An improved randomized ellipse detection algorithm applied in the swine gesture identification

Abstract

Based on ellipse characteristic of porcine contour, a simple gesture recognition algorithm was proposed to distinguish different gestures and mental states. Firstly, the porcine image was pretreated to detect edge. And all the points on the edge were fitted with an ellipse. Then, the eigenvectors of porcine gestures were determined according to the features of its head and neck, trunk, limbs in the spatial distribution. Additionally, the classifier base on support vector machine was used to classify different gestures into three categories: normal standing, standing with drooped head and lying. Finally, as different gestures corresponded to different mental states, the porcine mental state in the image was obtained. This method was adopted in the experiment to deal with the images form the self-building database. The experimental results demonstrate the validity of the above method.