Tracking Algorithm Based on Multi-feature Detection and Target Association of Pigs on Large-scale Pig Farms

摘要

In order to track pigs intelligently, we propose an algorithm based on multi-feature detection and target association. Firstly, the color feature and texture feature are used in pig self-adaptive detection, and the image segmentation using edge features is built to get the mask images of different pigs. Secondly, the correlation matrix of pig regions between previous and current frame is calculated, and the Kuhn-Munkres algorithm is adopted to determine the association between pigs and their image regions. The experiment results demonstrate that this pig tracking algorithm has relatively strong robustness and meets the real-time requirement.