Identification of Cattle Breeds by Segmenting Different Body Parts of the Cow using Neural Network

Breeds of dairy cattle are important to distinguish a steers for some reasons one of them is to forestall the deficiency of hereditary variety in domesticated animals’ populaces. The cattle breed is predicted by using different parts of the cattle as input. The body parts of the cow like hump, body size, muzzle, face and horn pattern are used for the identification of cow breed. The steps of this process consists of pre-processing of the cattle image, feature extraction, region based segmentation, classification and prediction from the cattle’s part in a single image. In our proposed method, the breed of the cow will be detected using the various body parts of the cows. The body parts used will be hump, horn, muzzle pattern and body pattern. The rare cattle breed will also be predicted if it is found in the image. The techniques which is used are CNN algorithm, SIFT techniques used for feature extraction along with region based segmentation are used. The multi SVM is used for identifying the group of cattle at the same time.

I. Introduction

Now days the breed prediction of cattle is the most difficult process and necessity. In this on-going field of dairy farming many research are going on. Still, it is difficult in categorizing the cattle into its particular breed using the existing techniques like RFID, muzzle pattern recognition, sensor based ear tagging. In the RFID and sensor-based ear tagging technique the sensors can easily get damaged by collision between the cattle when they are in groups, and it also causes high frequency radiation which will be harmful for the cattle ‘ s health. When the muzzle pattern technique is used the main drawback of this technique is that the deformation that occurs in their muzzle can predict the wrong breed of the particular cattle that will be used for the identification.