15 results about "Mixture modeling" patented technology

The invention relates to the field of computer vision, and discloses a defect detection method, system, medium and equipment based on adaptive dual-domain filtering and Gaussian mixture prior constraint, and the method comprises the steps: carrying out the multi-scale feature extraction of an ultrasonic C-scan image through a Vision Transform network after the ultrasonic C-scan image is preprocessed; respectively inputting the shallow fusion features and the deep fusion features into a frequency-space double-domain adaptive feature filtering module for filtering, inputting the filtered features into a Gaussian mixture modeling module Ada-GMM, and modeling normal feature distribution; carrying out Ada-GMM-Guided decoding, and carrying out interactive fusion on the corresponding deep semantic features and shallow texture features by adopting a deep and shallow multi-scale feature interaction mechanism; performing optimization by adopting cosine reconstruction loss, filtering consistency, entropy regularization loss and distribution alignment loss, and adaptively learning normal distribution characteristics according to an optimization process to obtain a model weight; and reasoning the input ultrasonic C-scan image by using the trained network weight to realize anomaly detection and positioning, and outputting an interpretable anomaly thermodynamic diagram.

The present invention belongs to the field of data processing technology and relates to a QAR data preprocessing method and system based on simulator backdrive. The method aims to address the problems of poor data quality, insufficient parameter coverage, and weak dynamic adaptability encountered in traditional methods. The method comprises: collecting raw QAR data, calculating initial credibility, and screening key parameters; utilizing key parameters to backdrive a full-motion simulator, combining deep reinforcement learning to optimize a PID controller and Monte Carlo simulation to obtain response data; constructing numerical, temporal, and physical three-dimensional difference features based on the raw QAR data and response data, and determining rationality through Gaussian mixture modeling and Bayesian decision making; fusing data based on rational functions and information entropy dynamic weights, optimizing iterative parameters with multiple objectives, and outputting verified high-precision QAR data that meets aviation standards. The present invention obtains full calibration data through a QAR key parameter backdrive simulator, enabling multi-dimensional and precise identification of data anomalies.

The invention discloses a human body posture estimation result generation method and device based on a generative model, relates to the technical field of image processing, effectively recovers a complete skeleton structure, realizes accurate recognition of human body postures in high-shielding and high-dynamic change scenes such as an electric power production field, and improves the adaptability to shielding scenes. The method comprises the following steps: acquiring a to-be-identified video stream, and performing initial attitude estimation on each frame of video image in the to-be-identified video stream by using a pre-trained attitude detection network model to obtain an initial skeleton point sequence; fitting the initial skeleton point sequence by adopting a Gaussian mixture modeling method to obtain a defect skeleton point sequence; inputting the defective skeleton point sequence into a pre-trained inverse diffusion network model, and optimizing the defective skeleton point sequence in combination with the adjacent frame skeleton point coding information of each frame of video image in the inverse diffusion network model to obtain a complete skeleton point sequence; and generating and outputting a human body posture estimation result based on the complete skeleton point sequence.

The invention belongs to the technical field of data processing, relates to a QAR data preprocessing method and system based on analog machine reverse drive, and aims at solving the problems that a traditional method is poor in data quality, insufficient in parameter coverage and weak in dynamic adaptability. The method comprises the steps of collecting original QAR data, calculating initial credibility and screening key parameters; the key parameters are used for reversely driving the full-motion simulator, and response data are obtained by combining a deep reinforcement learning optimization PID controller and Monte Carlo simulation; on the basis of original QAR data and response data, numerical value, time sequence and physical three-dimensional difference characteristics are constructed, and reasonability is judged by means of Gaussian mixture modeling and Bayesian decision; and based on the rational function and information entropy dynamic weight fusion data and multi-objective optimization iteration parameters, high-precision QAR data conforming to the aviation standard is output through verification. According to the invention, full-amount calibration data is obtained through the QAR key parameter reverse drive simulator, and multi-dimensional accurate identification of data exception is realized.

The invention provides double-granularity alignment efficient partial correlation video retrieval based on implicit fragment modeling and semantic decomposition, and aims to solve the problems of information redundancy and low efficiency in an existing video modeling method, the problem of granularity mismatching between sentence representation and video frame features and the problem that alignment between a text and a video is not refined enough. According to the method, the expression and modeling of video data are optimized by introducing a structure combining a Gaussian mixture model and Transform, and multi-scale local details with short time span are adaptively integrated by introducing a window attention mechanism and a cross attention mechanism, so that finer features are obtained, and cross-modal similarity score calculation of texts and videos is facilitated. The method comprises the following steps: data preprocessing and frame segmentation: preprocessing and segmenting an input video into frames, and preparing for feature extraction after each frame of image is subjected to standardization processing; implicit modeling is carried out on the fragment-level features, uniformly sampled frame-level visual features are input into a Gaussian mixture modeling module, adjacent frame focusing modeling is carried out through a multi-scale Gaussian attention mechanism, and fragment-level representation of different receptive fields is implicitly formed. Enhancing local details of the frame-level features, performing convolution operation on each frame by adopting windows with different scales, and calculating a feature relationship in each local window; local features of all scales are fused through a cross attention mechanism, semantic expression of frame features is enhanced, and it is ensured that each video frame can capture important local details.

A method for lane-level road congestion identification and forecasting includes identifying lane-level road congestion using Gaussian mixture modeling, predicting lane-level road congestion using a 2D Markov chain, and identifying routes and route changes for a host vehicle and applying the route and route changes to improve a host vehicle estimated time of arrival (ETA) at a predetermined finish location such that the ETA is shorter than a predetermined threshold.

This invention relates to the field of wind power equipment fault diagnosis technology, and provides a method for extracting fault features from wind turbine gearboxes based on Gaussian mixture modeling. The method includes: modeling a mathematical model of the wind turbine gearbox observation signal by superimposing impact fault feature vectors and multi-source noise vectors, wherein the fault feature vector is the product of a redundant dictionary D and a sparse coefficient vector; modeling the multi-source noise vector as a Gaussian mixture distribution; constructing an objective function within a Bayesian framework to solve for the sparse coefficient vector in the mathematical model using maximum a posteriori probability estimation; simplifying the objective function; and using the EM algorithm and ADMM algorithm in a joint alternating iterative solution to obtain the optimized sparse coefficient vector; and reconstructing the impact fault feature vector in the wind turbine gearbox observation signal. This method improves the accuracy and robustness of extracting fault features from the observation signal of offshore wind turbine gearboxes.

Methods for identifying disease subgroups are described. The methods may comprise, for example, receiving subject data for a plurality of subjects diagnosed with the disease; creating a plurality of candidate best fit latent class or mixture models by: i) providing an estimate of a number of subgroups; ii) generating a set of models, each model of the set comprising the same estimate of the number of subgroups; iii) selecting a candidate best fit model from the set; and iv) repeating (i) - (iii) at least once using a different estimate of the number of subgroups to obtain a plurality of candidate best fit models; selecting a best fit model from the plurality of candidate best fit models based on a fit statistic; and applying the best fit model to the subject data to identify a number of subgroups for the disease and an associated genomic profile for each subgroup.

The present application relates to the field of intelligent detection technology, and discloses an intelligent detection device and a detection method for corrosion-resistant aluminum alloy parts. The method includes: extracting features from an image sequence of the corrosion-resistant aluminum alloy cutting process to obtain a multi-dimensional feature map; performing three-dimensional space mapping and morphological feature analysis on the multi-dimensional feature map to obtain spatio-temporal feature data of the cutting trajectory; inputting the spatio-temporal feature data of the cutting trajectory into a hierarchical cutting parameter prediction model for hierarchical feature parameter prediction to obtain a cutting parameter prediction result; performing Gaussian mixture modeling and kernel density analysis on the cutting parameter prediction result to obtain probability density distribution data of the cutting trajectory, and optimizing the variable observation domain of the Gaussian process model to obtain an optimized sequence of cutting parameters; inputting the optimized sequence of cutting parameters into the Kalman filter-Hungarian algorithm for state tracking to obtain a control instruction for the cutting parameters, thereby realizing real-time and precise control of the cutting parameters and significantly improving the cutting processing quality.

The invention discloses a method for monitoring and managing illegal parking in a non-parking area in a large railway station and an airport, which relates to the technical field of image recognition and comprises the following steps of: acquiring image frame data based on a camera deployment scheme, performing standardization processing and constructing a camera deployment configuration list; then, the vehicle size is extracted according to manually labeled image data, a width-height pair set is constructed, and an anchor box parameter set is generated through a clustering algorithm and Gaussian mixture modeling; and inputting the anchor box parameter set into a single-stage multi-frame detection network, extracting candidate detection frames and vehicle confidence coefficients, and obtaining comprehensive confidence coefficients by combining frame difference analysis and a mask coverage ratio. And establishing a static cluster according to the first frame candidate detection frame, judging whether the static cluster is an illegal parking cluster through intersection-to-union ratio and continuous frame statistics, finally extracting coordinates of a central point of the illegal parking cluster, constructing a thermodynamic diagram data structure, and performing offline management. According to the invention, the recognition accuracy and stability of the illegal parking vehicles in the non-parking area are good.

The invention discloses a water depth estimation scene-oriented key frame extraction method, device and system. The method comprises the steps of S1, acquiring historical flood event scene video data; s2, performing foreground extraction on the video data through Gaussian mixture modeling and Otsu adaptive threshold processing to generate a foreground mask sequence; s3, carrying out image perceptual hash coding based on the grayscale image, dividing video clips according to inter-frame structural similarity, and adaptively estimating a frame sampling interval based on a hash difference value of adjacent frames; s4, performing inter-frame difference on the foreground mask sequence, and selecting a frame with significant motion mutation as a candidate key frame based on a statistical threshold of global motion amplitude; and S5, according to the candidate key frames, carrying out water body and texture feature combined screening. By adopting the technical scheme of the invention, the recognition efficiency and the analysis precision of the typical scene in the flood short video are improved.

This application discloses a method, apparatus, and device for pruning deep neural networks based on Gaussian mixture modeling, relating to the field of deep neural network pruning technology. It addresses the problem that single pruning strategies in existing technologies can easily lead to model accuracy collapse or convergence difficulties. The solution includes: obtaining a pre-trained deep neural network model; performing forward propagation on the deep neural network model based on sample data to obtain the activation feature values ​​of each neuron in each network layer of the deep neural network model; dividing each neuron into effective neurons and candidate neurons based on the activation feature values; for candidate neurons, using a Gaussian mixture model to fit the distribution of the activation feature values ​​of the candidate neurons to distinguish between signal distributions representing effective information and noise distributions representing redundant information; and pruning and removing candidate neurons belonging to the noise distribution based on the fitting results.

The present invention relates to the field of petrochemical material physical property calculation, and specifically to a physical information constrained petrochemical material mixed modeling physical property calculation system and method, comprising: a data processing module to realize the standardized cleaning and fusion of multi-source heterogeneous data; a multi-scale feature module to realize material property characterization through quantum-mesoscopic-macroscopic cross-scale modeling; a physical constraint module to embed thermodynamic constitutive equations and phase equilibrium criteria to ensure model self-consistency; a component interaction module to construct a non-ideal mixing effect prediction model based on deep potential energy field theory and establish a component interaction knowledge graph; an uncertainty quantification module to evaluate the prediction confidence interval using a Bayesian deep learning method, systematically integrating physical mechanisms and data-driven methods to achieve a 15% to 20% reduction in physical property prediction errors, and at the same time providing a visual analysis tool for molecular structure-physical property association to support process optimization and new product development decisions, meeting the petrochemical industry's core demand for high-precision and explainable physical property predictions.

The invention discloses an image signal preprocessing analysis method which is used for detecting surface defects of a workpiece product so as to improve the product quality and carry out quality control. A moving target in a complex environment is detected through a Gaussian mixture modeling method, and irrelevant information or noise in an image is eliminated through multiple image preprocessing operations, so that the image is clearer and more real, the output of a high-quality image is ensured, and necessary conditions are provided for the extraction of feature information in the next step. According to the invention, the accuracy and reliability of workpiece detection are improved.

The invention discloses an image generation method and system based on Gaussian mixture modeling, and belongs to the technical field of image generation, and the method comprises the steps: obtaining a plurality of image samples; encoding each image sample to obtain an implicit feature of each image sample after encoding; clustering the coded hidden features of all the image samples to obtain a plurality of feature clusters; generating Gaussian mixture distribution of all feature clusters; using Gaussian mixture distribution to randomly generate a new hidden feature vector; and decoding the generated new hidden feature vector to obtain an amplified sample. The problem that the current image generation quality is poor is solved, and high-quality generation of the image amplification sample is realized.