44results about How to "Prevent oversegmentation" patented technology

The invention relates to the field of medical image processing and aims to provide a fully-automatic three-dimensional liver segmentation method based on a convolution nerve network. The fully-automatic three-dimensional liver segmentation method based on the convolution nerve network comprises the following processes: preparing a training set, and training the convolution nerve network, processing CTA volume data of an abdominal liver by utilizing the trained convolution nerve network to obtain a segmentation result of the liver. The liver is segmented by means of the three-dimensional convolution nerve network; the three-dimensional liver segmentation method is fully automatic and can also prevent under-segmentation and over-segmentation phenomena well; and the accurate liver segmentation result is obtained.

The invention discloses a coronary artery CT (computed tomography) contrastographic image calcification point detecting method. According to the method, the existing coronary artery central axis is utilized, firstly, the local structural features of each voxel point in an interested region of a blood vessel are extracted, then, the spherical harmonic conversion is utilized for quantizing the local structural features, feature vectors are obtained, finally, a classification algorithm is adopted for classifying the obtained feature vectors, the local structural feature approximation degree of the voxel points to the calcification points, blood vessel lomens and image backgrounds in a training dataset is determined, and calcification point detecting results are finally obtained. The method has the advantages that calcification points positioned on the coronary artery blood vessel walls in the coronary artery CT contrastographic image can be precisely positioned, and the efficiency and the accurate rate of the computer auxiliary diagnosis are improved.

The invention relates to processing of a medical image and in particular relates to a method for enhancing the liver blood vessel and simultaneously dividing the liver from the blood vessel in a CTA (computed tomography imaging) image. The method comprises the following steps of: preprocessing an image by means of gaussian convolution; computing the anisotropic characteristic value of each point of the image, and further conforming the anisotropic oval neighbourhood of each point; computing a grey level histogram in each neighbourhood, initiating a liver region, and computing a grey level histogram in the liver region; computing the wasserstein distance between the grey level histogram in each point neighbourhood and the grey level histogram in the liver region; enhancing the blood vessel according to the wasserstein distance and the anisotropic characteristic of the neighbourhood; and dividing the liver from the blood vessel. According to the method provided by the invention, the negative effects caused by the low contrast ratio, the noise, the fuzzy boundary and the like can be overcome, and the blood vessel identifying and dividing accuracy rate can be greatly improved, thus the anatomical structure information of the liver blood vessel can be exactly obtained.

The invention discloses an improved adhesion particle target segmentation method based on concave point matching, and belongs to the technical field of image processing. The method comprises the following steps: firstly, carrying out image preprocessing to obtain a contour and a concave point of a to-be-segmented target; then, carrying out preliminary segmentation based on morphological operation, so that the number of pits for matching processing is effectively reduced; thirdly, performing segmentation processing based on local concave point matching, i.e., performing local concave point matching processing first, and then realizing first segmentation processing based on a matching result; and finally, carrying out second segmentation processing based on distance transformation processing, and the technical problem of isolated pits is solved. The method can be applied to the technical fields of agricultural seed counting, segmentation and the like, and is high in segmentation accuracy.

The invention relates to a power plant high-temperature pipeline defect detection and segmentation method based on OTSU and a region growing method, and the method comprises the following steps: obtaining an original infrared image, carrying out the graying, and extracting a pipeline region through the improved two-dimensional OTSU pre-segmentation; finding out a gray scale range corresponding tothe normal temperature of the pipeline through the gray scale distribution histogram of the pipeline area; judging whether the maximum value of the neighborhood gray average value exceeds a normal gray range or not to finish automatic detection and positioning of the multi-defect seed points; and taking the self-adaptive threshold adjusted according to the gray average value and the standard deviation of the grown region as a growth criterion, and taking a gradient amplitude threshold calculated based on a Prewitt operator as an additional condition to realize growth and extraction of the defect region. Compared with the prior art, the system has the advantages of high reliability and accuracy and good real-time performance, and can meet application requirements.

The invention relates to a superpixel-based method for parallel segmentation of a remote sensing image. The method comprises: a false color image is synthesized at a combined waveband; band-shaped block division is carried out on the false color image; parallel SLIC segmentation is carried out on all sub blocks; a regional adjacency map at a boundary is established; boundary region combination iscarried out; a region tag is updated; full-image tag recombination is carried out; and a segmentation result is stored. According to the invention, on the basis fusion of false color image generation,parallel segmentation, regional heterogeneity evaluation, boundary combination processing, and run length code compression and storage and the like, the high accuracy of the segmentation edge is ensured, the segmentation speed is increased, and the method is suitable for realizing engineering application of the algorithm.

The invention discloses an intelligent monitoring system for a power cable production process. The system comprises an image acquisition assembly, a sensor assembly, a wireless transmission module, anintelligent monitoring center and a danger alarm module. The sensor assembly is used for collecting environmental data and temperature data of the surface of power cable production equipment And transmitting the acquired data to the intelligent monitoring center; the intelligent monitoring center processes the received data and compares the processed data with a preset safety threshold value; andwhen the data is higher than a preset safety threshold value, the danger alarm module gives an alarm, image acquisition is carried out on the position where the sensor node for acquiring the data islocated through the image acquisition assembly, and the acquired image is displayed after being processed. The beneficial effects of the invention are that the system achieves the effective monitoringof the environment of a power cable production workshop and the operation state of power cable production equipment, and improves the monitoring intelligence of a power cable production process.

The invention discloses a cell nucleus segmentation method based on attention learning, and relates to the cell nucleus segmentation problem in intelligent pathological diagnosis. The intelligent pathological diagnosis technology utilizes a deep learning technology to segment and identify abnormal cells in a cell image. However, there are few models for cell nucleus segmentation in cell images. The following problems exist: (1) the problems of cell nucleus overlapping and unobvious boundary are not considered, so that the segmentation precision is low; and (2) the contextual information of thecell nucleus edge is not considered, so that under-segmentation or over-segmentation is caused, and the subsequent classification result is influenced. Therefore, the invention provides a cell nucleus segmentation method based on attention learning. Experiments show that the model can effectively solve the problem of segmentation of overlapped cells, and the problems of under-segmentation or over-segmentation of unclear boundaries and the like. The invention is applied to cell nucleus segmentation in intelligent pathological diagnosis.

The invention relates to medical image processing, and aims to provide a method for simultaneously segmenting liver and a blood vessel in a CTA (computed tomography angiography) image. The method comprises the following steps of (1) performing model initialization; (2) optimizing a new variational energy model, and simultaneously segmenting the liver and the blood vessel; (3) performing post-processing to obtain a liver contour and a blood vessel contour respectively. According to the method for simultaneously segmenting the liver and the blood vessel, the adverse impact of weak boundaries, noise and the like can be effectively overcome; particularly, due to the introduction of gray level distribution information of local neighborhoods, the liver can be well distinguished from adjacent organs or soft tissues, and the phenomenon of over-segmentation can be effectively prevented.

The invention belongs to the field of image segmentation, and provides a leaf image segmentation method and system based on color and morphological characteristics. The leaf image segmentation methodbased on color and morphological characteristics comprises the following steps: clustering crop plant images, reserving a foreground region to which a leaf part belongs, and removing a background region; screening out crop parts by using an ultra-green algorithm, and removing residual background regions; removing a weed area based on a processing method of a color difference value; removing adhered weed parts by adopting an area threshold method; and repairing and removing the image of the adhered weed part by using a closed operation, and finally obtaining a leaf image segmentation result.

The invention discloses a screening method for candidate nodules based on CT images and relates to the field of medical image processing. The screening method comprises the following steps: step 1, obtaining a to-be-detected lung CT image f0(x, y); step 2, binarizing the lung CT image f0(x, y) in the step 1 and extracting a lung parenchyma preliminary template f1(x, y); step 3, performing lung parenchyma repair and calculation on the lung parenchyma preliminary template f1 (x, y) to obtain a lung parenchyma region image f2 (x, y); step 4, binarizing the lung parenchyma region image f2(x, y) and screening candidate nodules to obtain a candidate nodule set R; step 5, constructing a 3DCNN network structure for classifying candidate nodules; step 6, training the 3DCNN network model according to the 3DCNN network structure and the library lung CT image in a LIDC lung image database; and step 7, determining the probability of each candidate nodule in the candidate nodule set R being a pulmonary nodule according to the 3DCNN network model.

The invention provides an image segmentation method based on region correlation. The method comprises the steps of automatic level set function initializing and level set evolution based on region correlation. Firstly, unlike a traditional level set method, the step of automatic level set function initializing automatically generates a smooth initial contour near a target boundary through sequence morphological filtering and a Gaussian mixture model, which effectively solve the problems of target boundary irregularity and image noise. Secondly, the step of level set energy function regularizing prevents the problems of weak boundary leakage and uneven gray value of a target region by calculating the correlation of the histograms of inner and outer regions of the level set contour. Furthermore, an improved boundary stop function is used to speed up the convergence of the level set function, which effectively prevents the phenomenon of boundary leakage caused by excessive evolution of the level set contour.

The invention discloses a multi-source path shortest distance-based interactive image segmentation method. The method comprises the steps of S1, calculating a side weight between pixel points in an image according to pixel value features of the image, and converting the image into an undirected weighted graph according to the side weight between the pixel points; S2, marking a plurality of pixel points as mark points of to-be-segmented objects in the to-be-segmented objects of the image, wherein the mark points of the to-be-segmented objects carry identifiers of the to-be-segmented objects; S3, searching for the mark points closest to unmarked pixel points according to the undirected weighted graph for all the unmarked pixel points in the image, thereby converting an image segmentation problem into a multi-source shortest path search problem; and S4, marking the unmarked pixel points with the identifiers of the mark points closest to the unmarked pixel points, obtaining the identifiers of the pixel points in the image, and marking a boundary of the pixel points with different identifiers as a segmentation boundary. According to the method, a more ideal and high-quality image segmentation result can be obtained.

The invention discloses a ground point cloud segmentation method and system, a ground modeling method and a medium, and belongs to the technical field of automatic driving, and the method comprises the steps: carrying out the preprocessing of ground point cloud data, and dividing the data into a ground two-dimensional grid; determining grid features corresponding to the ground two-dimensional grids through a pre-trained ground network model; processing the grid features through a ground network model to obtain a grid height prediction value; comparing the grid height predicted value with a corresponding grid height threshold value, and if the grid height predicted value is greater than the grid height threshold value, determining the corresponding ground point cloud data as an obstacle point; and if the grid height prediction value is not greater than the grid height threshold value, determining the corresponding ground point cloud data as a ground point. According to the method, the height of each point cloud grid is calculated through the pre-trained ground network model, the height is compared with the threshold value corresponding to each grid, whether the point cloud grid is a ground point or not is judged, and the accuracy of point cloud segmentation is improved.