Proper

Abstract

In a surveillance camera environment, detecting pigs in a pig pen is an important issue in terms of automatic management. In this study, we focus on depth-based standing-pigs detection with “moving noises” which appear every night in a commercial pig farm, but have not been reported yet. We apply a spatiotemporal interpolation technique first in oder to remove the moving noise occurred in the depth image. Then, we detect the standing-pigs by utilizing the undefined depth values around a standing-pig. Our experimental results showed that this method is effective for detecting standing-pigs at night in terms of both cost-effectiveness (using a low-cost Kinect depth sensor) and accuracy (pig-level and pixel-level accuracies over 100% and 93%, respectively), even with severe moving noises occluding a half of an input depth image.