Tracking A Group of Black Cows Using SORT based Tracking Algorithm

抄録

Tracking multiple objects that have the same color and texture features is a challenging task. The proposed system aims to track a group of black cows which are similar to each other with respect to colors and sizes by using the tracking-by-detection method. Before performing black cow tracking, the black cow’s regions with mask and bounding box are detected by using the robustness Detectron2 object detector. The Detectron2 outputs are then used as input of SORT based tracking algorithm to track the black cows by giving specific IDs. Although the traditional SORT tracking algorithm relies on bounding box information of the detection result, a fusion of mask and bounding box information from Detectron2 output is utilized in the SORT tracking algorithm for our proposed method.