67 results about "Connected component" patented technology

3D models often have a large number of small to medium sized connected components, with small numbers of large triangles, often with arbitrary connectivity. The efficiency of compact representation of large multi-component 3D models can be improved by detecting and representing similarities between components thereof, even if the components are not exactly equal. The invention uses displacement maps for encoding two or more different but similar geometry patterns differentially, based on clustering and a cluster representative surface. A method for encoding a plurality of geometry patterns comprises detecting and encoding identical copies of geometrical patterns, detecting and clustering similar geometry patterns, and detecting partial similarity. The detecting partial similarity comprises generating a cluster representative surface, generating for at least one clustered geometry pattern a displacement map, and encoding the common surface and the displacement maps.

In a text line segmentation process, connected components (CCs) in document image are categorized into three subsets (normal, large, small) based on their sizes. The centroids of the normal size CCs are used to perform line detection using Hough transform. Among the detected candidate lines, those with line bounding box heights greater than a certain height are removed. For each normal size CC, if its bounding box does not overlap the bounting box of any line with an overlap area greater than a predefined fraction of the CC bounding box, a new line is added for this CC, which passes through the centroid of the CC and has an average slant angle. Each large size CCs are broken into two or more CCs. All CCs are then assigned to the nearest lines. A refinement method is also described, which can take any text line segmentation result and refine it.

The invention discloses an automatic extraction method of data in a graph. The method mainly comprises the following steps: performing text region positioning and textbox classification; performing coordinate axis positioning and scale mark position positioning of the coordinate axis; performing legend positioning and information extraction; extracting a corresponding cylinder or broken line connected component according to the legend color, and filtering and classifying; determining X-axis key points and positioning an X-axis label corresponding to each key point; according to the X-axis keypoint positioning column body and the key point of the broken line, determining a marked numerical textbox corresponding to the key point, and carrying out numerical identification; calculating a numerical value corresponding to each pixel, and estimating a numerical value corresponding to the cylinder or broken line key point; and determining a final result according to the difference between theestimated value and the identified annotation value. The method can cope with various data graph types such as a horizontally-arranged histogram, a broken line and cylinder mixed data graph, a data graph without coordinate axes and the like. Experiments prove that the method has high accuracy and precision.

The embodiment of the invention discloses a text row arrangement analytical method and device for text area recognition. The text row arrangement analytical method for text area recognition comprises the steps that any two of individual-character areas are matched to form a candidate pair, wherein the individual-character areas are extracted from an image, and the two individual-character areas of each candidate pair have the most similar characteristic; candidate pairs, meeting preset aggregation conditions, in the candidate pairs formed through matching are aggregated to form a candidate set; the candidate set is set up to be a connected component, a maximum clique is searched in the connected component, and rows formed by the maximum clique serve as rows to be selected; the initial direction of a baseline and the initial direction of an apical line of the rows to be selected are calculated; mid-value separation distances of the individual-character areas in the rows to be selected are calculated according to the direction of the initial row of the rows to be selected, and the row, to be selected, with the minimum mid-value separation distance is selected to serve as a text row; the candidate pair contained in the text row is deleted from the candidate set, and a new text row is continuously selected out. According to the text row arrangement analytical method and device for text area recognition, the text row with the small number of characters can be accurately extracted from a manually-edited image, and accuracy of character recognition is improved.

The present invention relates to the field of image recognition, and for the problems existing in the prior art, provides a real-time train number recognition method. The method comprises: a step of initial positioning of a train number region: calculating an optimal gray level through a local binary model on input image data, finally obtaining a whole gray level value corresponding to the optimalgray level, and according to the gray level value, positioning the rectangular coordinate containing train number characters from the input image by using a linear two-class classifier; a step of character segmentation: using a dynamic binary method and a connected component analysis algorithm to divide the rectangular coordinate region containing the train number characters into individual characters; and a step of character recognition: performing character recognition on the train number according to the SVM classifier.

The invention provides a graph database-based power grid equipment reachability query method and system. The power grid equipment reachability query method comprises the steps of building a graph model for a physical connection of power grid equipment according to a topological connection relationship between the power grid equipment, and obtaining an initial connected component index tree; according to the initial connected component index tree, searching for a most recent common ancestor node between two pieces of the power grid equipment, and judging whether the most recent common ancestornode is a non root node or not; and when the most recent common ancestor node is the non root node, judging that reachability between the two pieces of the power grid equtipmen is realized. Accordingto the power grid equipment reachability query method designed by the invention, a reachability query function of the power grid equipment is realized; and the efficiency of power grid equipment reachability query is improved.

As set forth herein, a computer-implemented method facilitates pre-analyzing an image and automatically suggesting to the user the most suitable regions within an image for text-based personalization. Image regions that are spatially smooth and regions with existing text (e.g. signage, banners, etc.) are primary candidates for personalization. This gives rise to two sets of corresponding algorithms: one for identifying smooth areas, and one for locating text regions. Smooth regions are found by dividing the image into blocks and applying an iterative combining strategy, and those regions satisfying certain spatial properties (e.g. size, position, shape of the boundary) are retained as promising candidates. In one embodiment, connected component analysis is performed on the image for locating text regions. Finally, based on the smooth and text regions found in the image, several alternative approaches are described herein to derive an overall metric for “suitability for personalization.”

A non-invasive scattering imaging method beyond a memory effect range based on connected component optimization, including: calculating an autocorrelation image of collected speckle information, and obtaining an autocorrelation graph according to the autocorrelation image; obtaining a reconstruction result according to the autocorrelation graph, and optimizing the reconstruction result; calculating and normalizing autocorrelation of the optimized reconstruction result, to calculate initial autocorrelation of an object; obtaining a reconstruction result according to the autocorrelation of the object obtained in the previous step, calculating autocorrelation of an optimized reconstruction result by using a connected component and performing normalization, and using the autocorrelation graph to calculate autocorrelation of another object; and repeating the preceding steps of calculating autocorrelation of the two objects until a predetermined loop count is reached, and then using a phase recovery algorithm and the calculated autocorrelation of the two objects to perform spatial reconstruction to achieve non-invasive imaging.

The invention provides an image processor-combined algorithm for speed restriction board detection during automatic driving. An image with the resolution of 1600*1200 is reduced to an image with the resolution of 1200*900, and a red image with a similar red color as a speed restriction board is obtained by means of CUDA C language on a CPU and is given back to the CPU. A connected-component analysis algorithm is used to obtain information of each connected component, such as the position. Then, on the original color image with the resolution of 1600*1200, a judgment is made whether each connected component is a speed restriction board or not, by means of a support vector machine classifier. Finally, positions and pixel sizes of all speed restriction boards in the original image with the resolution of 1600*1200 are obtained. The invention proposes that a part of steps of the detection algorithm, where parallel computing is easy, are executed on the CPU, and in combination with calculation advantages of a central processing unit and an image processor, serial and parallel computing is carried out at the same time, thereby improving the speed restriction board detection rate and the speed restriction board detection speed. The algorithm of the invention has the advantages of high running frame rate, short development cycle, easy understanding and convenient maintenance.

Embodiments of the invention provide a data association method and device, and an electronic device. The data association method comprises the following steps: acquiring transaction data in a predetermined time period; constructing a directed graph corresponding to the predetermined time period according to the transaction data of the predetermined time period; obtaining a maximum communication sub-graph of the directed graph; and taking the maximum communication sub-graph as a connected component of the directed graph, obtaining transaction data belonging to the same connected component in the predetermined time period, and finally marking the transaction data belonging to the same connected component in the predetermined time period to identify a fund flow direction corresponding to thetransaction data.

The invention provides a multi-view subspace clustering method for self-weighted fusion of local and global information. The method comprises the following steps: firstly, preprocessing acquired multi-view data; secondly, when local information of the original multi-view data is mined through graph learning, adding self-representation learning to mine global information of the original multi-view data, so that a high-quality and high-robustness similarity matrix is obtained; and then fusing the similarity matrixes of all views in a self-weighting mode to form a consistent similarity matrix. Rank constraint is introduced into the consistent similarity matrix, so that the number of connected components in the consistent similarity matrix is equal to the number of clustering clusters, and a multi-view clustering result is directly obtained. In this way, it is avoided that after the consistent similarity matrix is obtained, additional clustering steps need to be executed to obtain the clustering result. According to the method provided by the invention, the clustering structure of the data is disclosed while the consistent similarity matrix is generated, and an additional clustering step is not required to be executed to obtain a clustering result. Therefore, the method provided by the invention has a good clustering effect on the multi-view data.

The invention discloses a super computer benchmark test acceleration method based on connected component generation optimization, and aims to minimize a communication path, maximize a memory access bandwidth utilization rate and accelerate a super computer big data benchmark test. According to the technical scheme, by means of the characteristic that a graph generated by Graph500 contains a plurality of connected components, the connected components are quickly found in the graph, the connected components are stored through a two-dimensional vector, path compression is conducted on the set membership of vertexes in the connected components, two connected components with different root vertexes are combined. And the vertexes of the same connected component are divided to physical nodes with shorter communication paths in the super computer, so that the communication overhead is small and the operation speed is high during graph traversal access. All connected components in the graph can be effectively and quickly stored, the merging speed is furthest improved, the query speed of the root vertex is accelerated, the occupation overhead of a stack in a memory is reduced, and the test speed of the big data processing capability of a super computer is improved.

The invention provides a system for measuring an artery input function in dynamic perfusion image post-processing. The system comprises: a preprocessing module, which is used for preprocessing original time density curves, and screening out curves with regular shapes as an initial curve set; a clustering analysis module, which is used for performing clustering analysis on the initial curve set, and selecting one type of curves with a minimum clustering center as candidate curves; a post-processing module, which is used for marking voxel positions corresponding to the candidate curves in a three-dimensional space, calculating connected components of the space and a number of voxels of each component, removing components smaller than a set volume, and calculating an average curve of each remaining spatial component to obtain target curves; a classification module, which is used for extracting characteristic parameters of the target curves, calculating the conformity of the target curves by using a linear weighting model, and screening out a final AIF curve according to the conformity. The system can efficiently and stably obtain an AIF curve.

The invention discloses a blood vessel center line determining method and a device and a readable storage medium. The method comprises the following steps: acquiring a blood vessel image; extracting afirst center line of each blood vessel in the blood vessel image, wherein the first center line comprises a first starting point, an intersection point and a segmentation line; constructing an undirected graph according to the first starting point, the intersection point and the segmentation line; determining each connected component of the undirected graph; determining each sub-graph in each connected component, wherein the sub-graph comprises at least one vertex, and when the sub-graph comprises a plurality of vertices, at least two connection paths existing between any two vertices in thesub-graph; for each connected component, constructing a spanning tree by taking the sub-graph comprising the first starting point as a root node, the sub-graph not comprising the first starting pointas a sub-node and a leaf node and a connecting line between the sub-graphs as an edge; obtaining at least one designated point selected by the user for the blood vessel to be named; and when the specified point does not include the second starting point, determining a first path from the root node to the specified point in the spanning tree to obtain a second center line of the blood vessel to benamed.

The embodiment of the invention provides an account fund transfer network diagram coloring method and device, and the method comprises the steps: respectively setting the identification of each node in a target account fund transfer network diagram as a natural number starting from zero, and determining the connection component to which each node belongs; performing initialization coloring processing on the connected components and a first preset number of nodes by applying a self-defined color number; coloring the nodes which are not colored at present; selecting one of the adjacent nodes with the same coloring color to delete the color; returning to execute the process of coloring the currently uncolored nodes until the execution frequency of the process of coloring the currently uncolored nodes is equal to or greater than a preset iterative threshold, and outputting a node coloring graph corresponding to the target account fund transfer network graph and used for dividing account node communities. User-defined coloring control over the account fund transfer network diagram can be achieved, and errors can be controlled within a certain range.

The invention relates to the technical field of network security, in particular to an APT attack detection method based on an event relation directed graph in network full traffic, which comprises the following steps of: S1, acquiring bypass traffic by using deep packet analysis to obtain real-time network metadata; S2, analyzing the network metadata by using big data based on a hacker network technology, and generating a network event relation directed graph; and S3, constructing a network event attack technology directed graph based on a preset APT attack technology weighted directed graph, performing matching analysis on the network event attack technology directed graph and a preset APT attack technology weighted directed graph, calculating a connected component weight of the network event directed graph, and finally detecting a current APT attack result and generating an alarm prompt. According to the technical scheme, the APT attack detection method based on the event relation digraph in the network full traffic is comprehensive and reliable in data analysis, higher in detection accuracy, capable of saving storage resources and convenient for backtracking, querying and analyzing historical data.

The invention provides a parameterization-free clustering algorithm and system based on a minimum spanning tree, and the method comprises the following steps: firstly abstracting a to-be-clustered data set into an empowerment complete graph WCG, where points represent vectors, and empowerment edges represent the similarity relation between data; converting the WCG into a fully connected minimum spanning tree MST; clustering the one-dimensional weight space of the MST edge set by using a K-means algorithm in which k = 2 to obtain a pruning threshold; and finally, performing pruning and noise filtering on the MST, and obtaining a connected component being a clustered cluster. According to the algorithm, an original high-complexity uncertain-category-number dimensional space clustering problem is converted into a low-complexity two-category-number one-dimensional space clustering problem; according to the method, the defects of the K-means algorithm are overcome, non-parameterized clustering is truly realized, the clustering efficiency is improved while the clustering time is shortened, and the dependence of the algorithm on empirical parameters is eliminated.