This paper presents the different unsupervised machine learning algorithms used for Ketosis detection, based on the color characteristics taken from the Ketocheck rapid colorimetric test. The level of ketone bodies in bovine’s urine is represented by three color categories, range of dark green (right ketone level), green (normal range ketone level) and yellow/orange (higher ketone level). The color image is converted into HSV color space for better color discrimination. The proposed technique enables detection of ketosis by clustering every pixel in the image using unsupervised K-means clustering and Fuzzy C Means (FCM) clustering algorithms. The results obtained have shown that K-means algorithm is faster and it also have low computational complexity. Two android application is developed. One with K-means clustering algorithm loaded in server and the other, directly programmed in android application.