3D zebrafish tracking with topology association

Recently, zebrafish has received more and more attention due to its wide range of applications such as regeneration promoting therapeutics and drug discovery. Therefore, vision-based trackers are utilized to record the swimming trajectory of zebrafish. In this paper, a re-association method is introduced in the 3D reconstruction process to generate missed targets caused by occlusion. Since the variation of the overall tracking targets has the property of continuity and stability, a topology association model (TAM) is proposed by point group similarity into the tracking framework. TAM describes the movement of zebrafish from the macroscopic level and utilizes the changes of the point group structure for tracking. Experimental results show that the tracking framework enhances the overall performance and promotes the trajectory integrity. On the latest 3D-ZeF20 benchmark, state-of-the-art results are achieved. In addition, TAM tracking framework is applied to 2D general tracking to prove that the method is useful and have great advantage in other scenarios with relatively stable amount of targets as well.