186results about How to "Realize automatic segmentation" patented technology

The invention discloses a CT image kidney segmentation algorithm based on a residual double-attention deep network. According to the method, the advantage that the residual error unit can repeatedly utilize the features and the excellent feature learning capability of the double attention mechanism are combined; a residual double attention module is designed; and a residual double attention moduleis used as a basic module to construct a U-shaped deep network segmentation model, and a loss function for segmentation is designed at the same time, so that the U-shaped deep network segmentation model can pay more attention to kidney region features, can effectively cope with shape changes of the kidney with cystic lesions, and can maintain robustness for the shape changes of the kidney with cystic lesions. Therefore, the boundary of the kidney area is accurately positioned, and automatic segmentation of the kidney area in the CT image is achieved, and a good segmentation effect is achieved.

The invention relates to a medical image recognition method, which aims at providing the method which can more accurately recognize the type of a new medical image. The method comprises the construction of a classification library of association rules, the update thereof and the medical image recognition step, the construction of the classification library of the association rules and the update step thereof comprise the following steps: data of medical image samples are prepared and carried out the pre-treatment; a density clustering-based medical image segmentation method is adopted to respectively recognize local region to be analyzed in each sample medical image; the characteristics of the local region in each sample medical image are respectively extracted to construct a medical image sample database T, the characteristics comprise: mean, variance, inclination, kurtosis, energy, entropy and clustering characteristics; the characteristic values are carried out the discretization; the characteristic values are carried out the discretization; a frequent itemset in the medical image sample database is excavated; and the classification library of the association rules is constructed according to the frequent itemset.

The invention discloses a method adopting image segmentation for automatically testing the LED indicator light of automobile instruments and provides a region-growing method based on the maximum between-cluster variance (Otsu's Method). The method comprises the following steps: automatically selecting the centroid in the light-emitting region of the LED light as the seed point to carry out the region-growing; realizing the region segmentation and extraction; and judging whether the brightness and color of the light are correct upon the seed point and the area of the region. The invention can effectively improve the testing accuracy rate and testing efficiency of the LED indicator light of the automobile instruments and provide effective means for the automatic production of the automobile instruments.

The invention relates to an intelligent blumeria graminis spore picture identification method. The method includes the steps of selecting different expert models to enable the intelligent identification accuracy to be capable of adapting to different requirements; pre-processing blumeria graminis spore pictures; dividing the blumeria graminis spore pictures; extracting the color, texture and shape features of the blumeria graminis spore pictures; conducting intelligent identification on the extracted features of the blumeria graminis spore pictures. By means of the intelligent blumeria graminis spore picture identification method, the blumeria graminis spore density in unit volume in air for a certain time can be rapidly calculated, and the basic conditions of blumeria graminis spore diseases are obtained. By means of the intelligent blumeria graminis spore picture identification method, automatic dividing of the blumeria graminis spore pictures is achieved, and the problem that an existing manual dividing method is low in efficiency is solved; automatic identification and automatic counting of the blumeria graminis spore pictures are achieved, and the problem that an existing manual (expert) identification method is low in efficiency and prone to making mistakes is solved; the intelligent blumeria graminis spore picture identification method can be suitable for the different manual (expert) using requirements and high in adaptability.

The invention discloses a registration-based CT (Computed Tomography) image total heart automatic cutting system. In the system, registration is applied to cutting to realize automatic cutting of a total heart by using similarity of heart CTs of different individuals. The system comprises the following modules: an input module, a cutting module, a rough registration module, a fine registration module, a conversion module and an output module. The system can realize automatic cutting of the total heart and can be used for simultaneously cutting a plurality of heart chambers.

The invention discloses a building facade three-dimensional reconstruction method based on knapsack type three-dimensional laser point cloud data, and relates to the technical field of geographic information. The method comprises the following steps: S1, acquiring building point cloud data; S2, automatically extracting building facade point cloud data; S3, automatically segmenting the single buildings; S4, acquiring a geometric position boundary of the building facade; and S5, building facade three-dimensional reconstruction. A point cloud filtering algorithm based on voxel projection densityis adopted to effectively filter non-building targets such as the ground and vegetation while a relatively complete building target is reserved, and then automatic segmentation of a single building isrealized by utilizing an image global search and profile analysis method. An RANSAC algorithm is used to carry out facade automatic segmentation and redundant facade elimination on the single building point cloud to obtain a building facade geometric position boundary; and the two-dimensional boundary line is used to constrain the original point cloud data and an RANSAC algorithm is combined tocarry out facade three-dimensional boundary straight line fitting so as to obtain a building facade geometric frame model.

The invention relates to an automatic segmentation method of pathology area based on deep learning, comprising the steps of S1 collecting multiple case data and conducting standardized preprocessing on medical images of a pathology position specific modal; S2 conducting edge mark on the pathology area layer by layer by a medical science audio visual technician, and using the marks as real data; S3conducting extraction of the training sample, wherein a plurality of voxels are extracted at random from voxels in the pathology area and within a certain distance outside of the pathology area, andimage blocks having fixed size are used as the training sample for the next step with the voxels being the center; S4 establishing a deep learning nerve network, and conducting training on the positive and negative samples of the above cases; conducting post-processing and segmentation precision assessment, and obtaining a segmentation model after the satisfied segmentation precision is obtained;S5 collecting medical science images of the same modal at the same position, and conducting standardized preprocessing for cases to be diagnosed; S6 automatically detecting the pathology area by meansof the segmentation model, and outputting the segmentation result.

The invention discloses an active contour model-based automatic positioning and segmentation method of a spinal CT image, and relates to the field of medical image processing. The method provides themethod of automatic positioning and segmentation on the CT image for the sensitivity problem which is of segmentation methods of spinal CT images and on initial positions and contours. n groups of spinal CT images are obtained by scanning of clinical CT instruments, and the CT slices are manually segmented by experts and are used as training samples; a random forest algorithm is used for positioning vertebral centers to determine the vertebral centers; initial contours of segmentation are placed at central positions determined by the random forest algorithm, and fuzzy contour segmentation is adopted to obtain vertebrae in the CT slice images by segmentation; and a trained model combination is output to obtain a complete vertebral CT image segmentation model. According to the spinal CT segmentation model provided by the invention, a vertebral center and an initial contour position of segmentation can be automatically positioned, a vertebra can be automatically obtained by segmentation,and segmentation steps and processes of the spinal CT image can be simplified.

Provided are a screening method and system for a fundus image of glaucoma based on deep learning. The screening method comprises the steps that the fundus image to be identified is acquired; an FCN model is extracted through an optic disk to position and cut the optic disk, and optic disk sub-images are extracted and obtained; the optic disk sub-images are identified through a classification CNN model, and a classification result of the fundus image of glaucoma is obtained. According to the screening method and system for the fundus image of the glaucoma based on deep learning, identificationis only conducted on the cut and extracted optic disk, so that information processing amount is reduced, and the screening method and system can assist a doctor in improving the detection efficiency of glaucoma.

The invention discloses an automatic segmentation method of a sandstone thin-section microscopic image, which comprises the following steps: 1) adopting a super pixel segmentation technology, pre-segmenting an orthogonal polarized light microscopic image of the sandstone into image blocks; 2) extracting color features and texture feature of image blocks and constructing feature vectors base on orthogonal polarizing microscopic images; 3) extracting image boundary features by adopt boundary detection technology based on that single polarized light microscopic sheet image; 4) trainning a supportvector machine classifier base on a sandstone particle sample data set; 5) using the trained classifier to predict the probability that each image block belongs to quartz, feldspar and debris, and marking the image block type through preset conditions; 6) predicting a type of an image block of an unlabeled type by using a label propagation algorithm; 7) merging the adjacent image blocks with thesame type and lower boundary characteristic intensity. This method utilizes image processing technology, machine learning method and data mining method, and combines the orthogonal polarizing microscope image and mono-polarizing microscope image obtained from the same sandstone slice to automatically segment the mineral particles contained in the sandstone slice, so as to reduce the time and economic cost of manual division of labor and improve the segmentation accuracy.

The invention provides a tongue image segmentation method and a device. The segmentation method comprises the following steps: acquiring a grayscale tongue image and an HSI colored image according to the original colored image; acquiring a first image based on the grayscale tongue image by utilizing a watershed algorithm; segmenting the HSI colored image based on the data of a channel H of the HSI colored image so as to acquire a second image; segmenting the second image based on the data of a channel I of the HSI colored image so as to acquire a third image; and calculating the set of the first image and the third image, thereby obtaining the tongue segmentation result map. The segmentation method and segmentation device are sensitive to the tongue boundary in the image, and the segmentation result is accurate; and the segmentation speed is high, and the method and device have high timeliness.

The invention belongs to the technical field of medical image processing, and discloses a hippocampus segmentation method for an automatic brain MRI (Magnetic Resonance Image) on the basis of multipleatlases. The method comprises the following steps that: (1) adopting a non-rigid registration method to carry out registration on atlas set and a brain MRI to be segmented; (2) calculating a similarity between an atlas image and a target image, and constructing and selecting a similar atlas which is most favorable for target image hippocampus segmentation; and (3) obtaining the confidence coefficient weighting probability matrix of the atlas image, establishing a context model based on the similar atlas, and combining the confidence coefficient weighting probability matrix of the atlas imagewith the context model to obtain a hippocampus segmentation result in the target image. By use of the method, image features and image used for segmenting the hippocampus in the atlas image can be mined, an accurate hippocampus segmentation result can be obtained under a situation that time complexity is controlled, and the problem in the prior art that the automatic segmentation accuracy of the hippocampus for the brain MRI image is low is overcome.

The embodiment of the invention provides a video data segmentation model training method, a video data segmenting method, a video data segmentation model training device and a video data segmenting device. The training method comprises the following steps of performing video feature detection on first video data so as to obtain information about one or more first video feature vectors; training by use of the information about the one or more first video feature vectors, so as to obtain a segmentation result; carrying out segmentation on the first video data by adopting a video data segmentation model, so as to obtain a segmentation result; judging whether the video data segmentation model meets preset verification conditions or not according to the segmentation result; if the video data segmentation model meets the preset verification conditions, outputting the video data segmentation model; if the video data segmentation model does not meet the preset verification conditions, training by use of the information about the one or more first video feature vectors again, so as to obtain the video data segmentation model. According to the embodiment of the invention, different video segmentation models are trained, so that automatic video data segmentation is realized, manual intervention operation is greatly reduced, the segmentation time is greatly reduced and the manpower cost is saved.

The invention discloses an image segmentation method based on cascade convolution, and relates to the technical field of image segmentation. The image segmentation method based on cascade convolutioncomprises the following steps of step S1, collecting medical image information of a plurality of lesion areas; step S2, performing artificial edge labeling on the lesion area of the collected medicalimage information layer by layer to obtain label information; step S3, carrying out standardized preprocessing on the label information to obtain a two-dimensional data set; step S4, establishing a multi-layer two-dimensional convolutional neural network based on cascade convolution, and training the multi-layer two-dimensional convolutional neural network by using a two-dimensional data set to obtain a neural network model; step S5, inputting medical image information of a patient to be segmented, and performing standardized preprocessing to obtain a two-dimensional data set to be processed;and step S6, inputting a to-be-processed two-dimensional data set into the neural network model, and automatically segmenting the medical image information of the to-be-segmented patient to obtain a lesion area of the patient.

The invention discloses a segmentation system based on a multi-scale double attention mechanism and a full convolutional neural network, which comprises a feature extraction backbone network, a feature pyramid network, a semantic segmentation network, a target detector, a coefficient predictor and a fusion module, and is characterized in that the semantic segmentation network comprises a first convolutional layer, an attention module and a second convolutional layer; the feature extraction backbone network is a VoVNet57 network and is used for extracting features of a training set image and a test set image and sending the features to the feature pyramid network; the feature pyramid network is used for performing same-level feature map fusion to obtain a P3-P7 feature map; a P3-P7 feature map obtained through the feature pyramid fusion network is input into an FCOS target detector, thus generating a suggestion box category and a position thereof pixel by pixel by the target detector, and performing a Soft NMS operation on the suggestion box to obtain a final detection box; a coefficient predictor performs weight prediction of instance information on the detection frame to generate an instance proportion corresponding to the detection frame; the semantic segmentation network is used for processing the P3-P6 feature map obtained through the feature pyramid fusion network to generate four segmentation maps; and the fusion module is used for superposing the four segmented images and the detection frame and outputting a final segmented image according to the corresponding instance proportion.

The invention discloses a magnetic resonance imaging blood vessel segmentation method based on a human brain tumor nuclear magnetic library. On the basis of a human brain magnetic resonance tumor segmentation image, an experience segmentation threshold for blood vessel segmentation is obtained from the human brain tumor nuclear magnetic library in accordance with data of a tumor area of the human brain magnetic resonance tumor segmentation image, the human brain magnetic resonance tumor segmentation image is analyzed to obtain a blood vessel gray threshold in accordance with the experience segmentation threshold, and the blood vessel gray threshold is adopted to perform blood vessel segmentation on the human brain magnetic resonance tumor segmentation image. By using the segmentation method, automatic blood vessel segmentation of a human brain tumor magnetic resonance image is realized, manual intervention is avoided, the segmentation accuracy is increased, and an algorithm for blood vessel segmentation of the human brain magnetic resonance tumor segmentation image is simpler, so as to be more favorable for system productization. Additionally, the invention further discloses a magnetic resonance imaging blood vessel segmentation system based on the human brain tumor nuclear magnetic library.

The invention discloses a lung anatomy position positioning algorithm based on a deep learning technology, which can accurately and quickly divide lung CT, and can simply, quickly and accurately realize automatic segmentation of lung lobes based on lung CT images, thereby realizing the anatomy position positioning of lung lesions. Compared with a traditional segmentation method, the method has theoutstanding advantages that (1) the process is simple, and the end-to-end segmentation mode does not need to pay attention to other processes; (2) the multi-stage and multi-output network architecture controls the network in different stages, so that the segmentation effect is better, and the segmentation precision can be ensured to the maximum extent through a semantic-based segmentation mode; and (3) the generalization ability is strong, and the data in the training process is enhanced, so that the model can learn different and diverse data, namely, the generalization ability of the segmentation model is ensured, meanwhile, the risk of over-fitting is also avoided to a certain extent, and the geometric deformation and illumination influence of CT (computed tomography) are insensitive when lung lobe division is performed on different CT.

The invention provides a bone joint CT image segmentation method and system. The acquired CT image is preprocessed so as to obtain the bone areas of the CT image; quasi-elliptic area locating is performed on each obtained bone area so as to obtain the bone areas imaged as the quasi-ellipsoids; at least one quasi-ellipsoid is selected from all the obtained quasi-ellipsoids according to the preset rules, and the selected quasi-ellipsoid is marked as the seed area; and the seed area is marked as the target area, and the non-bone areas are marked as the background areas. The connected bone area isseparated at the joints through the preset target area and the background areas by using the energy minimum segmentation algorithm based on the graph theory, and different identifiers are set for different bone areas so that CT image automatic segmentation can be realized and the efficiency of CT image segmentation can be enhanced.

The invention provides a remote sensing image road extraction method based on D-LinkNet, and the method comprises the following steps: S1, inputting a feature map into a D-LinkNet network, and completing the processing in an encoder sub-network based on a residual network and transfer learning; S2, inputting the feature map output in the step S1 into a feature extraction sub-network based on an expansion convolution and convolution block attention module for feature extraction; S3, enabling the feature map obtained after processing of the first two sub-networks to enter a decoder sub-network based on transposed convolution to realize image recovery. According to the invention, downsampling can be carried out on the road features in the remote sensing image, the problem of network degradation is well avoided, and the extraction of the road features is enhanced; the receptive field can be amplified by using the expansion convolution, the road characteristics in a larger range are sensedand the characteristics are extracted without increasing down-sampling, and the problem that the proportion of the road part in the remote sensing image is too small can be well solved.

The invention discloses a hyperspectral traditional Chinese medicine coated tongue quality classification method based on D-Resnet, and relates to the field of computer vision. Based on tongue coatingtongue quality classification of an RGB color space, the information amount is insufficient, a hyperspectral tongue image contains a large amount of spectral and spatial information, and the tongue coating tongue quality classification of a human body is realized by extracting the spectral reflectance change condition of a certain area of the tongue image and combining the spatial distribution information provided by the hyperspectral image. According to the method, an end-to-end D-Resnet network is provided to classify hyperspectral tongue images; the method comprises the following steps of:firstly, constructing a dense connection module to extract spectral information; then, a pre-activation bottleneck residual error module (pre-activation bottleneck residual error module) is constructed, so that space information can be extracted, and the spatial information of the pre-activation bottleneck residual error module can be obtained; according to the method, the coated tongue quality classification based on the hyperspectral image is realized.

The invention discloses a 3D vertebra CT image active contour segmentation method fusing weighted random forest, and relates to the field of medical image processing. For the problem of sensitivity ofa vertebra CT image segmentation method to an initial contour, a method for automatically locating a vertebra and segmenting a vertebra CT image is proposed. The method comprises the steps of firstly, proposing a weighted random regression and classification forest algorithm to determine a vertebra center; secondly, putting an initial contour ball of active contour segmentation in the vertebra center, and segmenting out the vertebra in the image by adopting a 3D active contour segmentation method in combination with an energy function; and finally, performing combination output on trained models to obtain a complete vertebra CT image segmentation model. A spinal CT segmentation model proposed in the method can automatically locate the vertebra center and can perform automatic 3D segmentation on the vertebra, so that the spinal CT image segmentation steps and processes are simplified.

The invention relates to the technical field of multimedia information processing, and provides a video reorganization system capable of automatically partitioning a shot and a video reorganization method thereof. The video reorganization system is as follows: a partition processor is respectively connected to an original video library and a reorganizer; and a generation and storage module is connected with the reorganizer. The video reorganization method comprises the following steps that: a user opens the original video library and extracts a video in need of being processed from the original video library; the partition processor performs automatic shot partition of the video, such that an independent video processing unit is obtained; a single-video effect library, a replacement background module and a multi-video operation module of the reorganizer are used for correspondingly processing a video processing unit respectively; and the processed video is transmitted to the generation and storage module and then stored and played. The video reorganization system disclosed by the invention is convenient to install and low in system hardware requirement and is capable of realizing operations, such as automatic shot partition, single-video effect processing, background replacement and multi-video mixing, on the same platform; therefore, the video processing efficiency of the video reorganization system is ensured; and thus, the video reorganization system has perfect functionality.

The invention discloses a medical image automatic segmentation method based on multi-spectral label fusion, in particular comprising the following steps: firstly, mapping the spectral gray image and the spectral label image to a target image through spectral registration; Then,selecting the map image blocks with high similarity to the target image blocks through the tag search area setting and thepreselection of the image blocks to be fused, so as to avoid the influence of the dissimilar map image blocks on the fusion result; Finally, acquiring the final image segmentation result by establishing a sparse representation model and target pixel point label assignment step, and dynamically setting the binarization threshold is according to the obtained fusion information in the target pixel point label assignment process, in order to reduce the number of pixel points with wrong label assignment and improve the segmentation accuracy. A method accord to that invention can obtain repeatableresults and is not affect by human subjective factors.