The Image Processing Technology Of The Milk Somatic Cells Based On The SOM Neural Network

In the cell micro image processing, digital image processing takes as a contemporary important means of the cell image research and an important basis for clinical diagnosis and treatment. Therefore, the study for the image processing technology of the milk somatic cells includes: the preprocessing, segmentation, count of the milk somatic cell images.The noise is removed through the bilateral filtering method. The milk somatic cell count images are segmented in the RGB color space. Experiments show that the segmentation accuracy using the SOM network in RGB color space is higher but the speed is not fast.Use the image edge contrast method to select four color spaces, and propose three segmentation methods 3PCA, 2PCA, 1PCA based on the principal component analysis. Through a large number of test and comparative analysis, the accuracy is the highest using the 3PCA method. Consideration from the speed and accuracy of the segmentation, the 1PCA method and the 2PCA method overcomes the large amount of calculation and the slow speed of the 3PCA method, but also overcomes the inaccurate disadvantage of the gray method. The 1PCA method or the 2PCA method can be applied in practice.The ultimate goal of segmentation is to count fastly and accurately. The milk somatic overlapping cells after the segmentation need to separate. In accordance with the count guidelines, the Fire Spread method is cited to count the cells. Experiments show that the count method is fast and accurate for milk somatic cell count images.The segmentation of the milk somatic cell images is realized under C++ Builder, and has been verified in MATLAB.