15 results about "Adaptive kernel" patented technology

The invention discloses a self-adaptive kernel selection fusion Transform network and method for hyperspectral image classification, and belongs to the technical field of hyperspectral image classification, and the network comprises three innovative modules: (1) a self-adaptive convolution module (ACBlock) improves the calculation efficiency while guaranteeing the feature extraction effect through a cascade structure of convolution and separable convolution; (2) a kernel selection fusion attention module (KSFA) adopts multi-scale convolution and dynamic attention fusion design, and feature enhancement and dynamic weighted integration are realized through a kernel selection strategy; and (3) the dynamic Tanh (DyT) replaces the traditional normalization layer, and two-dimensional convolution is introduced to compensate the information loss of the Transform layer. According to the innovative design, the AKSFormer keeps the advantages of Transform global modeling, and meanwhile, the problems that local feature extraction and small sample generalization are insufficient in ability and the like are effectively solved. Experiments on three common data sets and one homemade data set show that compared with the existing advanced method, the AKSFormer classification precision is improved.

This invention discloses an adaptive block width ultrasound speckle tracking blood flow velocity measurement method and system, relating to the field of ultrasound blood flow velocity measurement. The method includes: determining a reference frame image and a comparison frame image from a B-mode ultrasound image; dividing the pixels in the reference frame image corresponding to the blood vessel into multiple blocks; searching for the best matching block of each block in the comparison frame image; calculating the displacement of each block relative to its corresponding best matching block; calculating the displacement gradient and velocity of each block; when the displacement gradient of a block meets a set threshold range, the velocity of the block is taken as the blood flow velocity at that block; otherwise, the size of the block is scaled. When the displacement gradient of the scaled block meets a set threshold range corresponding to the scaled block, the velocity of the scaled block is taken as the blood flow velocity at the block's location. Based on the velocities of each block, a blood flow velocity profile is determined. This invention can improve the accuracy of ultrasound imaging blood flow velocity measurement results.

The invention relates to the technical field of 3D Gaussian sputtering reconstruction and geometric accuracy optimization, in particular to a building three-dimensional reconstruction method and system based on 3D Gaussian sputtering, and the method comprises the steps: recognizing a building right-angle component through a two-channel network, outputting a semantic segmentation map with geometric mutation features, and generating non-uniform Gaussian kernel distribution; the method comprises the following steps: reconstructing an isotropic kernel into an ellipsoidal anisotropic kernel aligned with a normal direction, establishing an independent texture channel, segmenting a cross-plane Gaussian kernel, clearing the texture weight of an overlapping region, introducing a right-angle coplanarity loss function to correct kernel coplanarity deviation, and carrying out gradient attenuation on kernel density in a wall surface normal direction; an auxiliary kernel is implanted in an edge low-confidence high-deviation area, coordinates are optimized through back projection calibration, the system comprises a building component semantic analysis unit, a self-adaptive kernel distribution control unit, a parallel texture processing unit and a geometric precision enhancement unit, and finally a high-precision reconstruction model which is free of right-angle distortion and high in edge resolution is output to meet the building construction drawing restoration requirement.

The present application relates to the control field, specifically to a kind of AGV relative position control method based on kernel function prediction, method includes: AGV relative motion state feature is collected and constructs timing sequence and expected timing sequence;For each feature, based on the bandwidth adjustment coefficient of multiple cycle length analysis calculation, the feature adaptive bandwidth adjustment is carried out to Gaussian kernel function, and the adaptive kernel similarity between current and historical time is calculated;Based on the similarity and Gaussian process regression model, obtain the initial prediction value under each cycle length, and are fused into comprehensive prediction value by confidence weight;Finally, according to the deviation of comprehensive prediction value and expected value, the relative position of AGV is closed-loop controlled.The present application realizes feature adaptive kernel similarity calculation and multiple cycle prediction fusion, and improves the control precision and robustness under the disturbance such as tire slip, load change etc.

The application provides a non-Gaussian noise suppression method and device with adaptive kernel width, and belongs to the field of inertial base combined navigation algorithm and state estimation. The method solves the problems that the robust filtering method based on fixed kernel width is difficult to adapt to time-varying noise characteristics, and the scheme depending on an optimization algorithm has the problem of insufficient real-time performance. The method comprises the following steps: initializing filter parameters according to a SINS / DVL combined navigation system; performing time updating to obtain a predicted state vector and a predicted state covariance matrix at the current moment; updating a measurement noise covariance matrix through a variational Bayesian method, wherein the measurement noise covariance matrix is modeled as an inverse Wishart distribution; adaptively updating a kernel width parameter according to a filter innovation at the current moment and the measurement noise covariance matrix; and updating a state quantity estimation value and a state covariance matrix at the current moment through a fixed-point iteration algorithm by using the updated kernel width. The method is used in the field of underwater resource exploration.

This invention discloses an online prediction method for battery SOC based on an incremental correlation vector machine (RVM) strategy using a multi-kernel ensemble approach. The method includes the following steps: Step 1, data preprocessing; Step 2, training set sampling; Step 3, kernel function selection; Step 4, model training; Step 5, model validation; Step 6, adaptive kernel parameters; Step 7, RVM model ensemble; Step 8, model prediction; Step 9, incremental learning strategy; and Step 10, online incremental prediction. From a practical perspective, this invention addresses the complexities and diverse needs of various applications. Drawing on the ideas of incremental learning and ensemble learning, it generates highly differentiated RVM individual learning models containing multiple kernel functions through dual perturbation of training samples and kernel functions. Combined with a novel incremental ensemble strategy, it avoids the problem of model overlearning, improves the model's generalization ability and robustness, and expands its application scope.

The present application relates to the technical field of 3D Gaussian sputtering reconstruction and geometric accuracy optimization, in particular to a building three-dimensional reconstruction method and system based on 3D Gaussian sputtering, which identifies building right-angle components through a double-channel network, outputs a semantic segmentation graph with geometric mutation characteristics, generates a non-uniform Gaussian kernel distribution, contains a right-angle inflection point ellipsoidal type forbidden area and a wall surface normal direction gradient attenuation kernel density, reconstructs the isotropic kernel into an ellipsoidal type anisotropic kernel aligned with the normal, establishes an independent texture channel, segments the cross-surface Gaussian kernel and clears the texture weight in the overlapping area, introduces a right-angle coplanarity loss function to correct the kernel coplanarity deviation, implants an auxiliary kernel in the edge low-confidence high-deviation area, and optimizes the coordinates through back projection calibration. The system includes building component semantic analysis, adaptive kernel distribution control, parallel texture processing and geometric accuracy enhancement unit, finally outputs a high-precision reconstruction model with right-angle non-distortion and high edge resolution, and adapts to the building construction drawing restoration requirement.

The application discloses a standard meter performance degradation online detection method, which comprises the following steps: constructing a six-dimensional state vector, lifting the standard meter error data to a high-dimensional phase space, and accurately describing the dynamic characteristics of the equipment. The probability density distribution and Jacobian matrix of the state vector are calculated by adaptive kernel density estimation and weighted least squares method, and the local entropy density, geometric weight and topological analysis are further fused to form a continuous and differentiable entropy field structure. The entropy field gradient is used to derive the metric tensor, affine connection coefficient and curvature tensor to complete the construction of the Riemannian geometry structure. Based on the curvature scalar, topological invariant and entropy density criterion, the performance degradation is accurately identified, and early warning is provided. The application can deeply mine the dynamic changes and system structure evolution in the equipment error data, significantly improve the monitoring accuracy and early warning ability of the performance degradation, has strong innovation and practicality, and is suitable for long-term monitoring and fault warning of power metering equipment.

The invention relates to the field of control, in particular to an AGV relative position control method based on kernel function prediction, and the method comprises the steps: collecting the relative motion state characteristics of an AGV, and constructing a time sequence and an expected time sequence; for each feature, calculating a bandwidth adjustment coefficient based on multi-cycle length analysis, performing feature adaptive bandwidth adjustment on a Gaussian kernel function, and calculating an adaptive kernel similarity between the current moment and the historical moment; based on the similarity and a Gaussian process regression model, obtaining initial prediction values under each period length, and fusing the initial prediction values into a comprehensive prediction value through confidence weight fusion; and finally, carrying out closed-loop control on the relative position of the AGV according to the deviation between the comprehensive predicted value and the expected value. According to the method, fusion of feature adaptive kernel similarity calculation and multi-cycle prediction is realized, and the control precision and robustness under disturbance of tire slipping, load change and the like are improved.

This invention discloses a method for characterizing wind and solar power uncertainties, comprising: fitting the output characteristics of wind power and photovoltaic power using an adaptive kernel density estimation method to obtain marginal distributions; using a Copula function to describe the correlation between wind power and photovoltaic power outputs and constructing a joint distribution model; combining the Latin hypercube sampling method with the Copula conditional distribution inverse transform to generate an initial wind-solar joint scenario; using the Kantorovich-SBR algorithm to reduce the initial scenario, constructing a transportation cost matrix by calculating the Euclidean distance between scenarios, selecting representative scenarios using a greedy strategy, and solving for the optimal transportation plan allocation probability weights through linear programming. This invention overcomes the shortcomings of the traditional MC-Kmeans method, such as low sampling efficiency and high omission rate in extreme scenarios, and can accurately characterize the non-Gaussian and nonlinear correlation relationships of wind and solar power outputs, providing reliable support for power system reserve configuration, risk assessment, and energy storage planning.

The invention relates to a multi-modal feature-based network construction type energy storage fault diagnosis method and system, and the method comprises the steps: collecting the operation data of an energy storage system, and carrying out the preprocessing of the operation data; performing feature extraction including a time domain and a frequency domain on the preprocessed operation data to obtain feature vectors of different modes; dividing the feature vectors of different modes into feature groups according to a physical domain, calculating the abnormality of each feature group, mapping the abnormality into a weight coefficient of each feature group through a full connection layer, and carrying out the weighted fusion of each feature group based on the weight coefficient, thereby obtaining a final fusion feature vector; and inputting the final fusion feature vector into an SVM classifier designed based on an adaptive kernel function, and carrying out fault type identification. According to the method, the problems of low diagnosis accuracy, poor real-time performance and high false alarm rate due to too single data of the existing diagnosis method are solved.

This invention discloses a point-line feature cascade matching method based on adaptive kernel density clustering and multimodal constraints. The method includes: detecting and extracting SIFT point features and LSD line features from image pairs; obtaining initial point feature matching pairs through bidirectional matching and performing initial image transformations; introducing adaptive kernel density parameters to perform spatial clustering on the initial point feature matching pairs; merging the clustering results of the features; iteratively finding the maximum number of point feature matching pairs; constructing point-line invariants; calculating and comparing the radiosity, geometric similarity, and NCC values ​​between line features; and obtaining robust and accurate line feature matching pairs through a voting matrix. This invention obtains point feature matching pairs through point feature kernel density clustering and guides the construction and calculation of multimodal constraints for line feature matching, fusing multiple conditions to match line features, thereby achieving accurate, efficient, and robust matching of line features.

The invention relates to the technical field of overhead line fault detection, in particular to an intelligent diagnosis method for galloping and disconnection of an overhead line, and the method comprises the steps: collecting acceleration data and attitude angle data of the overhead line in different operation states through an attitude sensor, forming a pre-labeled training sample set, and calculating an acceleration change rate and an angle change amount; establishing a time sequence feature set constructed based on a time window, and calculating extension features in the time window; applying a feature weight matrix; adopting a radial basis kernel function to calculate a sample minimum distance to obtain an adaptive kernel width parameter; setting a class imbalance compensation factor according to the sample quantity proportion; a weighted feature vector, a self-adaptive kernel width parameter and the class imbalance compensation factor obtained in the fourth step serve as input, a radial basis kernel function is adopted to construct a multistage support vector machine based on a binary tree structure, and the state of the overhead line is diagnosed; and establishing a confidence backtracking mechanism, and automatically checking a low-confidence classification result.

A multi-model power system inertia probabilistic prediction method based on porcupine optimization and adaptive kernel density estimation includes the following steps: acquiring power system inertia-related characteristic variables, constructing a data sample set for inertia prediction, and building a CNN-BiLSTM-MHAM deep learning model based on the data sample set; optimizing key hyperparameters of the model using the porcupine optimization algorithm based on the constructed CNN-BiLSTM-MHAM deep learning model to obtain optimized model structure parameters and inertia prediction results; constructing an error database based on the error between the inertia prediction results and actual values; and using the constructed error database, probabilistically modeling the prediction error using the adaptive bandwidth kernel density estimation method, and generating the probability interval for inertia prediction by combining the Bootstrap resampling method, thereby realizing the quantification and probabilistic expression of the uncertainty of the inertia prediction results. This method not only significantly improves the accuracy of power system inertia prediction but also more effectively characterizes the uncertainty and probability distribution characteristics of inertia fluctuations.

The application discloses a data and knowledge double-driven wheel set multi-parameter comprehensive state evaluation method and system, and belongs to the technical field of railway vehicle maintenance. The method comprises the following steps: firstly, a mapping relationship between geometric parameters and dynamic performance indexes is constructed through sampling and dynamic simulation, and global sensitivity analysis is performed to screen key dynamic performance indexes; secondly, the subjective and objective weights of the indexes are determined in combination with an analytic hierarchy process and an entropy weight method, and a dynamic optimization model based on Bellman equation is introduced to generate final combination weights; then, a multi-dimensional state space division model is constructed by using adaptive kernel density estimation and fuzzy C-means clustering, and the probability density and membership function of each index under different health grades are determined; finally, a simulation prediction is performed on a target wheel set, the predicted value of the target wheel set is input into the state space model, and the dynamic combination weights and D-S evidence theory are fused to calculate a comprehensive health index and output a grading result, so that the wheel set health state can be accurately and efficiently evaluated.