42 results about "Kernel clustering" patented technology

The invention discloses a secondary clustering segmentation method for a satellite cloud picture. According to the secondary clustering segmentation method for the satellite cloud picture, firstly, block processing is carried out to the whole wide-range satellite could picture; secondly, corresponding multi-channel spectral features and three-patch length between perpendiculars (TPLBP) textural features are sequentially extracted from each of sample points of each sub regional could picture for fine initial kernel clustering segmentation so as to obtain multiple sub regional could picture segmentation results; and at last, secondary kernel clustering segmentation is carried out to the global cloud picture on the basis of the initial kernel clustering segmentation results by utilization of the initial kernel clustering segmentation results as prior knowledge to extract a variety of grayscale average features and density indicator features of the initial kernel clustering segmentation results in a corresponding original sub regional cloud picture range so as to ensure integrity of cloud classification. The secondary clustering segmentation method for the satellite cloud picture has the advantages of being high in precision and robustness, and capable of identifying cloud classification in a fine mode according to huge and complicated geographical information, ocean information, atmospheric information and other information which are contained in the wide-range meteorological satellite cloud picture.

The present invention discloses a low-voltage area KFCM-SVR reasonable line loss prediction method. The method comprises the following steps: 1) presorting data samples twice after data screening; 2) clustering six types of data by using fuzzy kernel clustering to obtain a plurality of small types; 3 ) establishing a regression fitting model by virtue of a least squares support vector machine (LS-SVR); 4 ) calculating a residual error confidence interval limit value; 5) classifying to-be-predicted data into each small types according to a Euclidean distance principle; 6) after classification, inputting the to-be-predicted data into each respective corresponding LS-SVR model, and calculating to obtain a prediction residual error; and 7) comparing the prediction residual error with the residual error confidence interval limit value to provide a conclusion. In the low-voltage area KFCM-SVR reasonable line loss prediction method disclosed by the present invention, an algorithm has certain reasonableness under a big data condition, and a computed result of the algorithm can provide relatively reliable guidance for area line loss management of an electric power department, and provide a new idea for reasonable utilization and mining of big data in the current smart power network.

The invention discloses an MGKFCM (multipath Gauss kernel fuzzy c-means clustering algorithm). The MGKFCM comprises the following steps: 1, performing optimized division on a sample set according to the principle of minimization of an MGKFCM objective function; 2, initializing clustering centers and calculating fuzzy membership and an objective function value 1 with a gradient iterative formula; 3, estimating the clustering centers and calculating the fuzzy membership and an objective function value 2 with PSO (particle swarm optimization); 4, selecting a cluster center with smaller objective function value as an iteration path on the basis of the objective function value 1 and the objective function value 2; 5, obtaining the MGKFCM objective function through calculation. Two Gaussian kernel clustering algorithm iteration paths including a gradient method and the PSO are integrated, the path with the smaller clustering objective function value is taken as the parameter iteration path, optimal performance of the two algorithms is used effectively, and the clustering performance of the clustering algorithm is improved.

The invention discloses a dynamic weighted hybrid clustering algorithm based circuit breaker fault diagnosis method. The method includes the following steps: (1) capturing the energy changes of mechanical drive during the operation of a circuit breaker by utilizing three-axis vibration and two-way sound signals, decomposing the signals through local mean, and extracting the approximate entropy ofeach PF component as the characteristic quantity of a circuit breaker vibration signal; (2) optimizing the initialized clustering center of fuzzy kernel clustering by utilizing the maximum density peak decision of a density peak clustering algorithm, and considering different influences of different characteristics and different samples on clustering results; (3) performing checking on a clustering number K through a cluster validity index MIA; (4) inputting correctly classified characteristics into a multi-level classifier of a support vector machine to perform training; and (5) finding the optimal parameter of the support vector machine through mesh generation, and inputting test data samples to perform final fault classification prediction so that classification accuracy rates can be obtained;. The method has advantaged of being fast in fault diagnosis speed and high in accuracy rate.

A method for discriminating indoor positioning floor based on improved fuzzy kernel clustering includes collect received signal intensity (RSSI) of only WiFi access point (AP) visible in each floor ofpositioning area, and establishing original position-fingerprint database corresponding to each floor and the signal intensity distribution range database of all APs between floors; 2. by adopting the improved fuzzy kernel clustering algorithm, clustering the original position-fingerprint databases to obtain location-fingerprint database with classification markers; 3. according to the real-timecollected WiFi signal intensity data, the floor of the point to be detected is judged by rough classification and clustering center. The invention is based on the idea of rough classification and clustering, and can quickly and effectively judge the floor of the point to be detected in the indoor positioning.

The invention discloses a text matching method using a semantic parsing structure. The method comprises the following steps: defining an initial corpus set Cqa and a supplementary corpus set Cq; defining Semantic structure DP-tree corresponding to text by using a semantic dependency analysis method; Defining a kernel function of the text and a metric function of text similarity based on the semantic structure; Carrying out kernel clustering on the text; obtaining an aggregated text class function(shown in the specification), wherein i = 1, 2, ..., M, q'ij is ni sample points which are selectedfrom each cluster and are closest to the cluster; And through manual audit, approving the Ci class and marking the Ci class with a specific tag Ti. According to the invention, syntactic analysis structures such as a syntactic structure are used as a comparison basis; A convolution kernel function theory and tree kernels (tree kernel, TK) are combined to define a kernel function representing the distance between two tree syntactic structures, and internal and external knowledge of syntactic similarity, word vectors, word sense networks and the like is introduced, so that the similarity betweentexts can be accurately judged.

The invention discloses a newly increased requirement-based code variation quantitative evaluation method and a system thereof and belongs to the technical field of computer software development. The method comprises the following steps of: (1) carrying out feature extraction on N requirements of a certain software item according to set requirement features; (2) selecting feature extraction data of m requirements in the item as training data and calculating an adjustment coefficient gamma of the software item; (3) carrying out feature extraction on newly increased requirements of the software item according to the set requirement features; (4) carrying out clustering on each newly increased requirement and the N requirements by using the feature extraction data in steps (1) and (3); and (5) estimating a code quantity E of the newly increased requirements according to a formula E=gamma.C. The system comprises a feature extraction module, a kernel clustering algorithm calculation module, an adjustment parameter gamma training module and a code quantity prediction module. Compared with the prior art, the system of the invention can automatically predict the code quantity to be generated by the newly increased requirements and can visually display an estimation result.

The invention discloses an image segmentation method based on an immune clone algorithm and a fuzzy kernel-clustering algorithm. The image segmentation method aims to solve the technical fields of sensitive initial clustering center, easy local optimum falling, and low segmentation correct rate of a fuzzy C-means clustering method. The image segmentation method includes the steps of reading an image to test the image size and convert the image to a grey-scale image; setting the parameters of a fuzzy kernel-clustering method; filtering the grey-scale image by a non-local means filtering method capable of automatically adjusting parameters to obtain the filtered image; optimizing the grey-scale image of the input image by means of the immune clone algorithm to obtain an initial clustering center; obtaining a final membership grade matrix uki by means of membership grade matrix formula with filtering and obtaining the final clustering center value v2 by means of a clustering center formula with filtering; and outputting the segmented image by means of fuzzification. The image segmentation method has the advantages of local optimum falling free, high segmentation accuracy, and good robustness, and can be applied to the segmentation of artificial synthesis images, medical images natural images or other images.

The invention discloses an airport noise event recognition method based on a time series kernel clustering, and belongs to the airport noise recognition field. This method comprises the following steps of: acquiring a raw monitoring time series of airport noise; pre-processing the raw monitoring time series, creating a noise time series data set; using a k-means clustering method based on an overall alignment kernel to automatically classify the noise time series data set and to get multiple clustering results through cycles; adding a noise event category label corresponding to each type of the noise time series in the multiple clustering results; creating a noise event knowledge base; and finally carrying out the noise event recognition to the noise time series to be recognized according to the knowledge base. The method of the invention has a high degree of intelligence, and effectively improves the efficiency and accuracy of the airport noise event recognition.

The invention discloses an SAR (Synthetic Aperture Radar) image change detection method based on Memetic kernel clustering; and the SAR image change detection method mainly solves the problems that the conventional algorithm is high in time complexity and easy to fall into a local optimum value, and has bad performance in region consistency and edge retaining. The implementation process of the SAR image change detection method comprises the steps of: (1) inputting two SAR images at different times, and performing median filtering on the two images; (2) computing the logarithmic ratio difference striograph of the two time-phase images subjected to change detection; (3) setting initial conditions; (4) carrying out kernel clustering and computing a fitness function fk; (5) selecting optimal individuals after performing clone and dual mutation operation on current individuals; (6) selecting optimal individuals after performing clone and crossover operation on the optimal individuals obtained in the step (5); (7) selecting a final individual by an elitist strategy; and (8) judging stop conditions, and outputting clustering results if the conditions are satisfied, otherwise returning to the step (4). The SAR image change detection method based on Memetic kernel clustering has the advantages of rapid convergence rate, high detection precision and accuracy in edge retaining, and can be applied to target identification and change detection of the images in the image processing field.

The invention discloses a Gauss induction kernel based fuzzy c-means clustering algorithm, which comprises the steps of 1, performing optimized classification on a sample set according to an objective function minimization principle; 2, initializing a fuzzy membership degree or initializing a clustering center; 3, performing parameter estimation on the fuzzy membership degree and the clustering center according to an iterative calculation formula in a Gauss induction kernel clustering algorithm; and 4, acquiring an optimized objective function. The clustering performance of a kernel clustering algorithm can be improved by effectively utilizing the nonlinear expression ability of a kernel method in the clustering algorithm. The clustering center iteration formula does not contain the clustering center itself, conditions of the clustering algorithm for iterating a convergence proof are met, and thus the convergence of the algorithm is ensured theoretically.

The invention belongs to the technical field of music prediction model, and discloses a time series solving method and system for music traffic prediction. The method comprises the following steps that: carrying out model construction for a music user, and using a weighted fuzzy kernel clustering model to carry out a music use clustering operation; on the basis of artist dataset obtained by user clustering, carrying out the construction of the music traffic prediction model, and using algorithm based on feasible coefficient space segmentation for prediction; and verifying the efficiency and the accuracy of the artist music traffic prediction in ta music playing traffic prediction model. The method assists a player manufacturer to reasonably arrange a player background and a proper networkbandwidth so as to effectively lower the probability of occurrence of crash, meanwhile, through the research of the integral classification of music users and a classification prediction result of music, the analysis of the player manufacture for the development tendency of the music traffic is improved, and a result generated by prediction reduces loss for enterprises.

The invention discloses an electric power user credit evaluation method based on a global optimal fuzzy kernel clustering model, which comprises the following steps: establishing an electric power user credit evaluation index system, and preprocessing the data of the evaluation index; The global optimal fuzzy kernel clustering algorithm model is established. Input the pre-processed credit evaluation data of electric power users into the established global optimal fuzzy kernel clustering algorithm model to obtain the clustering results; Qualitative analysis is carried out on the obtained clustering results to obtain the credit rating of each category; Extracting characteristic indexes and corresponding eigenvalues of each cluster in the clustering results; According to the extracted eigenvalue index of each cluster and the corresponding eigenvalue, the credit result of the electric power users to be evaluated is evaluated again. It replaces the manual work of evaluating the credit of electric power users, lightens the investment of manpower, saves the management cost of electric power enterprises, and improves the scientificity and accuracy of the evaluation work.

The invention discloses a thermal error compensation calculation method, a thermal error compensation calculation device and thermal error compensation calculation equipment of a machine tool, and a storage medium. The thermal error compensation calculation method comprises the steps of: acquiring multiple physical parameters of composite cutting machining conditions, wherein each physical parameter comprises at least two state variables; calculating a state quantity combination of the physical parameters having the greatest influence on target parameters of the machine tool based on a Taguchiexperiment method; assigning the state quantity combination to the composite cutting machining conditions corresponding to the machine tool, so that the machine tool performs machining under the state quantity combination; acquiring temperature values and a thermal deformation amount after temperature and error synchronous measurement; grouping the temperature values by using a fuzzy C-mean clustering algorithm to obtain k temperature groups, and taking the center of each temperature group as a cluster temperature feature point; acquiring a kernel clustering factor based on the k cluster feature points and a rough set theory method; and establishing a linear regression thermal error compensation model based on the kernel clustering factor and the thermal deformation, and performing thermal error compensation calculation on the machine tool. The thermal error compensation calculation method, the thermal error compensation calculation device and the thermal error compensation calculation equipment improve the linear positioning and rotary positioning precision.

The invention belongs to the field of image processing, provides a method for quickly segmenting a finger vein image, and provides reliability guarantee for subsequent feature extraction and recognition of finger veins. Therefore, the technical scheme adopted by the invention is as follows. The invention discloses a finger vein image quick segmentation method based on an active contour model. A fuzzy kernel clustering KFCM algorithm is used to preliminarily determine a general contour of a vein area. And establishing a target functional, taking the preliminarily determined general contour of the vein region as an initial contour of the established target functional optimization, continuously iteratively solving a minimum value of the target functional, and continuously evolving the initialcontour in an iterative optimization process until an optimal segmentation result is achieved, thereby finally realizing image segmentation. The method is mainly applied to image processing occasions.

The invention discloses an electric reactor fault diagnosis method. The method comprises the following steps: decomposing a vibration signal of an electric reactor, and obtaining M IMF components; selecting several IMF components as data source for fault feature extraction; constructing fault feature quantity through the selected IMF components; and diagnosing an electric reactor fault state in combination with a kernel clustering algorithm. The method provided by the invention can improve accuracy of fault diagnosis.

The invention discloses a method for establishing a load soft measurement model of a cement combined grinding ball mill, which is established on the basis of a fuzzy clustering cognitive network, andcomprises the following specific steps of: determining fuzzy cognitive network nodes, and performing normalized representation on node state values; determining node weights according to causal relationships among the nodes; establishing a function relationship between the node state value and the weight and the state value at the previous moment; wherein the node weight is updated along with thesystem feedback value; calculating a node state estimation value by adopting a possibility kernel clustering algorithm; and establishing a soft measurement rule base, and replacing the system feedbackvalue of the fuzzy cognitive network with the node state estimation value according to the rule. The invention further discloses a soft measurement method applying the model. According to the soft measurement technology, by means of the function relation between the auxiliary variables easy to measure and the dominant variables difficult to measure, online measurement of the dominant variables, namely the mill load, is achieved, and the actual working condition of the mill can be accurately reflected.

An iris sorting scheme based on kernel clustering, belongs to the image processing techniques field, and relates to the technology of iris identification. When an iris database establishes clustering,firstly the iris normalization images are equally divided into an upper layer and a lower layer. The upper layer and the lower layer are analyzed by wavelet for three layers respectively, wavelet coefficients are chosen from seven channels to calculate energy and standard deviation; eigenvectors of iris database samples are obtained by which wavelet energy is divided by standard deviation of theupper layer and the lower layer; the eigenvectors is clustered based on kernel methods to obtain clustering results. When the iris waiting for testing is sorted, all eigenvectors in iris database areinput into a support vector machine to exercise, and support vectors and discriminant function are obtained; the corresponding eigenvectors of iris samples waiting for testing are input into discriminant function to have a test, and the sorting results are obtained. The invention has the advantage of that clustering which the iris sample belongs to can be rapidly and correctly found in favor of improving the iris identification accuracy rate and the identification efficiency.

The invention provides a convolutional neural network compression method and device and an image classification method and device, and the method comprises the steps: carrying out the convolution kernel clustering of each convolution operation layer in a pre-trained convolutional neural network, and obtaining a convolution kernel class cluster corresponding to each convolution operation layer; performing convolution kernel pruning on each convolution kernel class cluster corresponding to each convolution operation layer; and retraining the convolutional neural network after convolution kernel pruning to obtain a compressed convolutional neural network. According to the method, efficient convolution kernel pruning is achieved, meanwhile, the mode is not limited by different pruning sensitivities caused by convolution kernel distribution any more, and therefore consumption of computing resources and time resources of convolution kernel pruning is greatly reduced.

The present invention discloses a medical image segmentation algorithm used in a CT room. The segmentation steps of the algorithm are that: firstly, a CT image is segmented into different small areas via a watershed algorithm, then an average gray value of the small areas is mapped to a high dimension characteristic space according to a KFCM algorithm and by utilizing a Mercer kernel, so that the original characteristics which are not displayed in a watershed algorithm segmentation image are displayed. The algorithm comprises the following steps of (1) pre-processing an image, and carrying out the median filtering; (2) carrying out the watershed segmentation on the pre-processed image, and storing the label k of each small area; (3) calculating the average gray value x<k> of each area, wherein the average gray value x<k> represents a sample set of an input space, k =1, 2, ..., n, n is the area number formed after the image watershed segmentation; (4) selecting a classification number C, a threshold value epsilon and a fuzzy exponent m; (5) calculating a clustering center and a weighting matrix. The medical image segmentation algorithm used in the CT room utilizes the advantages of the watershed algorithm and the weighted kernel clustering, also greatly overcomes the disadvantages of the two algorithms.

The invention relates to application protocol screening and matching method research based on content filtering, and belongs to the field of power data exchange transmission technology innovation. The traditional power exchange network introduces the IMS technology to change the interaction mode of the B2B due to the limitation of the interface and system processing capability, thereby solving the development bottleneck of the business multimedia expansion of the power exchange network. In order to better meet the special requirements of a power grid on the aspects of safety, high efficiency and future service bearing, the invention provides an IMS capability-based open service access architecture, deployment authentication, firewall and other means to enhance the access safety of internal and external network services, and, through the processes of service protocol type matching, protocol recombination, kernel clustering protocol filtering and the like, the exchange service handling capacity is reduced. According to the invention, innovative technical contents are provided from the aspects of IMS power grid service development architecture, protocol processing, interface conversion and the like, and a basis with technical practice value is provided for power grid construction.

The invention discloses a multi-parameter intelligent threshold criterion of a high-speed train. Firstly, the working condition parameters and the state monitoring data of the high-speed train are collected, and the state monitoring data are divided according to the working condition parameters. Then, the fuzzy kernel clustering is used to analyze the condition monitoring data, and the center coordinates and the distance within the class are obtained. At last, the central coordinate, the distance within the class and the regression relationship between the multi-parameters of the working condition are calculated, and then the threshold value under various working conditions are intelligently calculated, and the high-speed train is intelligently alerted. The method is simple and applicableto the determination of intelligent threshold of monitoring data of high-speed trains under off-design conditions.

The invention discloses a later fusion multi-kernel clustering machine learning method and system based on proxy graph improvement. The later fusion multi-kernel clustering machine learning method based on proxy graph improvement involved in the invention comprises the following steps: S1, acquiring a clustering task and a target data sample; S2, initializing a proxy graph improvement matrix; S3, performing k-means clustering and graph improvement on each view corresponding to the obtained clustering task and the target data sample, and constructing a target function in combination with a kernel k-means clustering and graph improvement method; S4, solving the objective function constructed in the step S3 by adopting a circulation mode to obtain a graph matrix fused with basic kernel information; and S5, performing spectral clustering on the obtained graph matrix to obtain a final clustering result. The optimized basic division not only has information of a single core, but also can obtain global information through the proxy graph, and fusion of views is more facilitated, so that the learned proxy graph can better fuse information of each core matrix, and the purpose of improving the clustering effect is achieved.

The invention discloses an image classification method. The method comprises the steps of preprocessing a standard TWSVM algorithm to obtain a preprocessed TWSVM algorithm; executing a binary decisiontree construction operation according to an acquired training data set, and performing training processing according to the preprocessed TWSVM algorithm and non-leaf nodes in a constructed binary decision tree, thereby obtaining a decision function of each non-leaf node; and performing classification prediction processing on to-be-tested samples through the binary decision tree and the decision functions, thereby obtaining a classification result. According to the method, the TWSVM algorithm is expanded based on the binary decision tree; the efficiency of processing the multi-classification problem by the TWSVM algorithm is improved; and the accuracy of solving the nonlinear problem is improved through the binary decision tree constructed by kernel clustering. The invention further discloses an image classification training method, a classification prediction method, an image classification system, an image classification device and a computer readable storage medium, which have the abovementioned beneficial effects.

The invention discloses an optimal neighbor multi-kernel clustering method and system based on a local kernel, and the method comprises the steps: S11, obtaining a clustering task and a target data sample; S12, calculating a kernel matrix of each view corresponding to the target data sample, and carrying out centralization and normalization processing on the kernel matrix to obtain a processed kernel matrix; s13, according to the obtained processed kernel matrix, establishing an optimal neighbor multi-kernel clustering objective function based on a local kernel; s14, solving the established objective function by adopting a circulation mode to obtain a division matrix after view fusion; and S15, performing k-means clustering on the obtained division matrix to obtain a clustering result.