Abstract
Conventional cell tracking methods detect multiple cellsin each frame (detection) and then associate the detec-tion results in successive time-frames (association). Mostcell tracking methods perform the association task indepen-dently from the detection task. However, there is no guar-antee of preserving coherence between these tasks, and lackof coherence may adversely affect tracking performance. Inthis paper, we propose the Motion and Position Map (MPM)that jointly represents both detection and association for notonly migration but also cell division. It guarantees coher-ence such that if a cell is detected, the corresponding mo-tion flow can always be obtained. It is a simple but powerfulmethod for multi-object tracking in dense environments. Wecompared the proposed method with current tracking meth-ods under various conditions in real biological images andfound that it outperformed the state-of-the-art (+5.2% im-provement compared to the second-best).
Abstract (translated)
URL
https://arxiv.org/abs/2002.10749
