106results about How to "The clustering result is accurate" patented technology

The invention discloses an image appearance based loop closure detecting method in monocular vision SLAM (simultaneous localization and mapping). The image appearance based loop closure detecting method includes acquiring images of the current scene by a monocular camera carried by a mobile robot during advancing, and extracting characteristics of bag of visual words of the images of the current scene; preprocessing the images by details of measuring similarities of the images according to inner products of image weight vectors and rejecting the current image highly similar to a previous history image; updating posterior probability in a loop closure hypothetical state by a Bayesian filter process to carry out loop closure detection so as to judge whether the current image is subjected to loop closure or not; and verifying loop closure detection results obtained in the previous step by an image reverse retrieval process. Further, in a process of establishing a visual dictionary, the quantity of clustering categories is regulated dynamically according to TSC (tightness and separation criterion) values which serve as an evaluation criterion for clustering results. Compared with the prior art, the loop closure detecting method has the advantages of high instantaneity and detection precision.

The invention relates to a high sliding window data stream anomaly detection method based on layered clustering, and aims to solve the problem that the accuracy of a data stream anomaly detection result is reduced due to influences of stale data and historical data. According to the method, by means of a layered clustering algorithm, the final clustering result cannot be considered during clustering, arrival data are processed at a higher speed, and a data volume of an off-line layer is greatly smaller than the number of original data due to the fact that the off-line layer only utilizes a clustering structure to respond to a user query result, so that the data can be effectively stored, and a more accurate clustering result can be obtained. As for a sliding window model, a clustering characteristic index histogram structure is adopted, so that insertion of new data and deletion of stale data can be better finished. A cosine coefficient is taken as a metric function, so that good clustering and anomaly detection results can be obtained. The high sliding window data stream anomaly detection method is applicable to fields of sensors, network click stream, share dealing and the like.

The invention discloses a multi-stage fermentation process fault monitoring method based on a self-adaption FCM algorithm. The multi-stage fermentation process fault monitoring method based on the self-adaption FCM algorithm solves the following problems that clustering of multi-batch three-dimensional data can not be achieved, the number of divided stages needs to be appointed manually, the center of clustering is initialized at random, and the method is prone to being affected by sample noise and jump points when the standard FCM algorithm is used for dividing stages in the fermentation process. The method comprises the specific steps that firstly, similarity indexes of all time data matrixes are calculated to serve as clustering input samples, an initial clustering center set is obtained according to the maximum and minimum clustering rules, and then a clustering effectiveness function is introduced to determine the optimal number of clusters through the self-adaption iteration. The method achieves the division of the stages of the fermentation process based on multiple normal operation batch data, so that the stage division process is more objective and accurate, a staged modeling monitor model reduces the false alarm rate and false negative rate of faults, and the method has the important significance for achieving control over the fermentation process and fault detection.

A spectral clustering method for automatically determining the number of clusters based on a neighboring point method comprises the steps of 1) normalizing all dimensions of a data set; 2) calculating an interval sparse distance matrix by a neighboring point method and defining the matrix as local scale parameters of distance mean values of the neighboring points to obtain a whole sparse similarity matrix; 3) determining the local density of each data point and the minimum distance to other points with a higher local density by calling a CCFD method, and obtaining the number of singular points generated by the fitting outside a confidence interval; 4) calculating a degree matrix D and a Laplacian matrix L according to a formula and extracting an eigenvector group by eigen decomposition of L; 5) outputting clustering results; and 6) selecting and outputting the clustering result with the optimal number of neighboring points corresponding to the maximum Fitness function value. According to the invention, the local scale parameter of each data point can be estimated according to data distribution, the number of clustering centers is automatically determined, and the parameter adaptation of the number of neighboring points is realized.

The invention discloses a semi-supervised text clustering method and device fusing pairwise constraints and keywords. The method comprises the following steps of: fusing pairwise constraints to assist in text clustering to obtain an initial feature word weight; fusing the pairwise constraints and keywords and performing the semi-supervised clustering at the same time based on the obtained initial feature word weight; and evaluating and selecting a clustering result according to a user satisfaction degree. The device provided by the invention comprises a pre-processing module, a text clustering module fusing pairwise constraints, a semi-supervised text clustering module fusing pairwise constraints and the keywords, and an evaluation and selecting result module. Since the semi-supervised text clustering method provided by the invention continuously adds keyword information on the basis of fusing pairwise constraint information, the keyword information is used for adjusting the corresponding feature word weight while applying the pairwise constraints to learning the feature word weight; and therefore, the two prior information can be mutually influenced and promoted to obtain a more accurate clustering result.

The invention provides a client feature library generating method, which comprises the steps of: calculating the sum d of intra-class distances under different cluster numbers k, and drawing a k-d value curve; finding a turning point according to the k-d value curve, finding a k value corresponding to the turning point, looking for an optimal cluster number kopt within a range near the k value, and utilizing a cluster result corresponding to the optimal cluster number as a final cluster result. The invention also provides a client feature library generating device. The scheme of the invention can improve the cluster searching efficiency and precision, and the method of the invention is used for performing client analysis according to the obtained client feature library to improve the recommendation success rate and client satisfaction degree of a marketing campaign.

The invention discloses a density peak clustering algorithm based on K neighbors and shared neighbors. The density peak clustering algorithm is used for solving the technical problem that an existingdensity peak clustering algorithm is poor in clustering effect. The technical scheme is to improve a DPC algorithm based on the similarity between K-neighbors and shared neighbors, wherein the attribution of each data sample point is determined by KNN distribution information and SNN shared neighbor similarity; and if more points belonging to a certain class cluster in the KNN (i) of the i exist,and the Euclidean distance between the points and the i is closer, the similarity between the two sample points is greater, and the attribution value of the sample i relative to the class cluster to which the KNN (i) belongs is greater while the probability that the sample point i is distributed to the class cluster is greater at the moment.The clustering center appears in an area with higher local density. The density peak clustering algorithm provided by the invention avoids the defect that the DPC algorithm measures the sample density and the associated distribution error similar to the domino effect generated during sample distribution, and the clustering effect is good.

The invention proposes a characteristic vector group's best selected spectrum clustering method based on density self-adaptation. The method comprises the following steps: 1) first, performing pre-treatment; 2) calculating the sparse similarity matrix based on density self-adaptation; 3) calling the automatically determining clustering central algorithm; 4) decomposing the characteristics and seeking the characteristic vector group; 5) performing the standardizing treatment to all elements in the data set in the mapped characteristic vector group in the characteristic space, followed by the K-means clustering to obtain the clustering result; and 6) calculating the Fitness function values; performing iterations constantly; and selecting and outputting the clustering result of the number of the best clear-cut points corresponding to the highest Fitness function value. According to the invention, it is possible to introduce the data point density information to a similar function so as to improve the time gathering effect of multi-scale data of the algorithm processing, to select the best characteristic vector group and to utilize the Fitness function to realize the parameter self-adaptation of the number of clear-cut points.

The invention relates to an incomplete data fuzzy clustering method for information feedback RBF network estimations, which comprises the following steps: 1) presenting an information feedback RBF network model; 2) presenting an incomplete data fuzzy clustering method (IFRBF-FCM) of information feedback RBF value estimations; 3) selecting the corresponding training sample set for the incomplete data sample by using the nearest neighbor rule, and training the IFRBF network for each missing attribute by using the nearest neighbor training sample set, thereby realizing the estimation prediction of the missing attribute in the incomplete data sample and obtaining the complete data set after the estimation recovery of the IFRBF network; 4) determining the estimation interval of the attribute ofthe incomplete data to propose an incomplete data fuzzy clustering method (IFRBF-IFCM) of IFRBF interval estimations to obtain fuzzy clustering results. The invention adopts the IFRBF network to estimate the incomplete data set and recovers the intact data set. Compared with the comparison method, the clustering result of the intact data set is more accurate than that of numerical type estimations, and the robustness is better.

The invention discloses a possibility fuzzy c mean clustering algorithm based on multiple kernels. The possibility fuzzy c mean clustering algorithm is characterized by comprising the following steps of: 1, performing optimal partition on a sample set in order to minimize a objective function value; 2, acquiring an initial membership matrix and an initialized clustering center; 3, acquiring a membership value, a clustering center, and a typical value by iteration; and 4, acquiring an objective function with an introduced weighted index. The possibility fuzzy c mean clustering algorithm may accurately avoid a problem that FCM is sensitive to noise and is liable to generate consistency clustering so as to be further increased in accuracy, and may discover a most appropriate weighted value and a current membership value so as to be increased in reliability and convergence.

The invention discloses a multimodal surgical trajectory fast segmentation method based on unsupervised deep learning, belonging to the field of robot assisted minimally invasive surgery. Firstly, therobot system collects kinematics data and video data for a robot-assisted minimally invasive surgery process. Then wavelet transform is used to smooth and filter the short trajectory noise and data jitter in kinematics data. At the same time, a DCED-Net network structure is used to extract features from video data. The kinematics data after smoothing filtering and the video data image after feature extraction are sent to the improved TSC model for clustering, and the pre-segmentation results of n surgical demonstration trajectories are obtained. Finally, the PMDD merging algorithm is used tosimilarly merge the pre-segmentation results of each trajectory, and the merging result is the final trajectory segmentation result. The invention provides an optimization scheme and unsupervised deeplearning for over-segmentation and other problems, accelerates the extraction speed of video features, improves the quality of features and makes the clustering result more accurate.

The invention discloses a multi-channel spectral clustering method based on local density estimation and neighbor relation spreading. The multi-channel spectrum clustering method based on local density estimation and neighbor relation spreading mainly solves the problem that an existing clustering method cannot carry out clustering on data distributed unevenly in density. The multi-channel spectrum clustering method based on local density estimation and neighbor relation spreading comprises the steps that local density of a sample is estimated and is used as data characteristics and dimension lifting is carried out on original data; a distance matrix, a threshold value and a similarity matrix are calculated, and a neighbor relation matrix is initialized; the neighbor relation matrix and the similarity matrix are updated, and similarity of samples of a subset is updated by the adoption of a local maximum similar value, and an accurate affinity matrix is obtained; a similarity matrix and a normalized Laplacian matrix are calculated; a spectrum matrix is normalized, and a clustering result is obtained through the K-means algorithm. Compared with an existing clustering technology, the multi-channel spectrum method based on local density estimation and neighbor relation spreading enables a more real similarity matrix to be obtained, the clustering result is more accurate and the robustness is better.

The invention discloses a pedestrian re-identification method and device, electronic equipment and a readable storage medium, and belongs to the field of image processing, and the method comprises thesteps: initializing network parameters of an original network model to obtain an updated network model, and enabling the updated network model to comprise a feature distribution alignment module, a clustering module and a noise label correction module; training the feature distribution alignment module by using the source domain sample and the target domain sample to obtain an alignment network model; training a clustering module based on the alignment network model and the target domain sample to obtain clustering results and noise tags corresponding to the clustering results; training the noise label correction module based on the updated network model, the target domain sample and each noise label to obtain a target network model; and performing pedestrian re-identification on the to-be-queried image by utilizing the target network model. The accuracy of the noise label is improved through feature level alignment and iterative clustering, and the precision of the noise label is improved through a noise correction process, so that the recognition accuracy of the target network model is improved.

The invention provides a clustering method and apparatus for wind generating sets. The clustering method comprises the flowing steps of: preprocessing the predetermined data in the data of each wind generating set in a wind power plant to obtain the characteristic data of each wind generating set; combining the characteristic data of each wind generating set and the unpreprocessed data in the data of each wind generating set into a data set of each wind generating set; and clustering the wind generating sets by clustering the data sets of all the wind generating sets in the wind power plant. The clustering method and apparatus cluster the wind generating sets according to the data set of each wind generating set so as to achieve reasonable classification.

The invention relates to a gene classification method and device. An LLE algorithm and an AP clustering algorithm are combined, and a proposed mixed kernel function is utilized to improve a similaritymeasurement function. According to the method, first, the LLE algorithm is adopted to map an original high-dimensional gene expression dataset to a low-dimensional space to achieve the purpose of dimension reduction; second, a new global kernel function is proposed as an F-type function, the F-type function and a Gaussian kernel function are linearly combined into a new mixed kernel function, theproposed mixed kernel function is utilized to calculate similarity measurement, and a new similarity matrix S is constructed; third, data is clustered through the AP clustering algorithm and the similarity matrix, and a final clustering result is obtained through iteration; and finally the effectiveness and accuracy of the method are verified through comparison with other clustering methods.

The invention provides a semi-supervised clustering method for fusing pairwise constraint and attribute sorting information, wherein the method comprises the following steps of: obtaining initial attribute weights according to the pairwise constraint information provided by the user, adding the attribute sorting information provided by the user on the basis of the initial attribute weights to implement semi-supervised clustering, and then selecting clustering results according to the indexes of accuracy. Due to the addition of attribute sorting information on the basis of fused pairwise constraint information, the semi-supervised clustering method for fusing pairwise constraint and attribute sorting information provided by the invention has the advantages that the corresponding attribute weight can be regulated by using attribute sorting while utilizing the learning attribute weight of pairwise constraint, and the clustering result which is obtained by means of the method is more accurate by influencing and promoting the two priori information mutually.

The invention discloses a hyperspectral remote sensing image segmentation method based on K-means clustering, and belongs to the technical field of clustering segmentation of hyperspectral remote sensing images. According to the invention, the problem of inaccurate segmentation of the hyperspectral remote sensing image by the existing clustering segmentation method is solved. According to the invention, collaborative clustering is carried out on d dimensions of the hyperspectral remote sensing image; when K-means iteration is performed on a hyperspectral remote sensing image each time, clustering results of d dimensions are cooperated; and after each K-means iteration is finished, a probability value of the pixel point belonging to each category is calculated by utilizing d clustering results of each pixel point, then a new clustering center is calculated by utilizing the obtained probability value, and a next iteration process of each dimension is started according to the same clustering center until a final clustering result is obtained. Compared with an existing method, the method has the advantage that the segmentation result is more accurate. The method can be applied to the technical field of clustering segmentation of hyperspectral remote sensing images.

The invention relates to an image data clustering method and device, and belongs to the technical field of image processing. According to the method, firstly, a self-recommendation strategy is proposed to determine a local center point, the subjectivity of manually selecting a cluster center is eliminated, and the problem that the center point of a cluster with low density is ignored is solved; the local central point is taken as a micro-cluster central point to distribute residual data points, and further a plurality of micro-clusters are generated; and finally, a micro-cluster merging strategy is proposed, and the micro-clusters and the micro-clusters containing fewer boundary points in the adjacent clusters thereof are merged because the micro-clusters are closer to the cluster center and more likely to be located in the same cluster as the micro-clusters, so that the clustering result of the method is more accurate.

The invention discloses a text data stream clustering algorithm based on affinity propagation. The text data stream clustering algorithm is characterized by including the following steps: 1, carrying out dimension reduction processing on a text data set to obtain a corresponding text vector set; 2, obtaining clustering centers of all moments, and completing the clustering algorithm. By means of the text data stream clustering algorithm, the accuracy and the robustness of the algorithm can be improved without assigning the number of clusters in advance, and therefore the requirements for solving practical problems are met.

The invention belongs to the field of data mining. Based on different contributions made to classification by properties of multipaths, classification precision can be improved after multipath component distance is subjected to characteristic weighting operation via use of an information entropy principle. A technical solution adopted in the invention is that an improved multipath clustering method comprises the following steps: wavelet transformation peak detection technologies are used for replacing random selection, and stable clustering effects can be obtained; based on consideration of effects exerted by multipath power, the information entropy principle is introduced for calculating multipath component distance (MCD) of multipath property self-adaption weighting, and different multipaths can be assigned to different clusters according to different MCD. The improved multipath clustering method is mainly applied to data processing occasions.

The invention discloses an intrusion behavior-oriented tracing data clustering method and device, and belongs to the technical field of computers, and the method comprises the steps: collecting tracing information of a system kernel, and carrying out the preprocessing to filter nodes irrelevant to an intrusion behavior and dependency relationships related to the nodes; converting the preprocessed tracing information into a tracing graph, and constructing an adjacent matrix and a node attribute matrix of the tracing graph; selecting a plurality of root nodes from the tracing graph as seed nodes of random walk, and enabling the seed nodes to walk randomly according to a walk policy to obtain a walk path of the tracing graph and a weight of a tracing edge, wherein the walk policy is that any node in the seed nodes walks from an adjacent matrix to an adjacent node at a probability p, and walks from a node attribute matrix to a node having the same attribute with the node attribute matrix at a probability (1-p); and performing clustering according to the walk path and the weight of the tracing edge. Different tracing events can be accurately distinguished, and more accurate data can be provided for subsequent detection and query.

The invention discloses a Kmeans clustering method for the efficacy of traditional Chinese medicinal materials based on node similarity. The method comprises the following steps: collecting related traditional Chinese medicine data, and processing the data to form a prescription composition library, a medicinal material efficacy library and a channel-tropism binary table of the nature and taste ofmedicinal materials; summarizing and classifying the efficacy of the traditional Chinese medicinal materials according to 23 efficacy tables, and constructing a medicinal material efficacy matrix; constructing a prescription-medicinal material bipartite network based on the prescription composition library; calculating expected values of medicinal material pairs based on degree distribution, andtaking the expected values of the medicinal material pairs as the similarity of the traditional Chinese medicinal materials; establishing a Kmeans clustering model based on the similarity of the traditional Chinese medicinal materials; and clustering the traditional Chinese medicinal materials based on the clustering model to obtain potential effects possibly possessed by the traditional Chinese medicinal materials. According to the method, the accuracy of Kmeans clustering via a medicinal material similarity matrix can reach 0.728. Meanwhile, Kmeans is used for clustering the nature-taste channel-tropism data of traditional medicinal materials, an obtained final result is 0.646, which is about 0.08 higher; and therefore, clustering result is allowed to be more accurate through the method.