Zebrafishtracker: A multi-zebrafish tracking algorithm can effectively solve cross occlusion
Zhenhua Fu, Yiwen Wang, Xin Jie Zhao, Mingzhu Sun · 2021 IEEE International Conference on Robotics and Biomimetics (ROBIO) · 2021
Zebrafish is a model organism, which is widely used in life science and medicine research due to its high homology with human genes. How to handle cross-cases and achieve more comprehensive tracking is always a challenge. Zebrafishtracker proposed in this paper uses the instance segmentation and the skeleton stability analysis strategy to effectively solve the problem of detection error caused by the frequent crossover of multiple zebrafish. Then 2D tracking of zebrafish is realized based on identity information. The results show that in two relatively complex videos, MOTA of our algorithm in the top-view reaches more than 95%.