Research On Segmentation And Separation Method Of Milk Somatic Cell Image

At present, the identification and analysis of cells has become an important means in clinical diagnosis. Similarly, The diagnosis and analysis of milk somatic cell is the main way to dectect the cow mastitis, However, the key step is to count the milk somatic cells. Currently, The most common method for cell counts is by artificial reading. But this meathod is affected by subjective factors, the labor intention is great, and low working efficiency which lead to the lack of accuracy and objectivity of the analysis results. With the digital image processing technology and pattern recognition technology is widely used in computer, using the modern image processing technology to automate recognition and analysis the milk somatic cell images has become a hot issue in the field of cytology.According to compare the domestic and foreign cell image segmentation methods and the related research status, summaring their respective advantages, disadvantages and the present problems. The paper is mainly concentrates on researching the color milk somatic cell image preprocessing, image segmentation, the overlapping cells separation and so on, which provide the technical support for the automatic recognition technology of milk somatic cell image.The main content of the paper is as follows:(1) Using the basic theory and method of image processing technology, preprocessing the color milk somatic cell image which can remove the noise and the other irrelevant information to get the most suitable image.(2) Duing to the image segmentation methods of 3-D color spaces will consume large amount of computations and the accuracy of segmentation is not high in the low spaces, In this paper, the method of milk somatic cell image segmentation based on dimension reduction and fusion is proposed. Through analysising each color component of the color image, selecting the appropriate low dimensional space to implement K-means segmentation and using the region splitting and merging to fuse the result of image segmentation in the low dimensional space. So can achieve the color milk somatic cell image segmentation. Through experimental test and evaluation, this method has high accuracy and fast processing speed which lays the foundation for the separation of overlapping milk somatic cells image.(3) Aiming at the overlapping milk somatic cells image, a separation method for overlapping somatic cells based on the gradient information Hough circle detection is proposed. Firstly, using preprocessing method to the foremost segmentation somatic cell image, and the information of the target edge is extracted with the help of the 8-neighborhood chain code. Secondly, Separating the overlapping somatic cells by using the gradient information Hough circle detection, because of the parameter setting or other influence factors, the separation results have many false cells. Finally, removing false cells by established authenticity criterion of cells, and achieving the final separation of overlapping cells.