90 results about "Cut" patented technology

Disclosed is a method of segmenting one or more objects from one or more backgrounds in an image, the method comprising defining a plurality of image nodes, each said image node corresponding to one or more pixels of said image, connecting pairs of adjacent nodes with n-links, each said n-link weighted with an n-link cost, defining a source node, defining a sink node, defining one or more object seeds, said object seeds corresponding to image nodes within said objects, defining one or more background seeds, said background seeds corresponding to image nodes within said backgrounds, connecting said source node with each said object seed with a plurality of t-links, connecting said sink node with each said background seed with a plurality of t-links, wherein each said t-links is weighted with a t-link cost, and calculating a segmentation cut having the smallest total cost of all cuts separating said source from said sink, wherein said total cost of each said cut is defined as the sum of the costs of all said n-links and t-links that each said cut severs.

The invention discloses a depth image estimating method of a binocular stereo video, which comprises the following steps of: (1) collecting a binocular stereo video image from a binocular stereo camera; (2) performing parallax estimation based on graph cut stereo matching on the binocular stereo video image; (3) performing uniformity check on a parallax image obtained by the graph cut stereo matching, and deleting the unreliable matching by self-adapting matching to reduce the wrong matching of the depth image; (4) converting the parallax image into a depth image according to the relation of the parallax and the depth; (5) performing correction optimization on the obtained depth image by a multilateral filter; and (6) outputting the depth image to finish the depth image estimation of the binocular stereo video. The method disclosed by the invention has the advantages of effectively eliminating the errors in the depth image estimation, and finally obtaining an accurate and dense depth image so as to satisfy the demand of rebuilding image quality based on a real scene.

A visual target tracking method of a full convolution class and regression twin network structure comprises the following steps: (1) according to the position of a target in an image, cutting a targettemplate image and a search area image in an original training set, and forming a training data set by cut image pairs; (2) establishing a full convolution twin network to extract image features; (3)establishing a classification regression network; (4) in response to the fact that each pixel point on the image has a corresponding foreground score and a predicted bounding box, calculating the total score of each pixel point by combining the information of the foreground score and the information of the bounding box, wherein the pixel point with the highest total score is the center of the tracking target; and (5) training the full convolution twin network and the classification regression network by using the training data set to obtain the trained full convolution twin network and the classification regression network, calculating a score graph of a target in the to-be-tested image sequence by using the trained networks, and performing target positioning based on the score graph. According to the invention, the tracking precision and speed are improved.

A method for automatically decomposing a shape of an IC design layout into two or more shapes in order to resolve a double patterning loop violation involving the shape. The method decomposes the shape by introducing one or more splicing graphs on the shape. These splicing graphs serve as cuts to be made on the shape. By decomposing the shape into several shapes and assigning the shapes to alternating masks for the same layer, the method breaks the double patterning loop. That is, no pair of the shape and other shapes that form the loop will be assigned to the same color for a mask after the shape is decomposed. In some embodiments, the method introduces splicing points to more than one shape of the loop-forming shapes when necessary. Some embodiments minimize the number of splicing points introduced to the shape(s).

The invention designs a deformable convolution hybrid task cascade semantic segmentation method based on embedded balance, which is used for realizing image target recognition and semantic segmentation, and comprises the following steps: inputting a cut image into a pre-trained neural network; mapping the two samples to the same scale space through a feature pyramid network; performing informationfusion on semantic features extracted from different hierarchies; predicting a pixel-level segmentation result by adopting a convolution layer; performing feature extraction on the input image by adopting a deformable convolutional neural network at the convolution and pooling part of the feature pyramid network to obtain a feature map; dividing the feature map into parts with the same size; inputting a feature map obtained after passing through the feature pyramid network into a regional candidate network for training the network; wherein the region candidate network comprises a target detection classifier and a candidate frame positioning classifier, the target detection classifier outputs a target recognition result and prediction accuracy, and the candidate frame positioning classifier can provide accurate positioning for candidate regions and output candidate frames of a plurality of candidate regions. According to the method, the semantic segmentation positioning accuracy and the segmentation accuracy are improved.

A method of solving an energy minimization problem, the method comprising: a. constructing a graph representative comprising a set of nodes; two terminals; a set of N-links each connecting a pair of the nodes; and a set of T-links each connecting one of the terminals with one of the nodes; b. assigning a capacity to each of the N-links; c. assigning a capacity to each of the T-links; d. determining a first minimum cut / maximum flow solution which partitions the nodes into subsets, each subset containing one of the terminals; e. changing the capacity assigned to at least one of the N-links and at least one of the T-links in response to a change in the problem; and f. dynamically updating the first minimum cut / maximum flow solution determined in step d. to take into account the changed capacities.

Methods, and program storage devices, for performing model-based optical proximity correction by providing a region of interest (ROI) having an interaction distance and locating at least one polygon within the ROI. A cut line of sample points representative of a set of vertices, or plurality of cut lines, are generated within the ROI across at least one lateral edge of the polygon(s). An angular position, and first and second portions of the cut line residing on opposing sides of an intersection between the cut line and the lateral edge of the polygon are determined, followed by generating a new ROI by extending the original ROI beyond its interaction distance based on such angular position, and first and second portions of the cut line. In this manner, a variety of new ROIs may be generated, in a variety of different directions, to ultimately correct for optical proximity.

The invention provides a high energy beam processing method with an ejection point as a control target. The high energy beam processing method with the ejection point as the control target comprises the following steps of (1) vectorizing cut graphs and dividing the cut graphs into line segment units; (2) designing a pitching-in line and a pitching-out line; (3) designing a planned route of the ejection point moving on the lower surface of a processed workpiece; (4) designing a cutting scheme; (5) calculating multi-dimensional space tracks of the ejection point moving on the lower surface of the processed workpiece; and (6) processing the workpiece according to a multi-dimensional space of the ejection point. The high energy beam processing method with the ejection point as the control target selects an optimized strategy (the cutting scheme) according to the cutting route tracks of the ejection point moving on the lower surface of the processed workpiece and can effectively eliminate shape errors caused by different flow beam expression forms of the ejection point and an incidence point. Compared with the prior art, the high energy beam processing method with the ejection point as the control target adopts a cutting head backward inclining and quick cutting method to replace an original cutting-speed slowdown method, improves the cutting efficiency and avoids the shape errors caused by slowdown in the existing method.

The invention discloses a repeated image recognition method and device, equipment and a medium. The method comprises the following steps of: intercepting a cut image from the to-be-identified image; converting the cut image to obtain a grayscale image, a hue saturation image and a color video image; obtaining all normalized values of a local phase quantization characteristic histogram of the grayscale image through a local phase quantization method; obtaining all standard values of a first adjacent local binary pattern feature histogram of the hue saturation image through an adjacent local binary pattern method; and meanwhile, obtaining all standard values of a second adjacent local binary pattern feature histogram of the color video image through an adjacent local binary pattern method, performing sorting according to a mantissa zero padding method to obtain a matrix graph, inputting the matrix graph into a deep convolutional neural network model, and extracting texture signal frequency statistical features to obtain an identification result. According to the invention, whether the certificate or the image is copied or not can be quickly, effectively and conveniently recognized, and the recognition accuracy is improved.

The invention discloses an image foreground extraction method based on a graph cut model. The method is used for accurate segmentation of a human eye area in an image and belongs to the field of image recognition based on machine vision, and the method is suitable for automatic detection driving fatigue state of a driver. According to the invention, the problem that a present detection system cannot accurately recognize human eye opening and closing states in a complex driving environment is solved. According to the method, a graph cut model is firstly established on an original image; then, based on a segmentation target, an energy function of the graph cut model is constructed; and finally, the function is solved by a max-flow min-cut method so as to obtain the final segmentation result. By the segmentation algorithm, a binary image with a white background and a black eye area is constructed, thus providing stable feature description for follow-up human eye state recognition. It proves validity of the extraction method through experiments. The method provided by the invention can be widely applied in automatic detection of fatigue state of drivers of trains, interurban coaches and ''three danger'' vehicles.

The invention relates to a three-dimensional geologic model real-time sectioning and rendering processing method and a storage medium, and the method comprises the following steps: in a three-dimensional model shader, dividing a primitive into a reserved primitive and a cut primitive through a separation plane, abandoning the cut primitive in a rendering process, and only drawing the reserved primitive; processing a graph to be segmented into an entity closed model, specifically, a section line is drawn by means of a cache unit, so that a cutting plane in a specific range is obtained and closed; and finally, drawing the model twice, namely drawing the abandoned cut primitive model for one time and drawing the contour pixels of the cut plane for the other time, so as to obtain a final section model. According to the method, the section of the complex three-dimensional geologic model and the section filling effect can be efficiently generated, the complex three-dimensional geologic model can be conveniently displayed at a browser end, and real-time sectioning analysis of the complex three-dimensional geologic model is achieved.

A method and apparatus of in-painting an image using prioritized graph cut optimization is disclosed. In one embodiment, the method includes examining an image comprising a plurality of pixels that form a source region and a target region, wherein the source region comprises pixel information partitioning the source region into blocks, defining boundary areas comprising a portion of the source region and a portion of the target region, computing a plurality of energy values for the source region and the boundary areas, wherein energy values represent intensity comparisons between the boundary areas and neighboring blocks of the source region and assigning labels to the boundary areas using on a graph-cut technique, wherein each label is associated with a neighboring block and an minimal energy value for each boundary area and storing pixel information based on the pixel information of the neighboring blocks associated with the minimal energy values.

The invention provides a water supply pipe network valve layout design and optimization method which comprises the following steps: S1, abstracting valves and pipe points in a pipe network into points, abstracting pipe sections into edges, and constructing an NE graph; S2, identifying a turn-off unit by using the NE graph; S3, constructing an SV graph; S4, utilizing the SV graph to calculate the weighted deficient supply quantity; and S5, solving an optimal valve layout scheme set based on NSGA-II. The water supply pipe network valve layout design and optimization method has the beneficial effects that under certain economic cost, the NSGA-II algorithm is used for optimizing the valve layout, and the water consumption which cannot be supplied after the pipe network has an accident is reduced to the minimum, so that the reliability of the pipe network is improved; the reconstruction of the SV graph can be realized only by performing local modification on the SV graph, and the time complexity of the algorithm is effectively reduced by using the cut points in the graph theory when the weighting lacking supply is calculated; and the water supply pipe network valve layout design and optimization method can be applied to the design of new pipe network valve layout and the transformation process of an existing pipe network, so that engineering cost can be controlled easily, and valvescan be scientifically and reasonably arranged.

The invention relates to a defect detection method applied to a motor coil based on a cascade expansion FCN network, and the method comprises the following steps: 1), collecting a sample image, i.e.,a big image, of a target region needing defect detection; 2) traversing all the sample images, and marking the defect position of each image; 3) performing sliding image segmentation on the large image by adopting a window sliding method, and segmenting the marked large image into small images with fixed sizes for training; 4) performing data enhancement on the small graph cut from the sliding window, and expanding the cut small graph; 5) applying the expanded small graph to the training of a network model, evaluating the defect detection effect, and adjusting parameters; and 6) obtaining a feature map output by the last layer of the network model, i.e., a segmentation position of the defect, i.e., a final output result. According to the method, the trained model can be added to an industrial production machine, whether the product has the defect or not can be automatically recognized, the defective product can be automatically shunted, the detection efficiency can be improved, and thelabor cost can be reduced.

The invention relates to the technical field of artificial intelligence, in particular to a photovoltaic cell panel solder strip offset detection method and device based on artificial intelligence. The method comprises the steps: acquiring a front view of the surface of a photovoltaic cell panel, and cutting the front view to obtain and number a cut graph of each solder strip; carrying out edge detection on the cut image of the solder strip, and sampling an edge binary image to obtain coordinate information of sampling points; performing linear fitting on the sampling points to obtain a linearequation; carrying out polynomial fitting on the sampling points, setting an error function to evaluate and optimize the coefficient of the polynomial, establishing a fitting degree evaluation indexto determine the highest power of the best fitting degree, and obtaining a polynomial equation reflecting the actual trend of the edge of the solder strip; and determining whether the solder strip deviates or not through difference operation of the polynomial equation and the linear equation to obtain position information of the deviated solder strip. According to the method, the problems that thecurrent solder strip offset depends on manual detection, the efficiency is low and the assembly power is attenuated can be improved.

The invention provides an unsupervised image segmentation method based on a Chan-Vese model, and belongs to the technical field of computer application and the field of precision medical treatment. According to the method, firstly, an original image is cut, super-pixel segmentation is carried out on a cut image, then, an undirected graph is established according to a super-pixel segmentation result, an energy item of a ChanVese model is expressed by the weight of an edge, and by using a Markov chain, the relationship between the distance between super-pixels and the gray scale can be considered simultaneously in the edge assignment process. And finally, a corresponding segmentation result can be obtained by repeatedly using the maximum flow segmentation and updating the weight of the edge.According to the method, the image can be automatically and accurately segmented in the medical field lacking annotation data, so that accurate information of a segmented object is provided for subsequent medical research. According to the method, the required target can be accurately segmented in a shorter time and with less manual intervention.

The invention discloses a batch cutting placement graph generation method and device. The method comprises the following steps of responding to a data import command of a main menu in an EXCEL cuttingdetail table, automatically importing part cutting information of N parts into the table from a process flow card, responding to a data import command in a menu corresponding to CAD software, obtaining the cutting information of the N parts from the table, drawing N blank block diagrams, filling the N blank block diagrams with part cutting information, receiving N cutting graphs input by a user,responding to operation of the user for frame selection of the cutting graphs, judging the type of the cutting graphs, and crushing a face domain and converting the face domain into a closed multi-segment line contour if the type of the cutting graphs is the face domain; judging the contour type; according to materials and the plate thickness of the cut parts, automatically searching for a corresponding placement allowance from a cutting allowance comparison table in the table, placing an inner contour inwards, and placing an outer contour outwards; and generating cutting placement graphs. Thecutting placement working efficiency and correctness are improved.

The invention relates to an image recognition method based on edge extraction, and belongs to the technical field of image recognition, and the method comprises the steps: carrying out the graying processing of a to-be-recognized image, and obtaining a first gray-scale image; performing noise reduction and smoothing processing on the first grey-scale map to obtain a second grey-scale map; performing edge detection on the second grey-scale map to obtain an edge information map; carrying out binarization processing on the edge information graph to obtain a binarized image; determining a bounding box of the target area by traversing the binarized image, and obtaining the binarized image with the bounding box; cutting the to-be-identified image according to the bounding box in the binarized image with the bounding box to obtain a cut image; and carrying out image identification by using the cut image. According to the invention, the to-be-identified image is cut according to the bounding box, so that low-cost and high-efficiency extraction of the region of interest is realized, the adverse effect of excessive background information on the performance of an image identification algorithm is solved, and the improvement of the image identification precision is facilitated.

The invention discloses a distributed SPARQL query optimization method based on minimum attribute cut, which belongs to the field of distributed systems, and comprises the following steps: (1) reading an original RDF data graph, and storing an edge attribute set L; (2) calculating a weak connected component and a corresponding cost of each edge attribute; (3) selecting internal attributes as many as possible to obtain a coarsened graph of the data graph; (4) carrying out vertex division on the coarsened image, and carrying out anti-coarsening processing to obtain a final partition; (5) decomposing the SPARQL query into a group of sub-queries which can be independently executed; and (6) executing the decomposed sub-queries in each partition in parallel to obtain a matching result. According to the method, the query types which can be independently executed in the distributed RDF system are expanded, the connection among the partitions is reduced, the data communication time is shortened, and the query efficiency is improved.

The present invention relates to a novel method and system for crater detection in planetary data based on marked point processes (MPP), effective for various object detection tasks in Earth observation, and for planetary image registration. The resulting spatial features are exploited for registration, together with fitness functions based on the MPP energy, on the mean directed Hausdorff distance, and on the mutual information. Two different methods—one based on birth-death processes and region-of-interest analysis, and the other based on graph cuts and decimated wavelets—are included within the present framework. Experimental results confirmed the effectiveness of the present invention in terms of crater detection performance and sub-pixel registration accuracy.

The invention provides a clustering method and a clustering system based on residential area context space information, Detailed and accurate information on urban architectural patterns is important for urban design, landscape evaluation, Social analysis and urban environmental research are essential, in order to realize the clustering analysis of residential polygon, the invention provides a simple and novel method, a synthesize data with generated data attribute with Lidar point cloud and Footprint data attributes is obtained by data fusion. The data are used to determine the neighboring relationship between buildings and the weights of adjacent buildings, and the best forest is obtained as the clustering result by the method of cutting graph theory in the spectral clustering based on the minimum spanning tree. The clustering result obtained by the clustering method of the invention is more objective and reliable.

The invention designs a distributed keyword query method based on an RDF graph, and belongs to the field of information retrieval. The method comprises the steps of firstly, converting the RDF data graph into an RDF sentence graph; secondly, by utilizing a conditional depth-first algorithm and a simulated annealing algorithm, segmenting the RDF sentence graph according to two most basic principlesincluding minimum of an edge cut set and data balance between sub-graphs after segmentation; and finally, refining the segmented RDF sentence graph into the RDF data graph, obtaining a vertex cut setof the RDF graph, and by utilizing a reverse search algorithm and a Hadoop distributed computing framework, realizing efficient and fast query of keywords. Under the condition of guaranteeing the atomicity and semantic integrity of RDF data, the limitation of a traditional algorithm on the segmentation efficiency of a large-scale data set is effectively eliminated, and the query efficiency of thekeywords is greatly improved.

The invention discloses a method for segmenting a target by a binocular stereo image based on a Graph Cut theory, and the method comprises the steps: firstly inputting a left image and a right image in the binocular stereo image, and obtaining disparity maps of the two images through calculation through a stereo matching algorithm; secondly, calculating weight information of the image; then, constructing a graph model of a Graph Cut method according to the weight information to obtain a segmentation result of the three-dimensional image pair; and finally, filling small holes in the segmentation result by using morphological operations such as expansion and corrosion on the segmentation result, removing small connected regions in the segmentation result, and smoothing by using median filtering to obtain a final segmentation result. Compared with a traditional image segmentation method, the method disclosed by the invention strengthens a neighborhood system constructed in the traditionalmethod, improves the segmentation precision at the edge of the image, better uses the depth information of the binocular image, and obtains a better segmentation result.

The invention discloses an aluminum alloy micron-sized second phase quantitative statistical characterization method based on deep learning, which comprises the following steps of: obtaining a featuredatabase of a standard sample, training the feature database by using a deep learning-based image segmentation network U-Net to obtain a U-Net segmentation model, and selecting parameters corresponding to optimal precision to establish a U-Net target model; cutting a to-be-tested aluminum alloy image, inputting the cut to-be-tested aluminum alloy image into the U-Net target model, obtaining size,area and position information of a second phase through a connected region algorithm, performing statistical distribution representation on the data set in combination with a mathematical statisticalmethod, and restoring the position information in the test image to the surface of the to-be-tested aluminum alloy to obtain a to-be-tested aluminum alloy surface, and obtaining a full-view-field quantitative statistical distribution condition and a visualization result. Based on a deep learning image segmentation algorithm, an aluminum alloy micron-scale second phase is automatically recognizedand extracted, extracted features are positioned and counted, and the method has the advantages of being large in view field, complete in information, accurate and reliable.

The invention relates to the technical field of image processing, and provides an image segmentation and data enhancement method, system and device, and the method comprises the steps that a server receives a commodity image uploaded by a mobile terminal, and builds a training set and a test set; the server identifies a target area of the commodity image, and cuts the commodity image according tothe size of the target area; the server carries out data enhancement processing on the cut images, and a training set and a test set are expanded; the system comprises a receiving module, an image cutting module and a data enhancement module. According to the invention, the data set is expanded in a mode of combining image cutting and image enhancement, a user does not need to input a large numberof commodity images, so the convenience of the user is improved, the effectiveness of the training set is effectively improved, and then the fitting capability of the neural network model can be improved.

The invention provides a graph segmentation method, relates to the field of graph theory, in particular to the field of artificial intelligence such as image processing, can be applied to scenes such as statistical physics, network design, super-large-scale integrated circuit design and data clustering analysis, and comprises the following steps: acquiring a to-be-segmented graph, wherein the graph comprises a plurality of vertexes and edges between the vertexes; generating variable parameter vectors of the vertexes, and iterating a minimum loss function of the graph according to the variable parameter vectors of the vertexes; obtaining a target value of a variable parameter vector corresponding to the vertex at the generation moment of the minimum loss function; and for each vertex, determining the number of cut edges of the vertex according to the target value corresponding to the vertex. According to the method, the maximum cut problem is converted into the optimization problem of the variable parameter vector, the solving difficulty of the maximum cut problem is reduced, the solving efficiency of the maximum cut problem is improved, the practicability of the maximum cut problem is enhanced, and the solving efficiency of the maximum cut problem is enhanced.

When each pixel forming image data is classified into one of N labels (where N>2) which are ordered from 0 to N−1, a binary graph setting unit sets a binary graph where the i-th layer i is a boundary between the label i−1 and the label i of the N labels, a class to which a label equal to or higher than i is assigned corresponds to a virtual label 0, and a class to which a label equal to or smaller than i−1 is assigned corresponds to a virtual label 1, a layer-by-layer labeling unit performs a graph cut operation on the binary graph of each layer, and then a label determining unit determines which of the N labels each pixel belongs to based on the virtual labels assigned to all the binary graphs.

The invention relates to the field of artificial intelligence, discloses a data interpretation method, device and equipment based on image recognition and a storage medium, and is used for solving the problem of low efficiency in interpreting physical examination reports with different format contents in the prior art. The method comprises the following steps: receiving and extracting a physical examination report picture in a data interpretation request; recognizing a text position of a physical examination result in the physical examination report picture, and cutting the physical examination report picture to obtain a cut image set; performing text recognition on cut image blocks in the cut image set to obtain text content; determining an examination category and an examination result contained in the cut image blocks, and obtaining a corresponding standard index range; judging whether the examination result is in the standard index range or not; if not, marking the examination result and the examination category as an abnormal examination result and an abnormal examination category; and retrieving medical information in a medical knowledge graph, and outputting a data interpretation result. In addition, the invention also relates to a block chain technology, and related information of the physical examination report can be stored in a block chain.

The invention discloses a layer exporting method and device, equipment and a storage medium, and belongs to the field of UI design. The method comprises the steps that a node tree of a visual draft file of a user interface is acquired, the node tree comprises visual element nodes, and the visual element nodes correspond to graph layers of visual elements forming the user interface; visual elementnodes which do not support code construction in the node tree are determined; the visual element nodes belonging to the same level in the visual element nodes which do not support code construction are combined to obtain combined visual element nodes; and exporting the graph layers corresponding to the merged visual element nodes as first cut graph layers. In the image cutting process, manual operation is not needed, and the cut image layer can be exported only by processing the node tree of the visual draft file through computer equipment. The problems that errors are prone to occurring and efficiency is low due to manual operation can be avoided, and the technical scheme of full-automatic exporting and image cutting is provided.