A real-time computer vision assessment and control of thermal comfort for group-housed pigs

A real-time image processing system was developed to detect movement and classify thermal comfort state of group-housed pigs based on their resting behavioral patterns. This paper describes the theory, system structure, selection and analysis of image features, and image classification techniques. Image moment invariants, run-length frequency, pig body occupation ratio, and pig group compactness are extracted as feature vectors. Minimum Euclidian distance was used to distinguish cold vs. comfortable state of the pigs; whereas blob analysis was used to identify warm/hot state of the pigs. A sliding window was employed to update reference image feature sets so that classification is always based on the most recent information. The prototype system was initially developed with paper-cut pigs, followed by tests with live pigs. The results showed that this system effectively detects animal movement, and correctly classifies animal thermal behaviors into cold, comfortable, or warm/hot conditions. It also has the ability to adopt itself to different body weight or sizes of the pigs.