer to obtain a prediction result. According to the method, the problem that the traditional prediction model is difficult to process the time-space complexity at the same time is effectively solved by utilizing a decoupling type time and space multi-head self-attention mechanism, and the accuracy of traffic flow prediction is improved.

The invention provides a calcium imaging neural signal extraction method and product based on self-supervised pre-training, and relates to the technical field of artificial intelligence. Based on a pre-training data set containing a large amount of unlabeled two-photon calcium fluorescence imaging data, a pre-training decoding network and a corresponding pre-training task are self-supervised, and a pre-trained space-time coding network is trained by taking minimization of a pre-training loss function as a target; the method comprises the following steps of: firstly, establishing an encoder and decoder network, establishing fine-tuning data with accurate space-time marked neuron calcium signal positions through simulation, simultaneously performing fine-tuning on the encoder and decoder network by taking minimization of a fine-tuning loss function as a target, and checking the algorithm extraction precision based on various actual imaging data to obtain a calcium imaging signal extraction algorithm network with strong generalization performance. Therefore, a calcium signal extraction model which can be generalized in various imaging conditions and model biological brain regions can be obtained through training without manual labeling, a high-accuracy neural signal extraction result is obtained, the application difficulty that a deep learning model lacks data generalization ability is overcome, and the accuracy of a neural signal extraction result is improved. Calcium imaging neural activity analysis based on self-supervised pre-training and wide in application is achieved.

The invention discloses a sintering end point multi-step prediction method based on target correlation space-time coding-task decoding. The method comprises the following steps: constructing a space-time encoder formed by stacking a plurality of space attention enhanced time sequence convolution modules; completely adaptive noise set empirical mode decomposition is carried out on a target value historical sequence, a screened internal mode function is input into a gated neural network and then passes through a full connection layer, and a result is input into a task specific decoder constructed by adopting a time attention mechanism; historical industrial process production data is preprocessed to train a prediction model composed of a space-time encoder and a task-specific decoder; and preprocessing real-time industrial process production data, and inputting the data into the trained prediction model to obtain a prediction result. According to the method, features can be effectively extracted from multiple levels, implicit information in historical data of target variables can be deeply mined, and the prediction capability of the model on the sintering end-point state is remarkably enhanced.

The invention relates to the technical field of information, and discloses a digital right near-field real-time cancel-after-verification system based on edge computing, which comprises a space-time coding module used for generating a unique cancel-after-verification identifier according to the geographic position and time of a user, and transmitting the identifier to a dynamic weight module; a dynamic weight module which is connected with the space-time coding module, generates a right and interest distribution strategy based on a federated learning model and an optimization algorithm, and outputs the optimized strategy to a block chain fragmentation module; the block chain fragment module is connected with the dynamic weight module, dynamically divides a consensus network according to the real-time cancel-after-verification density, executes distributed verification in fragments and generates a cancel-after-verification authorization instruction; and the cancel-after-verification execution module is used for distributing cancel-after-verification tasks according to the authorization instruction and synchronously updating the multi-node inventory state. By adopting the technical scheme of gridding coordinate mapping, time window division and nonlinear hash of the space-time coding module, the technical effect that the verification identifier is unpredictable and uniquely bound is achieved.

The invention relates to the technical field of intelligent traffic, and discloses an end-to-end vehicle group collaborative decision-making method based on mixed deep reinforcement learning in a mixed traffic flow environment. According to the method, a graph structure modeling method aiming at a complex traffic scene is designed, a communication mechanism between traffic participants is realized from a bottom layer, then, a space-time encoder based on a graph convolution method realizes effective embedding of space-time information in environment observation, and the space-time information in the environment observation is effectively embedded. According to the method, the precondition is realized for a system to understand space-time coupling information in an environment, finally, an act-critic architecture and a Q learning architecture are deeply coupled, a novel reinforcement learning method suitable for a mixed action space is constructed, after model training is completed based on the method, model deployment can be realized by storing and loading model parameters, and the method has the advantages of being simple in structure, convenient to operate and high in practicability. And the deployed model can output the collaborative decision of the controlled vehicle group by inputting environment observation data.

The invention discloses a cross-modal large model construction method and system based on track spatio-temporal characteristics, and belongs to the crossing field of artificial intelligence and dynamic spatio-temporal data processing, and the method comprises the steps: carrying out the sliding sampling and spatial distribution difference judgment through the semantic dynamic segmentation of multi-scale track spatio-temporal data, and generating spatio-temporal data blocks with consistent semantics; designing a space-time encoder of a hybrid architecture, extracting track time sequence association and spatial features, and unifying dimensions; constructing a text space-time fusion mechanism, and dynamically adapting cross-modal features by means of an anchor interface and gating fusion; a staged instruction fine tuning strategy is adopted, semantic alignment of space-time and text features is optimized firstly, then model top-layer parameters are trained cooperatively, complex scene adaptation is enhanced in combination with instruction difficulty progression and hard sample mining, and space-time constraint regular terms are introduced to guarantee output rationality. According to the method, high-precision cross-modal reasoning capability is provided for scenes such as track analysis and track prediction.

The invention discloses a real-scene three-dimensional system geographic entity space-time coding method, system and device and a medium, belongs to the technical field of three-dimensional scenes, and aims to solve the technical problem of how to quickly and accurately carry out space-time positioning and coding on a geographic entity in a real-scene three-dimensional system and improve data processing efficiency and precision. According to the technical scheme, multi-level network division and unique identifier: multi-level and multi-scale grids on the surface of the earth are divided based on a GeoSOT model, the GB / T 40087-2021 earth space grid coding rule is followed, the grid level is determined by calculating the size of a bounding box of a geographic entity, coordinate binarization left-shift operation is carried out at a coordinate of a center point of the entity, and a unique identifier is obtained; generating a unique grid code containing south and north hemisphere marks and east and west hemisphere marks by combining a Morton code conversion technology and a quaternary conversion technology; bTC Beidou subdivision of time codes is carried out; a space-time embedding model; optimizing spatial data coding in combination with Morton codes and binary tree coding; and sorting the space-time codes through a heap sorting algorithm.

The invention relates to the technical field of food safety supervision, in particular to a food processing visual monitoring system and method based on order tracing, and the system comprises an intelligent collection terminal, a space-time coding server, a block chain evidence storage platform, an exception handling module and a user access terminal. The intelligent acquisition terminal comprises a multi-view camera array, a temperature sensor and an edge calculation module; the space-time coding server encrypts and binds order information with a three-dimensional space, and the block chain evidence storage platform stores video slices and associated metadata to generate a time sequence block chain record; the exception handling module automatically triggers order blocking and illegal evidence storage based on AI identification; and the user access terminal checks the video stream through the traceability code to implement dynamic privacy protection. Accurate tracing is realized through space-time encryption binding triggered by an order, edge calculation key frame extraction optimization storage and block chain evidence storage are combined, cross-system cooperative supervision is supported through dynamic privacy protection, and efficient, safe and low-cost food processing monitoring is realized.

The embodiment of the invention provides an image processing method and device based on computer vision and artificial intelligence, and the method and device achieve the privacy protection of data through innovatively designing a multi-dimensional data safety collection mechanism and through feature branch isolation and space-time coding. A hierarchical feature fusion network is constructed, and a safe and reliable target tracking system is established in combination with environmental constraints and physical laws. A dynamic weight adaptive mechanism is introduced, and the accuracy of a processing result and user privacy are ensured through multi-scale feature extraction and prediction correction. According to the method, the data security is protected, and meanwhile, the defects of the traditional technology in the aspects of data fusion, feature extraction, prediction constraint and the like are effectively overcome.

The invention discloses a multi-target radio frequency identification system based on a space-time coding metasurface. The multi-target radio frequency identification system comprises a space-time coding metasurface module, a baseband signal transmitting module, a received signal demodulation module, a signal preprocessing module and a target classification module, the space-time coding metasurface module adjusts the coding module in real time, dynamically regulates and controls electromagnetic waves, generates a plurality of detection channels, and realizes recognition scenes of various different targets; the baseband signal transmitting module generates and transmits a baseband signal, and after the baseband signal is modulated by the space-time coding metasurface module, a multi-order radio frequency signal containing the baseband signal is formed; the receiving signal demodulation module receives a radio frequency signal reflected from a target to be identified, extracts a specified harmonic wave and demodulates the radio frequency signal into a baseband signal; the signal preprocessing module transmits processed data of the demodulated signal to the target classification module; the target classification module uses a machine learning algorithm to classify and identify targets. According to the invention, simultaneous detection of multiple targets is realized, and the detection efficiency and speed are improved.

The invention discloses a multi-level electroencephalogram signal analysis method, device and equipment based on space-time coding. The method comprises the steps that multiple pieces of electroencephalogram channel data to be analyzed are acquired; dividing the electroencephalogram channel data into a parallel group or a series group according to channel information corresponding to the electroencephalogram channel data; extracting multi-scale time domain and frequency domain features from each electroencephalogram channel data of the parallel group, and constructing a first feature set according to the multi-scale time domain and frequency domain features; extracting a time sequence dependence feature from each electroencephalogram channel data of the series group, and constructing a second feature set according to the time sequence dependence feature; determining a primary feature set and a secondary feature set according to the first feature set and the second feature set; dividing the main feature set and the secondary feature set into row and column coding branches, and obtaining electroencephalogram signal feature representation according to the row and column coding branches so as to complete cognitive analysis according to the electroencephalogram signal feature representation.

The invention discloses a track user association method based on semantic perception and space-time coding, relates to the technical field of location service and user behavior analysis, and aims to solve the problems of excessive dependence of POI identifiers, limited space-time representation capability, insufficient cross-city generalization capability and the like of the existing TUL method in practical application. By introducing a pre-trained large language model to carry out POI category semantic coding, multi-frequency sine space-time coding and a double-flow transfer learning mechanism, the method can effectively improve the prediction precision and the model generalization ability.

The invention discloses a virtual-real sensing augmented driving equipment physical coupling field intelligent calculation method, and relates to the field of equipment physical performance simulation. According to the method, the real position and real sensing data of each real node are coded to obtain real sensing position features, based on the real sensing position features of each real node, a real domain space-time coding network is adopted to create virtual sensing position features of a virtual node, and then based on virtual and real sensing features of a virtual and real node, a real domain space-time coding network is adopted to create virtual sensing position features of a virtual node. And virtual-real sensing data of virtual-real nodes are created by adopting a virtual-real coding network, expansion of the sensing data in spatial distribution is realized, a temperature field neural network model and a stress field neural network model are constructed based on a temperature and stress mechanism equation, a multi-parameter nesting mechanism under thermal-mechanical coupling is established, and a multi-parameter nesting model is established. The training process of the stress field is standardized by using a temperature field calculation result and a strain-stress conversion relation, and intelligent and efficient calculation of the equipment temperature-stress coupling field is realized.

The embodiment of the invention provides a terminal track generation method and device, a medium and a product, and relates to the technical field of mobile communication. The method comprises the following steps: determining mapping data between signaling and a position of a terminal; acquiring a time feature and a position feature in the mapping data by using a space-time encoder; and inputting the time feature and the position feature into a preset attention mechanism model, and determining the actual position of the terminal. According to the scheme provided by the invention, the dependence on a single base station signal is reduced through deep association of the terminal signaling and the position mapping data, and the dependence on the positioning precision of the base station is broken through; a space-time encoder and an attention mechanism model are utilized to realize multi-modal mobile modeling; based on signaling and an attention mechanism, semantic and base station interaction relation modeling is enhanced, and the precision of track production or track recovery is improved.

The invention provides a communication and perception joint optimization method based on a space-time coding metasurface. The method comprises the following steps: step 1, modulating and sending symbols by adopting a quadrature amplitude modulation scheme; 2, performing real-time regulation and control on the programmable metasurface; step 3, the integrated multi-antenna receiver separates received signals with different frequencies by using orthogonality between harmonic waves, wherein different harmonic waves respectively correspond to a communication signal and a target echo signal; step 4, constructing a joint optimization problem model of communication and perception performance indexes; and 5, designing an improved algorithm based on a non-dominated sorting genetic algorithm to jointly optimize the communication SNR and the sensing CRB so as to improve the communication performance by using sensing assistance. According to the invention, the communication information demodulated by the zero-order harmonic can be utilized to realize the sensing of the other-order harmonic, and the sensing information of the other-order harmonic can be utilized to enhance the communication performance.

The invention provides a shale oil reservoir dessert intelligent identification method and system, relates to the field of shale oil reservoirs, and aims to obtain original logging data and extract curve deep space-time coding features by fusing a hybrid model of a convolutional neural network and a recurrent neural network. Then, multi-source logging features are dynamically fused through a cross-curve feature fusion algorithm, and spatial compression is carried out through feature principal component analysis; thirdly, implicit features strongly related to the geological signs of the sweet spots are screened through semantic level selection based on feature principal components, sweet spot probability prediction is achieved based on a lightweight classification model, an intelligent recognition closed loop without manual intervention is formed, and the dependence of a traditional method on expert experience is broken through; in this way, non-linear correlation in the logging curve is automatically mined through deep learning, and the interpretability and generalization ability of complex reservoir dessert recognition are improved.

The invention discloses a heterogeneous skeleton-based open vocabulary action recognition method and system. The method comprises the following steps of: constructing a heterogeneous open vocabulary skeleton data set; unifying the skeleton representation in the heterogeneous open vocabulary skeleton data set, and defining a unified space structure by establishing the maximum joint number and member number; a skeleton motion encoder model based on a Transform architecture is constructed, the skeleton motion encoder model comprises feature embedding, space-time encoding and a projection layer used for cross-modal alignment, and global time features, global space features and global visual features output by space-time encoding are subjected to cross-modal alignment through three parallel projection networks of the projection layer. Mapping to a semantic space aligned with text embedding from the pre-trained language model; training loss is constructed based on a multi-granularity motion-text alignment strategy including global instance alignment, stream specific alignment and fine granularity alignment, and a motion encoder model is trained. A wide range of experiments on a popular benchmark with heterogeneous skeleton data prove the effectiveness and generalization ability of the proposed method.