The lameness disease for diary cows have both been an economical and animal ware-fare problem for diary industry. Traditional lameness detection uses manual scoring which requires both large amount of time and resource and can be subjective to individuals. As a leading company in the area of farming and milking machines, DeLaval, has built up the system and collected videos of cows' walking movement. This thesis extracted cows' back curve from the videos from DeLaval's system and applied different machine learning algorithms, K-nearest neighbors(KNN), Large margin nearest neighbor(LMNN), Support Vector Machine(SVM) and Neural Network(NN). The classification results from each algorithm is compared, in which KNN, LMNN and SVM all achieved satisfactory results.