75results about How to "Clustering is fast" patented technology

The invention relates to a road and obstacle detecting method based on remotely piloted vehicles. The method includes adopting a four-wire laser radar as a distance sensor and calculating slope information of roads in a driving area according to the relative position correspondence of pavement data points on different scanning layers; fitting left and right road edges by the COBWEB algorithm and the least square fit improved on the basis of Euclidean distance according to characteristics of road edge data points, enhancing anti-jamming capability, accuracy and stability of road edge detection; applying DST (Dempster-Shafer theory) evidence theory to establish a raster map for the environment ahead of the remotely piloted vehicles, and estimating positions of each raster before integrating prior- and posterior-frame maps. Consequently, the problem of integration of prior and posterior raster cells in the local map is solved. Finally, dynamic faults can be detected by means of conflict coefficient in a driving area, and the dynamical obstacles can be clustered and information thereof can be extracted by the improved eight-neighborhood zone marker algorithm. The road and obstacle detecting method can stably and accurately detect road and obstacle information.

The invention belongs to the cross-media information technology field and particularly is an image optimization clustering method based on the typical correlation analysis. The invention mainly adopts the typical correlation to analyze while considering content characteristics of media data in various modes, maps the characteristics of the media data in various modes to an isomorphism sub-space of a united dimension through the sub-space mapping algorithm and obtains the final clustering result through optimizing clustering algorithm. The invention overcomes single-mode characteristic limitation in the multimedia field where only data is used, effectively solves the isomerism problem of the media data in various modes on the bottom layer characteristics, realizes the united measurement of the media object information between various modes, obtains results which are more accurate, more effective and more comforted to the needs in the large scale image data and has a wide application value in the cross-media processing and the retrieval field.

The invention relates to a search engine technology based on relevance feedback and clustering. By simultaneously utilizing user relevance feedback information and relavancy sequencing to direct the clustering of retrieval results, the invention ensures that the final partitioning of the retrieval results meet user query requirements; and in a clustering process, a large amount of documents and repeated webpage which are irrelevant to a user are removed, the clustering speed is improved and the retrieval results are optimized at the same time. In the clustering process, a clustering center is not modified by a clustering cluster irrelevant to the user, thereby result documents relevant to the user are ensured not to be lost when noise is introduced in irrelevant document clustering.

The invention brings forward a dancing robot indoor pedestrian tracing method based on a laser radar. The method comprises the following steps of: 1, laser data preprocessing; 2, density-based laser data clustering; 3, pedestrian identification based on laser acquisition points; 4, graphical display; 5, pedestrian tracking based on particle filtering; and 6, pedestrian movement track drafting. According to the invention, pedestrians with different forms at different distances are identified by use of a single laser radar, the pedestrians at different movement modes such as static, walking and the like can be tracked, the tracking scope is wide, the real-time performance is good, and the identification and tracking accuracy is high.

The invention relates to a modeling method for a parallel smart case recommendation model. The method comprises the following steps of obtaining existing patient cases from an electronic case database, carrying out denoising, clustering and word segmentation on the patient cases, and establishing a patient case corpus database; defining that TFIDFi, j shows the importance degree of a word or an expression in a case of the patient case corpus database, establishing an LSI vector space model according to the TFIDFi, j, and moreover, establishing a BOW word bag model according to all words and expressions in the patient case corpus database; calculating history case vectors and to-be-processed case vectors in the patient case corpus database through utilization of the LSI vector space model and the BOW word bag model; calculating cosine similarity among the history patient cases and storing the cosine similarity; and calculating the cosine similarity between the to-be-processed case vectors and the history patient case vectors, and searching similar cases of to-be-processed cases according to the cosine similarity. The model established through adoption of the method provided by the invention is high in accuracy and low in error. A recommendation result is high in quality.

The invention discloses a multilayer bitmap color feature-based image retrieval method. In the method, fast clustering is performed on an image with rich color information to obtain rational statistical distribution centers of each color cluster, and based on the rational statistical distribution centers, features capable of reflecting color differences among different distribution layers of the image are extracted to perform image retrieval. The method comprises the following steps of: first performing meshing on a color space of the queried image, counting the numbers of pixel points in each mesh and selecting the mesh with a number local maximum; then quickly generating each color cluster and the rational statistical distribution centers thereof by adopting a novel distance optimization algorithm and an equal-average nearest neighbor algorithm search (ENNS) algorithm in a K-average clustering algorithm, and on the other hand, performing space sub-block division on the queried image and calculating a Gaussian-weighted color average of sub-blocks; next comparing the color average of the image sub-blocks with the rational statistical distribution centers of the color clusters to extract the features of a K-layer bitmap; and finally performing the matched searching of the image features by combining the similarity measurements of the rational statistical distribution centers of the color clusters and the bitmap.

The invention proposes an image clustering method based on sparse orthogonal bigraph non-negative matrix factorization used for solving the technical problems of low accuracy and slow speed of image clustering in the existing method. The implementation steps are as follows: inputting image data; calculating a data space similarity matrix and a feature space similarity matrix; calculating a data space similarity diagonal matrix and a feature space similarity diagonal matrix; acquiring a label constraint matrix; defining and initializing three sparse orthogonal bigraph non-negative matrix factorization matrixes; setting the number of iterations; acquiring an updating formula of the three sparse orthogonal bigraph non-negative matrix factorization matrixes and an updating formula of the label constraint matrix; defining an updating formula of a coefficient diagonal matrix; updating the three sparse orthogonal bigraph non-negative matrix factorization matrixes, the label constraint matrix and the coefficient diagonal matrix; defining and calculating a low-dimensional data representation matrix; and performing image clustering and output. The image clustering method based on the sparse orthogonal bigraph non-negative matrix factorization provided by the invention can be used for texts, image clustering and face recognition and other practical applications.

The invention discloses a professional field-oriented on-line theme detection method. The method comprises the following steps: obtaining a text vector matrix of a preprocessed text set, and extracting a dictionary from the text set; modeling the text vector matrix; calculating a mixed weight p (thetak|d) from a text d to a theme thetak and a frequency p (w|thetak) that a feature word appears in each theme thetak; obtaining the similarity between two texts di and dj, defining a theme model-based theme distance between the texts into a relative entropy distance of a text vector, and calculating a similarity matrix; compressing the text set, thus obtaining a new text sample sect; calculating a similarity matrix of the new text sample set, and selecting a deviation parameter p according to the similarity matrix; combining clustering results, thus generating a new clustering result; calculating distances between all texts in the original text set and compressed classified texts, and performing classification; outputting a text set theme and a final clustering result. Compared with the prior art, the professional field-oriented on-line theme detection method disclosed by the invention has the advantage that by the adoption of an optimal clustering algorithm, the accuracy and the efficiency of the clustering effect are improved.

The invention discloses a ship path planning method based on a ship trajectory and an ant colony algorithm, a clustering object is changed from a point to a track segment, information of the track segment relative to a trajectory point, a ship course, a ship speed and the like is reserved, and a clustering result is more accurate during clustering. In similarity measurement, similarity measurement calculation is carried out by adopting more multi-dimensional angles, and four dimensions of a horizontal distance, a vertical distance, a steering angle and a speed are considered. Thus, the overall clustering result is more accurate. According to the method, an improved DBSCAN algorithm is used, information such as the speed and the direction contained in a ship track is reserved, and meanwhile the clustering speed is increased. According to the improved ant colony algorithm provided by the invention, experiments show that compared with the previous ant colony algorithm, the ant colony algorithm suitable for ship route planning provided by the invention improves the convergence speed when solving the problem of ship route planning.

The invention provides a method and system for possibly fuzzy K-harmonic means clustering. The method includes the following steps: determining an initial clustering center; setting a parameter value of a clustering algorithm; calculating covariance of sample data; calculating a fuzzy membership value of the possibly fuzzy K-harmonic means clustering; calculating a typical value of the possibly fuzzy K-harmonic means clustering; calculating a clustering center value of the possibly fuzzy K-harmonic means clustering; judging whether an iteration termination condition is met, if on yes judgment, terminating iteration, if on no judgment, performing iteration calculation continuously; and utilizing the fuzzy membership value and the typical value to achieve division of data sets finally. The method and system for the possibly fuzzy K-harmonic means clustering effectively processes data containing noise and can obtain the fuzzy membership value and the typical value. The typical value does not belong to the fuzzy membership value and does not have probability constraint conditions, therefore the method and system for the possibly fuzzy K-harmonic means clustering is insensitive to noise, high in clustering accuracy and rapid in clustering speed.

The invention discloses a deep convolutional neural network model adaptive quantization method, and designs a deep convolutional deep network low-bit quantization algorithm suitable for FPGA calculation, which mainly comprises preprocessing of network model parameters and a grouping adaptive quantization method of a parameter set. the method includes: acquiring dynamic thresholds to perform coarse-grained cutting on the original parameters of the model; constructing an initial clustering center point set suitable for FPGA shift calculation; grouping and clustering the preprocessed model parameters based on a mode length minimization method; finally, overlaying the clustering center point set with the non-null parameter class, achieving self-adaptive low-bit quantization of different networks through optimization; the quantization algorithm is moderate in complexity and quite conforms to the calculation characteristics of the FPGA, hardware resource consumption on the FPGA is reduced, and the model reasoning speed is increased while the model reasoning precision is guaranteed.

The invention discloses a dynamic partition electricity price calculation method based on partition clustering analysis. The method comprises constructing a direct-current optimal power flow optimization target, determining a power system operation constraint condition, solving to obtain optimized hydroelectric output, and then solving node electricity price, conventional unit and new energy unitoutput, power flow data, an active dual factor and node injection power; generating an effective clustering attribute; determining an optimal clustering number; generating a dissimilarity matrix through standardization processing and a Euclidean distance method, and dividing a clustering algorithm to realize clustering; dividing each electricity price area according to the clustering result, and finally determining the unified pricing of each area; feeding back a calculation result of the regional electricity price, and carrying out settlement on market participants in each region according tothe regional electricity price. According to the invention, the electricity price of each subarea can be determined, and finally, each node in the same area is settled in a unified pricing manner, sothat the practical operability is improved, and the demand of obtaining relatively stable price by most users is met.

The invention provides a retrieved result clustering system in a coal mine search engine. The search result clustering system in the coal mine search engine comprises a retrieved result clustering and category label drawing device. The retrieved result clustering and category label drawing device comprises a search engine server, a text retrieved result clustering module and a category label drawing module. The coal mine search engine server processes inquire requests submitted by a user, and a generated initial retrieved result passes through the text retrieved result clustering module and then returns to the user. By the adoption of the retrieved result clustering system, the clustering speed of text sets can be effectively increased, and subjectivity and randomness caused when a similarity calculation method is selected can also be avoided. When data objects are combined into clusters, the similarity relation of the data objects is measured by calculating mutual information loses generated when the data objects are combined, and retrieved result documents can be grouped in a high-quality mode on the basis of the similarity relation.

The invention discloses a rapid night vehicle detection method applied to a self-adaptive high beam, and the method comprises the steps: step 1, enabling an image collection module to collect a road traffic image in front of a vehicle, and transmitting image data information to an image processing module; step 2, processing the image data information by an image processing module, and judging a suspected vehicle lamp region by adopting a grid clustering algorithm; step 3, determining a halo range of a suspected vehicle lamp region by adopting a corrosion algorithm, calculating a halo color through a rapid algorithm, and judging a headlamp and a tail lamp; step 4, conducting pairing according to the geometrical relationship, recognizing vehicles, calculating vehicle coordinate position information and achieving night vehicle detection; and step 5, enabling a data transmission module to transmit the vehicle coordinate information calculated by the image processing module to the high beamcontrol module. The final vehicle lamp information obtained through image processing can serve as the control basis of the self-adaptive high beam lamp and can also provide support for other modulesneeding the vehicle lamp information.

The invention discloses a short-term power load rapid prediction method based on a spark framework. According to the method, two models are trained; a BIRCH parallelization algorithm is used to cluster historical load data and weather data to obtain a model for anomaly detection;, historical load data and weather data are trained through a light GBM algorithm based on the spark technology, a loadprediction model is obtained, and then the two models are sent to a Spark Streaming cluster to be used for clustering and predicting real-time data streams; in clustering and prediction of real-time data streams, a kafka cluster is used to receive power load data streams sent from various terminals; the data flow is transmitted to a Spark Steaming cluster to be processed; real-time feature extraction and normalization processing is completed on a Spark Steaming cluster, real-time clustering is performed by using an anomaly detection model to find out whether abnormal data exists or not, and then a load value of a next time period is predicted by using a load prediction model by using non-abnormal load data.

The embodiment of the invention provides a network news hotspot mining method and device, and the method comprises the steps: carrying out preprocessing of original network news data to acquire network news information; extracting text feature vectors in the network news information through a bilingual LDA topic model and a bilingual LSA model; and according to the text feature vector in the network news information, performing parallel operation on a Spark platform by utilizing a Single-Pass clustering algorithm to obtain news hot topic information. According to a text feature extraction method based on combination of the bilingual LDA model and the bilingual LSA model, entity information with high distinction degree for each topic is contained in the topic model; according to the networknews hot spot mining method, the semantic relation between text contexts is also considered, and the Spark-based parallelization Single-Pass clustering algorithm is utilized, so that the clustering speed is increased, and the network news hot spot mining is more effectively and accurately realized.

The invention discloses a phishing website detection method and system in an Android environment, and the method comprises the following steps: obtaining a website URL when a cellphone user browses a webpage, and transmitting the website URL to a server; obtaining the webpage contents when the server receives the URL transmitted by a client, and constructing a feature vector according to the URL and the webpage contents; carrying out the clustering of K-means through a sample dataset, forming two clustering centers, and training a naive Bayes algorithm; discriminating the feature vector through an algorithm combining the K-means and an improved naive Bayes algorithm, and transmitting a discriminating result to a client user; enabling the client user to give a corresponding prompt according to a returned result, displaying a prompt box for reminding a user if a website is a phishing website, or else browsing the webpage normally. The method makes the most of the advantages that the K-means algorithm is high in clustering speed and the naive Bayes algorithm is high in accuracy, and greatly improves the classification speed and precision.

The invention discloses a base station control method based on an overlay tree. The method comprises the following steps: acquiring location data of a mobile terminal in a target area and building theoverlay tree; selecting a cluster center according to the overlay tree and an expected number of clusters; clustering various nodes according to the distance between other nodes and the cluster center; and finally, adjusting a working parameter of a base station according to a cluster result. According to the method of the invention, clustering is carried out based on the overlay tree; on the onehand, only the distance between various nodes needs to be calculated when the overlay tree is built, and on the other hand, since the structure of the overlay tree itself reflects a distance relationship between the nodes, a new cluster center does not need to be calculated after the cluster center is selected, in consideration of the above two points, the method saves the calculation consumption, speeds up the clustering speed, and improves the service quality of the base station to the mobile terminal. In addition, the invention further provides a base station control device and equipment based on the overlay tree, and a computer readable storage medium, and the functions of which correspond to the above method.

The invention discloses an infrared spectroscopy tea quality identification method mixed with GK clustering in a tea detection technology. A linear discriminant analysis method is used to learn a compressed training sample to acquire a training sample with identification information and a test sample with identification information. Fuzzy C average value clustering is carried out on the training sample with identification information to acquire the initial fuzzy membership degree and an initial clustering center. The fuzzy scattering matrix and the fuzzy membership degree value are calculated, and then a typical value is calculated. The clustering center is calculated according to the typical value. The Euclidean distance from the average value of the training sample with identification information to the clustering center of the test sample is calculated. If the Euclidean distance from the clustering center to the average value of training tea is the minimum, the tea variety of the clustering center and the tea variety of the training sample are the same. The tea and the category of the test sample are determined according to the fuzzy membership degree value. According to the invention, the typical value is added into a function, which can significantly reduce the probability of noise data processing errors.

The invention discloses a tea variety classification method based on fuzzy inter-cluster separation and clustering. The method comprises the following steps: S1, carrying out Fourier near infrared spectrum collection on a tea sample; S2, carrying out pretreatment on the near infrared spectrum of the tea leaf sample by using multivariate scattering correction; S3, realizing near infrared spectrum dimension reduction treatment by using principal component analysis; S4, realizing identification information extraction of the near infrared spectrum data by using linear discriminant analysis; And S5, carrying out tea variety classification by using fuzzy inter-cluster separation clustering. According to the method, the problem that the clustering effect is not ideal when a complex data structureis processed by using traditional fuzzy inter-cluster separation and clustering is solved. The method has the advantages of high detection speed, nondestructive detection, capability of processing complex spectral data, high tea variety classification accuracy and the like.

The invention discloses a large power network graded voltage control method under wind power access. The control method comprises the steps of (1) measuring the power flow calculation data at the input current moment of the system in real time; (2) carrying out online selection on data leading node based on clustering by fast search and find of density peaks; (3) carrying out three-level voltage control optimization calculation according to the current state, and determining an adjusting target value of the voltage of the leading node according to the optimization result and the selected leading node; and (4) sending down the leading node target value to a two-level voltage control system, implementing two-level voltage control, ensuring the voltage of the leading node at the target value level, and maintaining the overall voltage level of the system; judging whether the system runs to a moment corresponding to a next three-level voltage control cycle or not; if so, returning to the step (1); and or otherwise, returning to the step (4). By adoption of the large power network graded voltage control method, it is ensured that the control machine set, the control target and the constraint conditions do not need to be replaced frequently in the process when the target issued by the three-level voltage control is maintained by the two-level voltage control, so that the feasibility of the practical engineering application can be ensured.

The invention discloses a competition and cooperation clustering method based on maximum clearance segmentation of a dynamic bounding box, and provides a method for acquiring an initial seed point by adopting the maximum clearance segmentation of the dynamic bounding box, i.e. firstly calculating a bounding box of data in a multi-dimensional characteristic space, projecting data points in the bounding box towards a longest axis, dividing the bounding box into two parts by finding out positions with a maximum distance of two adjacent projection points, carrying out the recursion until the entire space is segmented into sufficient subspaces, and finally calculating a center of the subspaces to be used as the initial seed point. The invention also provides a method for merging clusters by adopting a distance radius analysis method and capable of self-adaptively combining a plurality of segmented clusters into a complete cluster aiming at the phenomenon that the same cluster is segmented into a plurality of clusters. By adopting the competition and cooperation clustering method, the missing phenomenon caused by the random seed point can be avoided, the clustering segmentation phenomenon can be avoided, and a real cluster result can be rapidly acquired.

The invention discloses a protein family phylogenetic analysis method based on amino acid sequence alignment, which comprises the following steps: obtaining a combined multi-sequence alignment result based on an amino acid sequence alignment fusion method; digitizing a combined multi-sequence comparison result, and constructing a fractional matrix; performing dimension reduction and clustering processing on the fractional matrix to obtain an input sequence; identifying a specific site and a conserved site of the input sequence; performing quasi-time analysis on the input sequence to obtain a track sequence of the input sequence; and obtaining a development trajectory of the input sequence based on the trajectory sorting. According to the method, fractional matrix construction and dimension reduction analysis are carried out through the sequence site features, so that the clustering and evolutionary relationship between gene families is deduced, the sequence gene clustering speed is effectively increased under the condition that the sequence clustering stability is guaranteed, and a new tool and method are provided for gene phylogenetic analysis and development trajectory analysis.

The invention discloses a high-dimension data soft and hard clustering integration method based on a random subspace. The method comprises the following steps of (1) inputting a high-dimension data set; (2) performing data normalization; (3) generating the random subspace; (4) performing kmeans and fuzzy cmeans clustering; (5) generating a fusion matrix; (6) using a clustering validity index to obtain an optimum clustering number; (7) constructing a decision attribute set; (8) improving rough set attribute reduction to obtain a simplified fusion matrix; (9) performing consistency function division; (10) obtaining a clustering purification rate. By using the method provided by the invention, the random subspace is used for solving the problem of processing difficulty of high-dimension data; the combination of soft clustering and hard clustering is used; original data and intermediate result information are sufficiently utilized for performing the intermediate result redundant attribute reduction; the clustering accuracy is improved; meanwhile, the clustering speed is also accelerated; the problems of incapability of sufficiently utilizing clustering information and removing redundant information in the prior art are solved.

The invention relates to the field of electronic evidence collection, in particular to a method for detecting tampering that deleting operation is conducted on the motion foreground of a video under a video static background. The method includes the steps that a difference frame and a block are acquired, and a difference frame delta I of the difference frame and the block is acquired; self-adaptation sparsification is conducted; a measurement matrix theta is selected for sparse measurement; feature vectors acquired through sparse measurement are classified, feature clustering is conducted on categories of the feature vectors through a k-means clustering algorithm; through clustering, the method for detecting video tampering of overcomplete dictionary training based on sparse representation is achieved. The method is visual in detection result, high in robustness and antijamming capacity, accurate in detection result, high in actual application value and small in parameters needing configuring, and brings great convenience to users, and the influences of trees, flowers and plants which swing with wind in a shooting scene can be effectively avoided.

The invention discloses a multilayer bitmap color feature-based image retrieval method. In the method, fast clustering is performed on an image with rich color information to obtain rational statistical distribution centers of each color cluster, and based on the rational statistical distribution centers, features capable of reflecting color differences among different distribution layers of the image are extracted to perform image retrieval. The method comprises the following steps of: first performing meshing on a color space of the queried image, counting the numbers of pixel points in each mesh and selecting the mesh with a number local maximum; then quickly generating each color cluster and the rational statistical distribution centers thereof by adopting a novel distance optimization algorithm and an equal-average nearest neighbor algorithm search (ENNS) algorithm in a K-average clustering algorithm, and on the other hand, performing space sub-block division on the queried imageand calculating a Gaussian-weighted color average of sub-blocks; next comparing the color average of the image sub-blocks with the rational statistical distribution centers of the color clusters to extract the features of a K-layer bitmap; and finally performing the matched searching of the image features by combining the similarity measurements of the rational statistical distribution centers ofthe color clusters and the bitmap.

The embodiment of the invention discloses an incremental clustering method and device, an electronic device and a computer readable medium. An embodiment of the method comprises the following steps ofclustering an original data set by utilizing a density-based clustering algorithm with noise, and determining a cluster in the original data set; and in response to detected update of the original data set, performing incremental clustering on a new data set obtained through updating based on the update type of the original data set so as to update the cluster in the original data set. Accordingto the embodiment, the clustering efficiency is improved.

The invention relates to the field of load prediction of power systems, in particular to a power consumer characteristic analysis method and system based on BEMD and kmeans. The method comprises the steps of firstly obtaining user power load data and storing the data as a database; preprocessing the power load data based on an empirical mode decomposition method; carrying out kmeans algorithm clustering on the hierarchical power load data, and selecting a Pearson distance as an evaluation index of a sample similarity degree; According to the kmeans clustering result and the actual user power utilization characteristics, characteristic analysis is carried out for different time intervals. The analysis method can adapt to time interval load data, is also suitable for loads with high volatility and poor stability, can realize hierarchical clustering of power consumer power utilization characteristics, and has a stable and good clustering effect on the premise of considering the operationspeed.

The invention relates to a weak and small moving target detection method based on superpixel adjacent frame feature comparison. The method comprises the following steps of: 1, performing super-pixel segmentation on a target frame image by using an SLIC (simple linear iterative clustering) algorithm; 2, generating a graph theory model according to the adjacency relation of the superpixels, designing superpixel features, calculating feature differences among the superpixels, taking the feature differences as edge values in the graph theory model, and realizing potential target extraction by utilizing a graph segmentation algorithm; and 3, performing adjacent frame color feature comparison on the potential target area, and if the color feature difference exceeds a set threshold, marking the potential target area as a moving target. According to the method, inter-frame search is avoided, and the efficiency of detecting the weak and small moving target is improved.