The Segmentation Of Overlapping Milk Somatic Cell Based On Watershed Algorithm

The quality of milk and the health level of the daily cow rely on the number of milk somatic cells. The correct image segmentation of milk somatic cells determines the accuracy of automatic recognition and analysis of somatic cells. Meanwhile, it can also improve the quality of the milk and the accuracy and efficiency of mastitis checks on the daily cow. However, the structure complexity, the size variety and the conglutination and overlap of the milk somatic cells have impacts on the accuracy of cell image segmentation.This essay will provide a method of image segmentation for overlapping cells, based on the features of images of milk somatic cells and the existing techniques. The first step is to recognize the existence of overlapping cells in the images, according to the shape factor. Then, converting the target images into gray images which consist of only cell nucleus, cell cytoplasm and the background by k-means and extracting the cell image (cell nucleus and cell cytoplasm). The binary images of the target images are under image preprocessing by mathematical morphology operators-eliminating noise and debris, filling empties and smoothing edges. Finally, operating image segmentation by means of distance transformation and watershed Algorithm based on image reconstruction. Experimental results show that this method is highly effective in image segmentation of milk somatic overlapping cells, which can basically resolve the problem of over segmentation in the watershed Algorithm and provide efficient techniques and strategies for further image analysis and cell counting.