33results about How to "Simplify the segmentation process" patented technology

The invention discloses a data safety protection method and device based on data partitioning. The method includes the steps that private data are extracted out of original data according to a preset private data extraction rule and partitioned into private data and non-private data; padding data are added to the private data, and encrypted data are obtained; a mapping table is generated according to the serial number of the private data, the padding data and the encrypted data; meanwhile, an index table is generated according to the serial number of the private data, nouns of the mapping table and original file names corresponding to the private data; an information safety card of the private data is generated according to the mapping table and the index table, and the information safety card of the private data is stored in a database according to preset private data storage strategies. According to the data safety protection method and device, the data are partitioned, the private data generate the mapping table and the index table, the information safety card of the private data is established, storage is conducted through different strategies, storage and reading efficiency is improved, and data portioning and encrypting difficulty is greatly simplified.

The invention discloses a polarization SAR image segmentation method based on characteristic value measurement spectral clustering. The method mainly solves the problems that the number of parameters is large and adaptive adjustment is difficult to carry out in the existing polarization SAR image segmentation process. The method includes the steps of 1, carrying out eigenvalue decomposition on a polarization SAR image to form a feature sample set x, 2, solving corresponding mean values of three eigenvalues in an 8 neighborhood of each pixel to form a mean feature sample set, 3, respectively building a similarity matrix for the feature sample set x and the mean feature sample set utilizing mahalanobis distance so as to obtain a mixed similarity matrix w' according to the two similarity matrixes, 4, obtaining a clustering label C1 of the mixed similarity matrix w' through the spectral clustering algorithm, and 5, repeating the steps 3 and 4, integrating obtained class label sets by utilizing the MCLA so as to obtain a final segmentation result. The polarization SAR image segmentation method has the advantages of being high in adaptivity, low in complexity, detailed and accurate in segmentation result, and capable of being used for target detection and target recognition of the polarization SAR image.

The present invention provides a CT image interaction segmentation method and device. The method comprises the steps of: regulating a CT intensity threshold range, obtaining portions to be displayed having CT intensity values belonging to the intensity threshold range, and performing marking and coloring of the portions to be displayed, employing the volume rending technology to simulate contour surface display, and displaying an original data volume rending view; determining areas to be segmented in the original data volume rending view through a marking tool, employing a filling tool to calculate the areas to be segmented, generating an initial segmentation result, and employing a corrosion expansion tool to regulate the initial segmentation result; and on the basis of the initial segmentation result, employing a marking tool to select areas without segmentation in the areas to be segmented, generating a segmentation result, and displaying a segmentation volume rending view. The CT image interaction segmentation method and device can employ a visual interaction means to simplify the segmentation process so as to save a lot of segmentation time and facilitate users' usage.

The invention discloses a deep learning model training method and device. The method comprises the following steps: the deep learning model is trained through the chest medical image marked with the frosted glass nodule, the trained deep learning model is obtained, the frosted glass nodule segmentation efficiency is improved through the trained deep learning model, in addition, the segmentation accuracy is effectively improved, and the segmentation process of the frosted glass nodule is simplified. Furthermore, a multi-stage feature fusion monitoring module is added to the deep learning model,so that the response of the deep learning model to the focus area in the image is enhanced, the whole image is processed, and the operation process is simplified; moreover, an autonomous grading supervision module is also added, so that the deep learning model can autonomously adjust the influence of the features of different levels on the segmentation result. In addition, the decoded feature mapis subjected to up-sampling operation, so that the ground glass nodules can be segmented more accurately by the deep learning model.

The invention provides a method for segmenting and marking a multi-branch tubular structure in a three-dimensional image, and belongs to the field of medical image processing. According to the method,a multi-branch tubular structure is regarded as a set of a plurality of single-branch tubular structures, and for the single-branch tubular structures, a center line is obtained through key point detection and neural network fine adjustment methods; then a cross section image of the tubular structure is acquired along the center line based on the center line, and finally a segmentation result ofthe tubular structure is acquired by using a segmentation network. The specific multi-branch tubular structure existing in the image can be segmented and automatically marked, and especially for someimages with lesions, such as aortic dissection, aortic aneurysm and coronary artery stenosis images, a good segmentation result is achieved.

The embodiment of the invention discloses a head image area segmentation method and device, electronic equipment and a storage medium, a point cloud registration method and device, electronic equipment and a storage medium, and a face modeling method and device, electronic equipment and a storage medium. The head image region segmentation method comprises the steps of determining an image region where a body part in a target image is located from the acquired target image; calculating a width value of the body part in the target image based on the determined image area; based on the width value of the body part in the target image, determining the segmentation position of the trunk object and the head object in the target image; and performing image segmentation on the foreground image region in the target image at the segmentation position to obtain a head image region. According to the embodiment of the invention, the head image region can be obtained based on the width value of the body part in the target image, so that the segmentation accuracy of the head image region is improved, and the segmentation process is simplified.

The embodiment of the invention discloses a video segmentation method and device, equipment and a medium. The method comprises the steps: obtaining a to-be-segmented video, and determining the corresponding relation between knowledge point data and video frames in the to-be-segmented video; and segmenting the to-be-segmented video according to the corresponding relationship to obtain at least onevideo clip. According to the video segmentation method provided by the embodiment of the invention, the to-be-segmented video is segmented according to the knowledge point data in the to-be-segmentedvideo, so that the segmentation process of the video is simplified, and the segmentation accuracy of the video is improved.

The invention discloses a geological radar-based railway tunnel internal disease detection method and a device thereof. The method comprises the following steps: acquiring a lining radar image in a railway tunnel; carrying out coarse precision segmentation on the grayscale profile map; calculating a gray variance of the radar segmentation image I; determining an adaptive threshold by maximizing the ratio of variances, and screening out a region of interest and a background region in a radar segmentation image I according to the adaptive threshold; expanding the contrast of the region of interest and the background region; performing high-precision segmentation on the grayscale profile map; determining a discrete grayscale image two-dimensional entropy, taking an area of which the grayscaleimage two-dimensional entropy is greater than a preset two-dimensional entropy as a target area, and finally screening out a target disease area where the target area is located; and outputting the radar image with the target disease area and the mileage of the disease area. According to the method, tunnel disease conditions can be detected and measured on a large scale, and the tunnel disease detection efficiency and precision can be improved.

The invention discloses a complementary model method and device and a computer readable storage medium, and the method comprises the steps: carrying out the first model training of a blood vessel datasample, and obtaining a first neural network model; performing second model training on the blood vessel data sample to obtain a second neural network model complementary to the first neural networkmodel; and segmenting the blood vessel data through the first neural network model and the second neural network model to obtain a blood vessel data segmentation result.

The invention relates to a data packaging method capable of matching multiple transmission modes and a system. The method comprises steps: S1, service data are divided into fragmented files with fixed lengths; S2, according to a broadcasting plan, the fragmented files and auxiliary data are added in an interleaving mode in a broadcasting queue; S3, the fragmented files are packaged into a plurality of UMCP (Unit Multi-Cast Protocol) blocks and whether data check is added to the UMCP blocks according to the link quality condition; S4, one or more UMCP blocks is packaged into network protocol frames according to the needs of network adaptation. Through carrying out pre-treatment on the fragmented files with fixed lengths, the data dividing process is simplified, service data concurrence in multiple channels can be realized, the broadcasting efficiency is improved, reliability of transmission is improved through mixed broadcasting of service data and auxiliary data and the check function, selective data receiving is realized through service needs, service flexibility and receiving efficiency are improved, the method can be adapted to multiple kinds of data link layer network protocols, and cross network collaboration service in different network conditions can be realized.

The invention relates to a medical lung MRI image segmentation method based on an adaptive contour model, and MRI equipment. The equipment comprises a main magnet system, a gradient magnetic field system, a radio frequency system and an operation and image processing system; the radio frequency system comprises a plurality of radio frequency coils which are respectively arranged in coil fixing devices, and different radio frequency receiving coils correspond to different detection parts on a patient. The method comprises the following steps: 1) preprocessing an input fused MRI image; 2) setting an initial contour line as c0 and setting the initial contour line as a circle according to the characteristics of the lung tumor image, and calculating an initial level set function phi 0 accordingto the c0; 3) updating a level set function phi n, and calculating c1 and c2 according to the current phi n; 4) checking whether iteration is convergent or not: if so, determining that c is the optimal contour line, and otherwise, continuing iteration; 5), after the target area is obtained, removing noise and some tiny protruding parts by using image morphological opening operation, smoothing theboundary, connecting the fracture part by using closed operation, and filling a cavity to obtain a target image. The method is high in segmentation precision and clear in edge, and is suitable for segmenting lung MRI images with fuzzy targets, weak edges, smooth boundaries, discontinuous boundaries and complex topological structures.

The invention relates to an MRI image fusion method based on Laplace pyramid transformation and MRI equipment. The MRI image fusion method comprises the following steps: S1, for the features of a brain MRI image, decomposing a plurality of source images by using Laplace pyramid decomposition to obtain different frequency layers, and adopting different fusion rules in the different frequency layersso as to reserve feature information of each source image in the different frequency layers in a fused image; S2, respectively calculating the regional mean value of the top layer and the point sharpness of other layers to serve as fusion scales; S3, performing normalization processing on the regional mean value and the point definition; S4, comparing the normalized region mean value and point definition value of each layer of different source images, and obtaining a fusion result of each layer of images by adopting different fusion strategies; and S5, for each layer of the Laplace pyramid ofthe obtained fusion image, performing recursion downwards layer by layer from the top layer, and finally obtaining the fusion image. By adopting the MRI image fusion method, the multi-focus fusion image with low noise and clear edge can be obtained.

The embodiment of the invention discloses an image reconstruction method and device, equipment and a medium. The method comprises: acquiring collected under-sampling data, inputting the under-samplingdata into a pre-trained target image reconstruction model, and obtaining the target image reconstruction model by generalizing a data fidelity item of an original image reconstruction model and thensolving the data fidelity item; and obtaining a reconstructed image output by the target image reconstruction model. According to the image reconstruction method provided by the embodiment of the invention, the data fidelity item of the original image reconstruction model is generalized and then solved to obtain the target image reconstruction model. According to the method, the reconstructed image is obtained based on the obtained target image reconstruction model and the undersampled data, the network freedom degree of the neural network is improved, the data consistency in the image reconstruction process is guaranteed, and the quality of the reconstructed image is improved.

The present invention discloses a lesion volume measurement based MRI (Magnetic Resonance Imaging) automatic image segmentation method. The method comprises: firstly, acquiring a gray image of a to-be-segmented MRI scanned image; secondly, dividing the gray image into sub-blocks that do not overlap mutually; thirdly, segmenting the gray image by using a marker-based watershed segmentation algorithm to obtain a plurality of preliminary lesion regions; and finally, inputting respective pixel values, as input parameters, of all pixel points in each sub-block corresponding to each preliminary lesion region into a non-linear optimization model for optimization to obtain a corresponding final lesion region. The advantage is that preliminary segmentation is performed on the gray image by using the marker-based watershed segmentation algorithm, so that not only does the obtained preliminary lesion region have high accuracy, but also the segmentation efficiency is high; after the preliminary segmentation, the optimization is performed by using the non-linear optimization model, so that an accurate lesion region can be obtained only by performing segmentation once in combination with optimization; and a segmentation process is simple, the segmentation efficiency is high, and the computation amount is small, so that the method is suitable for massive image segmentation operations.

The invention discloses a method for generating an SCMA (Sparse Code Multiple Access) codebook. The method comprises the steps of: building a factor matrix according to a subcarrier number K and a user number J as well as a predetermined dimension number L of a unit matrix; selecting a QAM (Quadrature Amplitude Modulation) master constellation diagram according to the user number actually transmitted by each subcarrier, optimizing the QAM master constellation diagram by aiming at maximizing the minimum euclidean distance between constellation points, dividing the QAM master constellation diagram by the method of equally dividing the constellation points to obtain sub-constellation diagrams with preset number; and randomly replacing non-zero elements in the factor matrix with the differentsub-constellation diagrams to form a constellation matrix, and generating a codebook according to the constellation matrix and coordinates of the constellation points in the sub-constellation diagrams. The method of the invention can be used for generating the SCMA codebook, and enables the minimum euclidean distance between the constellation points in the constellation diagram corresponding to users to be maximized, thereby reducing error rate of a transmission system using the codebook under the condition of wireless transmission, and achieving high wireless communication application value.

The invention discloses a complementary model method, equipment and computer-readable storage medium, comprising: performing first model training on blood vessel data samples to obtain a first neural network model; performing second model training on blood vessel data samples, Obtaining a second neural network model complementary to the first neural network model; segmenting the blood vessel data by using the first neural network model and the second neural network model to obtain a blood vessel data segmentation result.

The embodiment of the invention discloses an image processing method and device, which are used to rapidly segment a foreground image and a background image in an image. The method in the embodiment of the present application includes: acquiring an original image, and judging whether the original image is a single-channel image; if not, converting the original image into a single-channel image to obtain a grayscale image of the original image; Perform high-frequency filtering on the grayscale value of each pixel in the grayscale image to obtain a first grayscale image; according to the grayscale image and the first grayscale image, the grayscale value of each pixel in the grayscale image value to perform Kalman filtering to obtain the template image of the grayscale image; calculate the distance range between the grayscale value of each pixel in the template image and the grayscale value difference of the corresponding pixel in the first grayscale image If the distance norm is greater than the preset threshold, the current pixel is defined as the foreground image, otherwise, the current pixel is defined as the background image.

The invention discloses a text segmentation method and device, and relates to the technical field of cloud computing and text processing. The specific implementation mode comprises the steps: acquiring a target text and performing sentence segmentation processing on the target text, thus obtaining a sentence segmentation result; in response to any sentence segmentation result, determining that thelength of the sentence segmentation result exceeds a preset length threshold, and performing word segmentation processing on the sentence segmentation result to obtain at least three vocabularies; performing vocabulary combination on at least two vocabularies in the at least three vocabularies to obtain word groups; comparing the length of the current word group with a preset length threshold; and in response to the fact that the length of the word group does not exceed a preset length threshold, taking the word group as a segmentation result of the target text. According to the method, the text segmentation process can be simplified, computing resources are saved, and the text processing efficiency is improved. Moreover, the length of the text segmentation result can be controlled, and the problem that the obtained segmentation result is too long to achieve the segmentation purpose is avoided.

The invention discloses a segmentation method, system, medium and device based on a U-Net network. The method comprises the following steps: acquiring a brachial plexus ultrasound image; performing down-sampling on the brachial plexus ultrasound image for four times to obtain a first feature map of which the size is 1 / 2 of the size of the image before down-sampling; performing U-Net-based jump connection on the first feature map to obtain a second feature map with significant features; decoding the second feature map based on U-Net, and recovering to the size of the brachial plexus ultrasound image; and carrying out convolution on the decoded features to obtain a segmentation mask of the brachial plexus ultrasound image. According to the invention, no operator is used for image preprocessing, and the image segmentation process is simplified; the features are extracted by using an efficient OfficientNet series network, so that the speed is high and the effect is good; U-Net jump connection is designed, more significant features are extracted on a large receptive field by using cavity convolution and maximum pooling operation, and the effect of noise suppression is achieved.

In a method and apparatus for recording a magnetic resonance dataset of at least one foreign body in a target region of a patient, a magnetic resonance sequence having an ultra-short echo time, which is less than 500 μs is used for recording the magnetic resonance data.

The invention provides a large joint tissue segmentation method and a segmentation system, which solves the technical problem of low efficiency of joint tissue segmentation in the prior art. The method includes: forming an image segmentation criterion; forming the tissue category and position feature of a single pixel in the MRI image according to the image segmentation criterion to determine the tissue plane profile; combining the image segmentation with the tissue plane profile in each of the MRI images The fiducials form a three-dimensional outline of the tissue. The use of bone features simplifies the tissue segmentation process, the use of related image and information features to form automatic processing of object contours and the improvement of recognition accuracy, and the use of segmentation benchmarks to form image quantification rules, pixel connotation information, and modeling frameworks. Organizations can achieve automatic subject segmentation, accurate contour positioning, and automatic 3D modeling in MRI images, effectively improving the identification efficiency of professional human identification resources.

The invention relates to a contrast medium for ex vivo microangiography for digital imaging of the vascular system of a mouse or rat or other laboratory animals, and of individual animal and human organs, comprising an iodized, esterified oil, a polyurethane, and a hardener. The invention further relates to a method for producing the contrast medium; to a kit of parts, comprising the various containers having the components of the contrast medium according to the invention that are to be mixed; and to a preferred use of the contrast medium according to the invention for imaging by means of a micro-CT device.

The present invention relates to a data encapsulation method and system matching multiple transmission modes. The method includes: step 1, dividing business data into fixed-length fragment files; step 2, dividing the fragment files and auxiliary Data interleaving is added to the broadcast queue; step 3, the fragmented file is encapsulated into multiple UMCP blocks, and it is judged whether to add data verification to the UMCP block according to the link quality; step 4, according to the needs of network adaptation, one or more A UMCP block is encapsulated into a network protocol frame. The present invention simplifies the data segmentation process through fixed-length slice preprocessing, realizes multi-channel concurrency of business data, improves broadcast efficiency, and improves transmission reliability through mixed broadcast and verification functions of business data and auxiliary data , realize the selective reception of data through business requirements, improve business flexibility and reception efficiency, and can adapt to a variety of data link layer network protocols to realize cross-network collaborative services under different network conditions.

The present invention discloses a lesion volume measurement based MRI (Magnetic Resonance Imaging) automatic image segmentation method. The method comprises: firstly, acquiring a gray image of a to-be-segmented MRI scanned image; secondly, dividing the gray image into sub-blocks that do not overlap mutually; thirdly, segmenting the gray image by using a marker-based watershed segmentation algorithm to obtain a plurality of preliminary lesion regions; and finally, inputting respective pixel values, as input parameters, of all pixel points in each sub-block corresponding to each preliminary lesion region into a non-linear optimization model for optimization to obtain a corresponding final lesion region. The advantage is that preliminary segmentation is performed on the gray image by using the marker-based watershed segmentation algorithm, so that not only does the obtained preliminary lesion region have high accuracy, but also the segmentation efficiency is high; after the preliminary segmentation, the optimization is performed by using the non-linear optimization model, so that an accurate lesion region can be obtained only by performing segmentation once in combination with optimization; and a segmentation process is simple, the segmentation efficiency is high, and the computation amount is small, so that the method is suitable for massive image segmentation operations.

The invention provides a segmentation method and a segmentation system for large joint tissue, solving the technical problem of low segmentation efficiency of the existing joint tissue. The segmentation method comprises the steps: forming an image segmentation reference; forming tissue types and position features of a single pixel in the MRI image according to the image segmentation reference so as to determine a tissue plane contour; and forming a tissue three-dimensional contour according to the tissue plane contour in each MRI image in combination with the image segmentation reference. Thetissue segmentation process is simplified by utilizing skeleton features; related image and information characteristics are used to form automatic processing of an object contour and identification accuracy is increased; and an image quantization rule is formed by using a segmentation reference, and pixel connotation information and a modeling framework are effectively fused, so that a large jointcomposition tissue can realize automatic main body segmentation, accurate contour positioning and automatic three-dimensional modeling in an MRI image, and the identification efficiency of professional human identification resources is effectively improved.

The present invention provides a CT image interaction segmentation method and device. The method comprises the steps of: regulating a CT intensity threshold range, obtaining portions to be displayed having CT intensity values belonging to the intensity threshold range, and performing marking and coloring of the portions to be displayed, employing the volume rending technology to simulate contour surface display, and displaying an original data volume rending view; determining areas to be segmented in the original data volume rending view through a marking tool, employing a filling tool to calculate the areas to be segmented, generating an initial segmentation result, and employing a corrosion expansion tool to regulate the initial segmentation result; and on the basis of the initial segmentation result, employing a marking tool to select areas without segmentation in the areas to be segmented, generating a segmentation result, and displaying a segmentation volume rending view. The CT image interaction segmentation method and device can employ a visual interaction means to simplify the segmentation process so as to save a lot of segmentation time and facilitate users' usage.

The invention provides a three-dimensional MRI image-based breast tumor automatic segmentation method. The method comprises the following steps of performing image preprocessing: providing an initial MRI image and performing preprocessing on the initial MRI image by adopting a non local mean filter; performing breast tumor locating: building a multilayer processing model for a training set, performing hierarchical abstraction on characteristics of a training object by adopting a convolutional neural network, automatically extracting segmentation characteristics, and outputting a probability distribution graph of a tumor position; performing breast tumor boundary segmentation: providing a breast three-dimensional MRI image, determining a seed point based on the probability distribution graph of the tumor position, and finishing segmentation initialization to obtain an initial region C0 of a tumor; and performing accurate segmentation on the tumor by using a region growth algorithm.

The embodiment of the invention discloses a model separation method. The method comprises the following steps: obtaining a three-dimensional model containing a heart model according to the obtained three-dimensional model, constructing a three-dimensional matrix of the three-dimensional model; performing corrosion operation in morphology; performing operation on the three-dimensional matrix by adopting the constructed first-class three-dimensional matrix; obtaining a three-dimensional matrix after corrosion operation; performing expansion operation in morphology; calculating the three-dimensional matrix subjected to corrosion operation by adopting the constructed second-class three-dimensional matrix to obtain a three-dimensional matrix subjected to expansion operation, and subtracting thethree-dimensional matrix of the three-dimensional model from the three-dimensional matrix subjected to expansion operation to obtain a difference matrix so as to separate the auricle model from the model corresponding to the difference matrix. The embodiment of the invention also discloses a model separation device, a terminal and a computer storage medium.