Abstract: In order to deal with various background situations, the complex lighting situations and the goals-background-mixing situations in piggery, a new kind of tracking method which is based on traditional compressive tracking algorithm was proposed. Firstly, to reduce the tracking error, we changed the search window to oval which is closer to the pig body. Secondly, to increase the stability of feature extraction and reduce drift, we combined gray feature with texture feature, and improved the random measurement matrix of traditional compressive tracking algorithm. Lastly, the piggery was divided in different areas. Based on the location of the target pigs we can analyze and assess its current behavior. Test results of different video samples and tracking results show that this algorithm improves the accuracy significantly in the piggery scene. The mean value and the variance of central point error in the proposed method were 25.44, those were 60.32%，33.33%，32.57% of the mean value of central point error in the CT method, TUT method and Camshift method. The tracking rate and it reaches to 19.3 frame/s.