9results about How to "Effective modeling" patented technology

The invention discloses a real-time infrared small target detection method based on a linear global scanning network. In order to solve the problems that an existing model is high in calculation complexity and poor in real-time performance, a U-Net-like lightweight architecture giving consideration to both detection precision and reasoning speed is constructed. In the encoding stage, shallow local textures are extracted through Stem and ResBlock, a linear global scanning (LGS) module is introduced into a deep layer, and the core of the LGS module captures anisotropic long-distance semantic dependency with linear complexity by utilizing space scanning GRU. The coding end multiplexes the GRU by using a linear context aggregator (LCA) and combines a channel reweighting enhancement feature. In the decoding stage, jump connection semantics are aligned through PlainBlock, and upsampling features are fused through a StandFusion module and are refined through cascade convolution. And finally, an Inception module is combined with depth supervision to generate a high-precision prediction result. The method is low in calculation overhead, high in detection precision and suitable for high-frame-rate real-time infrared monitoring scenes.

The invention belongs to the technical field of human body posture recognition, and relates to a human body posture estimation method using diffusion model reconstruction, which comprises the following steps: acquiring data sets, integrating the data sets, uniformly converting the data sets to an SMPL coordinate system to form a fusion input vector, and constructing local input vectors of a left ear and a right ear; embedding the local input vectors of the left ear and the right ear into a feature space of a specific dimension, and stacking to obtain embedded feature vectors of the left ear and the right ear; the embedded feature vectors of the left ear and the right ear are sequentially input into a time modeling module and a space modeling module to obtain output after interaction of two channels, then the output of the two channels is spliced and mapped into a unified space-time fusion feature, and a feature sequence is output after all time steps are stacked; and inputting the feature sequence into a de-noising diffusion probability model by adopting the de-noising diffusion probability model, and obtaining an output model after forward diffusion and backward diffusion. The frame can complete space-time modeling and reconstruction of human body postures only by depending on the two IMU sensors worn on the two ears.

The invention relates to the technical field of photovoltaic power generation prediction, in particular to a photovoltaic power generation prediction method based on global confrontation and local contrast transfer learning, and the method comprises the steps: respectively developing ST-Net models in a source domain and a target domain based on distributed photovoltaic station data, so as to fully model the spatial features and time features of the photovoltaic power generation power in a modeling region; designing a global adversarial discrimination mechanism, learning domain invariant feature representation through a minimum-maximum game, effectively relieving a domain offset problem, finally designing a local contrast learning strategy, and enhancing the capturing ability of the model to local key features by optimizing a sample similarity relationship in a feature space, so as to obtain a local key feature model. Therefore, the problem of local information loss possibly caused by global adversarial training is solved. Experimental results show that the algorithm provided by the invention can significantly improve the photovoltaic power generation prediction precision of a newly-built power station under the condition of data scarcity.

This application provides a method for extracting voiceprint features from speech segments with specific content. The method includes: obtaining an acoustic spectrum feature segment through preprocessing; constructing a time-delay neural network module; constructing a residual time-delay neural network module based on the time-delay neural network module, a weighted activation mechanism, and a residual structure; constructing a residual attention time-delay neural network module based on the time-delay neural network module, the residual time-delay neural network module, and an attention pooling mechanism; and inputting the acoustic spectrum feature segment into the residual attention time-delay neural network module to obtain the voiceprint features of the speech segment with specific content. The voiceprint feature extraction method provided here extracts deep-level information from features at multiple scales and, combined with residual networks, weighted activation, and attention pooling mechanisms, can effectively extract voiceprint features from speech segments with specific content.

The application relates to the technical field of electroencephalogram signals, in particular to a multi-instance depression recognition method based on high-order microstate characteristics, which comprises the following steps: acquiring an electroencephalogram signal; extracting a microstate time sequence of the electroencephalogram by using an electroencephalogram microstate template; mining a frequent microstate transition mode by using the microstate time sequence, and constructing a global transition rule library; calculating the matching degree of the microstate time sequence of a sliding window and rules in the global rule library, and realizing rule activation recoding of the microstate time sequence; encoding a rule index sequence by using a multi-instance learning model to obtain a microstate high-order transition feature; and inputting the microstate high-order transition feature and a microstate time sequence feature after splicing into a classification network to output a depression judgment result. The application solves the problems of insufficient feature expression and unreliable diagnosis model of the existing method in early screening of depression.

This invention provides a method and system for agricultural product quality assessment and environmental decision-making based on a causal model, relating to the fields of agricultural product preservation and supply chain environmental control. The method includes: acquiring environmental data of the environment in which the agricultural product is located; determining respiratory characteristic parameters characterizing the respiration state of the agricultural product based on changes in oxygen and carbon dioxide concentrations; pre-constructing a causal model based on agricultural product category information; performing causal inference using the respiratory characteristic parameters and environmental data as inputs to obtain the remaining shelf life under current environmental conditions; determining target environmental conditions based on a preset preservation strategy, and inputting the target environmental conditions as intervention variables into the causal model for inference to obtain the expected remaining shelf life; generating a quality damage index based on the remaining shelf life and the expected remaining shelf life, and generating environmental control strategy information. This invention, by introducing a causal model and intervention inference mechanism, achieves quantitative assessment of changes in agricultural product quality and generates environmental control decisions.

The invention discloses a digital holographic microscopic imaging coherent noise suppression network model and method based on deep learning. The network model adopts a double-branch encoder-single decoder structure, and a double-branch encoder extracts local details and global noise features of an image in parallel; the innovative double-branch intensive attention fusion module fuses multi-scale features through an adaptive weighting and enhancement mechanism; and the decoder integrates information through jump connection and reconstructs a clear image. The training method generates a high-fidelity data set based on a physical simulation system. According to the method, speckle noise and parasitic fringes can be efficiently suppressed, the phase details of the object are effectively reserved while the peak signal-to-noise ratio and the structural similarity index are remarkably improved, and the method has excellent generalization ability and calculation efficiency and is suitable for real-time high-quality imaging of a digital holographic microscopy system.

The application provides a context semantic modulation method for video subtitle generation, and belongs to the field of video subtitle generation. In order to solve the problem that in the prior art, the semantic alignment effect is poor, and key information is easily ignored. The application integrates the context information of each output token, further learns the dynamic attention weight between tokens, so as to realize the fine-grained semantic alignment between the to-be-generated token and the generated token. The method can not only more effectively model the semantic correlation, but also adaptively adjust the semantic weight of the key information, so as to enhance the representation of important features and avoid the key information being ignored.

The invention discloses an application, a system and a method of an improved brake disc model in wind power plant yaw simulation, the improved brake disc model is embedded into a dynamics simulation loop of a wind generating set to accurately analyze non-axisymmetric aerodynamic load distribution of a wind turbine rotor under a yaw working condition, and a wake flow rotation effect of the wind turbine rotor is further represented. The system comprises a data input module, an improved brake disc model module, a yaw execution and flow field coupling simulation module and a result output module which are in communication connection in sequence. The method is implemented by integrating an improved brake disc model into a CFD simulation process based on CFD simulation of a yaw wind turbine of an OpenFOAM platform. According to the scheme, on the premise that the simulation calculation efficiency is guaranteed, the prediction precision of the output power of the wind turbine and the wake flow field evolution process in yaw control strategy simulation is remarkably improved, and therefore efficient and reliable numerical simulation support is provided for intelligent cooperative control optimization of the wind power plant level.