48 results about "Tree clustering" patented technology

The invention discloses an internet of things-based intelligent building monitoring managing system, which mainly realizes remote monitoring and managing on an intelligent building. The system comprises four parts of a building equipment managing system, a firefighting linkage alarm system, a positioning navigation system and a security system; the building equipment managing system realizes concentrated managing on building equipment based on a tree cluster network established by an intelligent sensor module; the firefighting linkage alarm system adopts an optimized A*algorithm to search for an optimal fire escape path; the positioning navigation system is achieved by a mobile navigation robot covering floors of the building, and if a user inputs a destination, the navigation robot directly inserts virtual scene information into a reality scene, and displays a direction mark; the security system is realized by a mobile patrolling robot based on a GPS/DR combined navigation technology, and an intruder is alarmed after being found, the current intelligentized managing requirements of the building are met and the system is suitable for a high-rise building such as a commercial building with a large scale.

According to one aspect of the invention, a method is provided in which a set of multiple mixture monophone models is created and trained to generate a set of multiple mixture context dependent models. A set of single mixture triphone models is created and trained to generate a set of context dependent models. Corresponding states of the triphone models are clustered to obtain a set of tied states based on a decision tree clustering process. Parameters of the context dependent models are estimated using a data dependent maximum a posteriori (MAP) adaptation method in which parameters of the tied states of the context dependent models are derived by adapting corresponding parameters of the context independent models using the training data associated with the respective tied states.

The invention provides a single-plant tree point cloud automatic extraction method based on vehicle-mounted laser scanning data, and the method mainly comprises two big steps: obtaining point clusters, called as a candidate tree clusters, from the vehicle-mounted laser scanning data based on the rough classification of cluster features, wherein the point clusters may comprise trees; extracting a single-plant tree point cloud from each candidate tree cluster based on two-stage voxel regional growth. The method combines the overall and local geometrical characteristics of trees, can automatically recognize the seeds of single-plant trees in a complex environment, respectively sets different growth rules for the growing of tree trunks and crowns, and achieves the extraction of the single-plant trees. Meanwhile, compared with other single-plant tree extraction algorithm, the method achieves greater improvement on the extraction integrity and accuracy of trees, the extraction precision of a single-plant tree point cloud and the extraction in the complex environment, and can be used for further information extraction of trees.

The invention discloses a complex network community mining method based on an improved genetic algorithm and belongs to the technical field of complex network community mining method research. The complex network community mining method based on the improved genetic algorithm uses the improved genetic algorithm based on clustering and double population thought fusion to mine communities in a complex network. The complex network community mining method based on the improved genetic algorithm uses a normalization common information similarity standard as the standard for measuring the similarity between individuals in the population and fuses the clustering and double population thought. The complex network community mining method based on the improved genetic algorithm includes that introducing the clustering thought, using a minimum spanning tree clustering method to classify the population, introducing the double population thought, and determining the main type and auxiliary type for the clustering. The main type maintains the population evolution direction to get close to the optimal solution of an objective function; the auxiliary type is mainly used for duly providing diversity for the main type so as to enable the main type to be capable of coming out to search the other solution space to realize the complex network community mining when the main type is located at the local optimum.

The invention discloses a method for automatically extracting point clouds of an electric tower from airborne LiDAR data. The electric tower is important contents in high-voltage line patrol. Based on the plane position and spatial geometric characteristics of the electric tower, a method for extracting the electric tower from point clouds of an airborne laser radar is proposed and comprises the steps of (1) performing electric tower point cloud coarse extraction on original point cloud data, obtained by the airborne laser radar, of a power transmission line according to a two-dimensional grid neighborhood clustering method; (2) preprocessing a coarse extraction result by adopting Kd-tree clustering and spatial grid region growth methods; (3) through a spatial geometric structure of the electric tower, extracting a main region of the electric tower, and determining a tower body edge line equation in combination with an RANSAC spatial linear fitting method; and (4) removing noisy points based on growth of a point cloud model of a main region of a line tower, and removing bottom noisy points of the electric tower by adopting a specific method in combination with different bottom structures of the electric tower, thereby finishing fine extraction of the point clouds of the electric tower. According to the method, fine classification of the point clouds of the electric tower is realized directly through structural characteristics of the point clouds of the electric tower; the problem of relatively poor quality of point cloud data of the electric tower can be solved to a certain extent; and the classification efficiency and precision are high.

The invention discloses a mobile Sink data collection method for a wireless sensor network based on a tree cluster. The method comprises the steps of establishing a distributed tree cluster; calculating node weights by taking surrounding energy and density information of nodes into comprehensive consideration; selecting suitable root nodes as resident points RPs of a mobile Sink through distributed information exchange of a local range; accessing all RPs by the mobile Sink for data collection in a stable data collection stage; providing a new conditional re-clustering method; setting data collection number of rounds; and selecting new RPs for alternation. According to the method, the hotspot problem can be effective mitigated; the network load can be balanced; the energy consumption of nodes can be reduced; and the network lifetime can be prolonged.

The invention relates to an energy-saving dormancy awakening method based on a ZigBee technology. The method comprises the following steps that: S1, a coordinator node establishes a tree cluster type network; and a terminal node and a routing node are added into the network established by a coordinator; S2, the network is graded; time synchronization of a network node and the previous stage is carried out stage by stage; and finally, the whole-network time synchronization with the coordinator is realized; S3, the terminal node and the routing node enter a working period; the terminal node performs acquisition and transmission of data; and the routing node forwards the data to the previous-stage node; S4, after receiving the data uploaded by the terminal node, the coordinator sends dormancy control information to the routing node; and the routing node forwards the dormancy control information to the terminal node; S5, the node receives a dormancy instruction, sets a dormancy time control parameter and enters a dormancy state; and S6, when the dormancy time is up, the node is synchronously awakened to add the network and then enters the working state. By means of the method, the energy consumption of the system can be reduced; the energy utilization rate can be increased; the use time of the system is prolonged; and simultaneously, the maintenance cost of the network can also be reduced.

The invention provides a suffix tree clustering method on basis of word segmentation and part-of-speech analysis. The method involves three parts, namely a document word segmenting module, a part-of-speech analyzing module and a suffix tree clustering module, accomplishes word segmentation, part-of-speech tagging, word weight calculation and essence extraction to a document, can realize dimension reduction process to an original document, reduces the complexity of the processing procedure of the suffix tree clustering method, and meanwhile, ensures the clustering accuracy.

Systems and methods for training a region clustering forest include receiving a set of medical training images for a population of patients. A set of image patches is extracted from each image in the set of medical training images. A plurality of region clustering trees are generated each minimizing a loss function based on respective randomly selected subsets of the set of image patches to train the region clustering forest. Each of the plurality of region clustering trees cluster image patches at a plurality of leaf nodes and the loss function measures a compactness of the cluster of image patches at each leaf node in each of the plurality of region clustering trees.

The invention discloses a tower inclination early warning method based on airborne liDAR point cloud data. The tower inclination early warning method based on the airborne liDAR point cloud data comprises the steps that the tower point cloud data scanned by an airborne lidar based on Kd-tree clustering and spatial grid growth is preprocessed to remove noise points; secondly, point clouds in a tower trunk area are extracted based on tower geometric features; then angular point coordinates of the point cloud data of each layer in the tower trunk area are calculated based on a coordinate system rotation method again; finally, center point coordinates of each layer are calculated by using the angular point coordinates and a space linear equation is fitted to obtain the inclination radian and inclination angle of a tower; and safety warning is given to the tower with the inclination angle exceeding a certain safety threshold value. The defect that traditional manual detection is not suitable for efficient detection in a complex environment is overcome, the inclination and inclination radian of the tower can be automatically obtained based on the airborne LiDAR tower point cloud data, high measurement efficiency and high precision are achieved, and early warning can be timely given to the tower with serious inclination to prevent more serious consequences in the later period.

The invention relates to a clustering multi-hop routing method based on a maximum and minimum distance method. A maximum and minimum distance method is adopted to select a cluster center, re-clustering is carried out according to the cluster center, the selection of a cluster head is carried out according to a maximum weight principle, a new cluster is formed according to distance between a node and the cluster head, the insufficiency of random selection of a LEACH (Low Energy Adaptive Clustering Hierarchy) protocol cluster head is overcome, so that the energy consumption of a network is evenly consumed on each node. A link with a shortest communication path is generated between the cluster head and Sink by adopting a continuous Hopfield neural network, the link is optimized to form a multi-hop tree cluster type link which takes the Sink as a center, and energy consumption generated by long-distance communication caused by adopting a single-hop way is lowered. The deficiencies of the cluster head selection and a node single-hop mechanism in the LEACH protocol under the energy heterogeneous environment of a wireless sensor network can be eliminated, and aspects on network stable phase prolonging and energy balance are obviously improved than a LEACH algorithm.

The invention discloses a robot positioning and navigation method based on a bag-of-words tree cluster model. The method comprises the following steps: 1) inputting a long-time video stream containing a low dynamic indoor scene change; 2) performing off-line feature extraction on an image of the video stream; 3) clustering the obtained features through a k-means++ algorithm; 4) performing iterative clustering on the subspace obtained after clustering; 5) building a bag-of-words tree model for features after iterative clustering; 6) building a bag-of-words tree cluster model; 7) summarizing the motion law of a low dynamic object through a statistical principle; 8) analyzing and determining a bag-of-words tree model corresponding to the current moment based on the motion law according to the bag-of-words tree cluster model; and 9) using the determined bag-of-words tree model for loopback detection to realize robot positioning and navigation in a low dynamic environment. According to the invention, long-time self positioning and navigation of a robot can be realized in a low dynamic environment such as indoor space and the like at low cost under the condition of avoiding the use of expensive laser sensors.

The invention provides a wireless sensor network routing protocol based on improved tree clustering; in a wireless sensor network, a cluster head is elected out and clustered; a tree structure is constructed in the cluster; with the cluster head as a root node, each sensor node in the cluster directly communicates with the cluster head if being close to the cluster head; otherwise, other sensor node relatively close to the cluster head is selected to communicate with the cluster head; with the cluster head closest to an aggregation node as the root node, all cluster heads form a tree, and theroot node directly communicates with the movable aggregation node; after every sensor node is elected out to be the cluster head for one time, a new round of clustering is started for the whole network; and residual energy of the sensor nodes is considered when electing the cluster head. According to the routing protocol, the sensor nodes are prevented from premature death, and the service life ofthe wireless sensor network is prolonged.

The invention claims an image retrieval method based on tree cluster vector quantization. The method comprises the following steps: S1, preprocessing a picture and extracting vector characteristics ofthe picture; S2, clustering the vectors of the images; S3, dividing the data space by using recursive clustering; S4, stopping clustering according to the formulated rules; And S5, storing the path fingerprint of the picture and the tree model. According to the method, the high-dimensional feature vector of the image is extracted by using the feature extraction capability of the deep learning model based on the tree clustering vector quantization algorithm, the content and semantics of the image are fully expressed, and the search capability of the tree clustering algorithm is combined, so that the image retrieval accuracy and the retrieval speed are improved.

The invention relates to an infrared multi-target segmentation method based on minimum tree clustering. The method is characterized in that an image is preprocessed by the adoption of a background differencing method and morphological operation, and a binary image with a target area and a background area separated is obtained; uniform sampling and nearest neighbor classification are carried out on all pixel points in the target area in the binary image so as to initially classify the pixels of the target area into subclasses; the minimum tree is constructed with the subclasses as top points and with the space, obtained by calculation through an L1 norm and a Gaussian kernel function, between the subclasses as the side length; the obtained side length threshold is calculated, the long side of the minimum tree is cut, a final clustering result is obtained, isolated points and noise points are removed, and each residue class serves as a target; all target centroids are marked through crosses in an original infrared image, and all target special ranges are marked through rectangular boxes.

The invention discloses an author disambiguation device based on subject tree clustering. The author disambiguation device comprises the following steps: obtaining text data with authors; processing the text data to extract key information; extracting representative words according to the text data; sequentially matching the text data with each subject node of the subject tree based on the representative words and the key information of the text data, and calculating a matching degree; and selecting the subject node with the highest matching degree, connecting the subject node with the text data, and taking the author with the same name under the subject node as the author. According to the method, the subject tree is constructed, clustering calculation is carried out on the basis of the subject tree to eliminate the problem of text data classification errors under the condition of different persons with the same name, the accuracy of author recognition of the text data is improved, then the text retrieval quality is improved, and an effective auxiliary analysis means is provided for computer semantic analysis.

The invention discloses a clustering algorithm on the basis of minimum spanning trees and initial clustering centers. The clustering algorithm includes steps of S1, inputting to-be-clustered data setsD and category numbers K of the data sets D; S2, creating the minimum spanning trees T<D> of the data sets and computing geodesic distances between optional two nodes in each minimum spanning tree T<D>; S3, selecting the initial clustering centers; S4, creating the minimum spanning trees T of the initial clustering centers; S5, disconnecting K-1 edges of each minimum spanning tree T<D> of thecorresponding data set. The clustering algorithm has the advantages that minimum spanning tree clustering algorithms are combined with selection of the initial clustering centers, the to-be-deleted edges are constrained on paths between the initial clustering centers and can be accurately found by the aid of comprehensive values of the density and the distances of the upper edges for forming the paths, accordingly, the clustering accuracy can be further improved, and interference of noise in the data sets can be prevented.