Research on the recognition of pig behavior based on contour features

In the modern pig farms, the detection of the early symptoms and abnormal behaviors for sick pigs usually relies on manual observation. This method is labor-intensive and could not find sick pigs in time. In order to solve these problems, a target recognition method based on contour features was proposed to classify pigs' behavior. Firstly, the contour of moving objects was extracted by background subtraction, and the hue and saturation information of HSV color space model was used to eliminate the influence of shadow on the target detection. Secondly, a model of contour eigenvector was built based on the edge invariant moments and morphologic features. Finally, pigs' behaviors were classified into four categories: normal standing(walking), drooped standing, high spirited standing, and lying. It was realized by analyzing and comparing the Euclidean distance between the contour eigenvector and each standard template. The experimental results show that the behaviors of pigs could be detected using this method. The recognition accuracy is above 80%. This study has provided a valuable exploration on the recognition of abnormal behaviors of pigs in modern farms.