17 results about "Spatiotemporal pattern" patented technology

This invention provides a skeleton gait emotion recognition method based on identity and gender suppression, belonging to the field of emotion recognition technology. This invention effectively solves the problem of interference from individual differences such as identity and gender on emotion recognition through the following innovative designs: First, a hierarchical cross-scale attention mechanism is used to adaptively fuse multi-scale features extracted from spatiotemporal graph convolutional blocks. Through cross-scale attention complementarity, time-channel alignment, and gating weighting, the spatiotemporal pattern representation related to emotion is strengthened. Second, an emotion-conditional adversarial training strategy is innovatively introduced: while optimizing the identity / gender classification branch through a three-attribute subset EGI-Gait, adversarial constraints are implemented by concatenating emotion label encoding with shared features, forcing identity and gender features within the same emotion category to be indistinguishable. This method improves the model's cross-subject generalization ability and significantly reduces the interference of irrelevant individual features on emotion discrimination while maintaining a lightweight inference structure.

This invention relates to the field of migratory bird distribution detection technology, specifically disclosing a method for detecting migratory bird distribution in wetland protected areas based on image sensors, including the following steps: Step S1: Divide the protected area into several sub-regions, set up image acquisition points in each sub-region, acquire migratory bird images in real time, identify the instantaneous number of migratory birds in the images, and simultaneously record the acquisition timestamp and acquisition coordinates; Step S2: Construct a spatiotemporal counting matrix based on the number of migratory birds at different times at each acquisition point, with acquisition coordinates as rows and time as columns, and obtain several basic spatiotemporal patterns and the corresponding linear weights of each pattern through matrix decomposition; Step S3: Based on the change of the linear weights of each basic spatiotemporal pattern over time, identify the migration events of migratory bird flocks between sub-regions, combine the migration events with the spatial distribution preference vectors of each pattern, determine the migratory bird flock gathering areas and migration channels, and generate the final distribution detection results.

The invention discloses a space-time pattern prediction method based on forest carbon sink climate change. The method comprises the following steps: collecting vegetation data, soil data, meteorological data, topographic data and special interference factor data of a target forest area; preprocessing the collected multi-source data, and performing space-time matching on the preprocessed data; quantifying a climate index of the target forest region; taking a Biome-BGC ecological process model as a core, and integrating a graph neural network to construct a carbon sink-climate index-terrain coupling model; setting various climatic change scenes, and predicting a carbon sink spatial and temporal distribution pattern under different scenes based on a carbon sink-climatic index-terrain coupling model. The space-time pattern prediction method based on the forest carbon sink climate change is adopted, and the prediction precision is improved.

An apparatus for diagnosing an autism spectrum disorder (ASD) based on a graph neural network includes a preprocessor configured to acquire brain image data, designate a region of interest (ROI) of the brain, and generate preprocessed data for neural network input, a spatial feature extractor configured to extract spatial features from the preprocessed data, a temporal feature extractor configured to analyze changes in brain activity over time and extract attention-based temporal features, a spatial and temporal convergence feature unit configured to analyze spatiotemporal correlations by combining the spatial features and the temporal features, a graph generator configured to convert connectivity between regions of interest into a graph structure and implement the same as nodes and edges, and a graph classifier configured to analyze a spatiotemporal pattern of the connectivity through the graph structure and classify whether or not there is an ASD.

This invention discloses a method for identifying the physiological state of adult product users based on multi-sensor data acquisition, relating to the fields of intelligent health monitoring and biosignal processing technology. This invention uses non-negative matrix factorization (NMF) technology to decompose multi-channel non-steady-state electromyographic signals collected from the pelvic floor and core muscle groups into multiple muscle coercive elements and their temporal activation coefficients. Each coercive element represents a fixed pattern of muscle cooperating under neural drive, while the activation coefficient reflects the change of this pattern over time. Sexual arousal, as a specific neurophysiological process, induces the activation of coercive elements with specific spatiotemporal patterns. By tracking the activation and evolution of specific coercive patterns, the neural control fingerprint reflecting sexual arousal is extracted from the original signal contaminated by motion noise. This effectively distinguishes between ordinary muscle contractions caused by device use and physical activity and autonomous neuromuscular activities related to sexual responses, overcoming the defects of susceptibility to interference and insufficient specificity.

The main repeating patterns in the recorded spatiotemporal data are identified, extracted, and / or refined by: converting the data into phase data; determining the global cycle length for each cycle with discrete time points; determining the start time in the original recording; calculating the optimal global phase order over 2π and the entire cycle length using the time points and start time; assigning singular values ​​to each spatial location; and generating a visual representation of the pattern. Variations of the method can be used to identify features of the pattern, such as spatial movement, focal position, drift, and stability.

The invention relates to the technical field of big data processing, in particular to a geological disaster emergency response method and system based on big data, and the method comprises the following steps: building a monitoring data set through collecting and synchronizing a multi-source monitoring sequence, carrying out the abnormality elimination to obtain a change rate, calculating a ground stress gradient and an elevation difference through space mapping, and analyzing the continuity; the method comprises the following steps: calculating an energy change rate by combining rainfall and underground water level fluctuation, extracting associated features, tracking phase difference and tension potential direction change to form phase response parameters, fusing a multi-dimensional feature identification space-time mode, clustering to complete risk level assessment and response priority output, and generating a geological disaster emergency response result. Dynamic continuity is enhanced through time synchronization and change rate, regional difference is highlighted through spatial mapping, unstable factors are described through energy change correlation analysis, an evolution trend is shown through phase difference tracking, a space-time mode is formed through multi-dimensional fusion, and risk identification and response precision is improved.

The invention provides a fine simulation method and system for total primary productivity of tree species in the technical field of ecological remote sensing and carbon cycle simulation. The method comprises the following steps: S1, acquiring leaf hyperspectral reflectivity data, photosynthetic physiological parameter data and canopy hyperspectral remote sensing image data of different tree species in a target area; s2, extracting characteristic wave bands based on the maximum carboxylation rate in the hyperspectral reflectivity data of the leaves and the photosynthetic physiological parameter data, constructing a characteristic wave band combination, and constructing a Vcmax hyperspectral inversion model based on the characteristic wave band combination; step S3, tree species classification is carried out by using the canopy hyperspectral remote sensing image data, and a dynamic Vcmax distribution diagram is generated in combination with the Vcmax hyperspectral inversion model; and S4, replacing static Vcmax parameters of the ecological process model based on the dynamic Vcmax distribution diagram, and inputting driving data into the ecological process model to simulate a GPP space-time pattern. The method has the advantages that the accuracy and practicability of tree species total primary productivity simulation are greatly improved.

This invention relates to a method and system for locating faults in optical fiber networks, comprising: collecting raw monitoring data of optical transmission through a multi-vendor network management system; constructing a partial optical fiber topology map including user-end equipment nodes, optical line terminal port nodes, and splitter nodes within a preset monitoring period; generating a global optical fiber topology map based on the uplink port identification information of the user-end equipment after conflict resolution and integration processing; converting the topology map into a set of optical transmission triples, and identifying directly connected and indirectly connected port groups based on the triple set; constructing a target fault propagation loop with the port to be located as the starting and ending point based on the two types of port groups; extracting the optical performance feature dataset of the ports within the loop, and determining whether the loop represents a real physical layer fault by combining topology features and using spatiotemporal pattern matching and optical performance feature offset analysis. This method achieves accurate fault location and tracing in a multi-vendor environment.

The invention discloses a traditional dwelling decoration value factor construction and classification method and system, and the method comprises the steps: obtaining a decoration image and point cloud data through multi-source data collection, and generating a decoration element data set after semantic segmentation and standardization processing; multi-level features are extracted to construct a function dependency graph, and a large language model is used for filtering and reordering optimization; constructing a Text2SQL expansion framework based on the optimized function dependency graph, and extracting a multi-dimensional value factor of the decoration element; building a multi-granularity spatio-temporal flow mode through spatio-temporal mode enumeration and optimization, and forming a decoration value factor system; and importing the system into an atlas database, establishing a bidirectional mapping relationship, and realizing multi-dimensional classification and associated retrieval of the decoration elements. The problems that traditional decoration value evaluation is high in subjectivity and single in classification system are solved, and systematic and standardized construction and intelligent retrieval of the decoration value are achieved.

This invention discloses a data prefetching method, product, device, and medium based on multiple memory access patterns, relating to the field of computer architecture technology. The method includes: monitoring for L2 cache memory access miss events in the processor; if the miss status register hits the event, splitting the miss address to obtain the base address of the spatial region and the offset address within the region. When the program counter pattern training table hits the base address, stride and spatiotemporal pattern training is performed on the offset address to obtain a fixed stride, space / time related access pattern. Valid memory access patterns are filtered by confidence level and stored in the corresponding LLC pattern table, with an index recorded in the prefetch pattern record table. Subsequently, based on the memory access miss event and index information, the target pattern is read from the corresponding pattern table to calculate the prefetch address, generate a request, and execute the prefetch. Through this scheme, it is possible to adapt to multiple scenarios and accurately identify memory access patterns for more precise data prefetching.

This invention discloses a method and related apparatus for identifying the spatiotemporal evolution characteristics of watershed water resources. The method includes: acquiring an overview of watershed zoning and water resource quantity information; using a point conversion tool to convert the surface features of the zoning into point features at their geometric centers; using a spatiotemporal cube construction tool to create a spatiotemporal cube of water resources quantity in the watershed; using the Mann-Kendall statistical method to determine the trend of water resource quantity change in each zoning; using Sen's Slope Estimator to quantify the rate of change of water resources quantity in each zoning; and finally, visualizing the evolution characteristics. This invention, by constructing a spatiotemporal cube and embedding the time dimension into a spatial analysis framework, achieves joint analysis and dynamic presentation of the spatiotemporal evolution of water resources. It can quickly and accurately identify the evolution trend of watershed water resource patterns and has advantages such as comprehensive spatiotemporal analysis, high visualization, and dynamic expression.

The invention discloses a power transmission line active protection method based on bird-related risk dynamic assessment, and relates to the technical field of power transmission safety protection, and the method comprises the following steps: obtaining multi-source sensing data of the periphery of a power transmission target line; recognizing a continuous behavior state of the bird based on the multi-source sensing data, and generating a state transition chain; generating a dynamic risk representation of a bird-related risk evolution trend by using a state transition chain in combination with a field environment condition and a historical space-time law; according to the dynamic risk characterization, reverse planning is carried out along a risk evolution time axis, and a serialized active protection strategy containing multi-stage response actions is generated; and cooperatively scheduling a multi-position protection device according to the active protection strategy, and constructing a protection interception surface dynamically changing along with the risk direction and strength. According to the method, the fundamental transformation from static threshold judgment to dynamic behavior prediction and from isolated action triggering to systematic collaborative protection is realized, and the protection accuracy, initiative and systematic flexibility are remarkably improved.

The application discloses a new energy grid-connected inverter cluster online parameter identification method and related device, including the following steps: dynamic mode decomposition is carried out on the three-phase voltage and total current signals of the point of common coupling, and a dominant spatiotemporal pattern set representing the overall dynamic behavior of the cluster is extracted; a parameterized virtual observer model is constructed for each inverter in the cluster, and a unified input excitation is generated based on the three-phase voltage signals of the point of common coupling and applied to all virtual observers; a collaborative loss function is set; an online adaptive optimization algorithm is used to minimize the collaborative loss function, and the multi-dimensional parameter vectors of all virtual observers are updated in parallel and in real time until convergence, and the identification result is output; the method and related device can be operated online and in real time, and can simultaneously identify the multi-dimensional key parameters of the new energy grid-connected inverter cluster.

This invention discloses a method for assessing the affected population based on dynamically corrected disaster-reported population data. The method includes the following steps: Step 1, collecting basic data of the study area; Step 2, initializing the population density field; Step 3, constructing a spatial adjacency graph; Step 4, determining the dynamic flood inundation range; Step 5, comparing the reported data to correct population errors. This method, even with limited reported data, dynamically optimizes local population density estimation by combining the spatiotemporal patterns of flood evolution, while ensuring total population conservation, non-negativity, and spatial smoothness. It achieves dynamic updates of population density at the grid scale within administrative regions, thereby obtaining a more accurate and realistic spatial distribution of the population and the number of affected people. This provides a basis for decision support such as population relocation and precise rescue, demonstrating strong practicality.

The invention provides a global sea surface temperature prediction method and system based on knowledge and data coupling. The method comprises the following steps: firstly, constructing an OKG (Ocean Knowledge Graph) containing five entities including ocean currents, climate zones, monsoon systems, regions and special sea areas, and embedding and learning the structure and semantic information of the OKG through the knowledge graph; secondly, in the time sequence feature coding step, time sequence features of multi-region SST observation data are extracted, and fine-grained alignment of ocean knowledge and observation data is achieved through a cross-modal attention mechanism; and then, inputting the fused multi-modal representation into a pre-trained large language model so as to carry out space-time mode modeling and temperature trend prediction. And finally, a sea surface temperature prediction result at multiple moments in the future is generated by using a Transform decoder. According to the method, numerical observation and domain knowledge can be fused, precision, interpretability and global adaptability are considered, and high-reliability sea surface temperature intelligent prediction is achieved.