Identification of group-housed pigs based on Gabor and Local Binary Pattern features

Abstract

A novel method for the identification of group-housed pigs based on machine vision is proposed. It benefits to the automatic detection and analysis of the behaviour of pigs. Top-view videos of pigs were obtained and the images of individual pigs extracted. The Gabor features were extracted by convolving pig images with Gabor filters and the local structural features using the Local Binary Pattern (LBP) identification. Principle Component Analysis (PCA) was then used to reduce the feature dimension and the features were concatenated to form the feature vectors. In order to evaluate the performance of the proposed method, standing posture images of pigs were used to conduct the experiments in terms of Support Vector Machine (SVM) classification. The experimental results demonstrated that the combination of Gabor and LBP features produced better results. The average recognition rate achieved 91.86% by SVM with a linear kernel and the PCA parameter varied from 0.85 to 0.99.