Milk Somatic Cells Recognition based on Gray-Scale Difference Statistics

Abstract

In order to solve the problems in recognition of milk somatic cells and improve the recognition efficiency and accuracy, we put forward a classification algorithm based on the gray-Scale difference Statistics(GLDS). First, the algorithm extracts the gray difference matrix of the image, and from the matrix we can calculate the texture features of cell images. Then uses the K-means algorithm to segment the cells, extracts morphological features from the nucleus. We extract the 8 morphological feature parameters, 4 texture feature parameters, and use the K-fold cross-validation and random forest classification(RF) to classify samples. Experimental results show that the correct rate is 95.36% when using the random forest classifier.