Multiple Object Tracking For Occluded Particles
IEEE ACCESS(2021)
摘要
The precise detection and tracking of multiple particles under a microscope are of significance in the research of the individual and cluster behavior of dynamic bacteria and subcellular structures. However, the existing detection algorithms cannot separate occluded particles from each other, and most of the tracking algorithms aimed to address the occlusion involve several uncertainties. In this paper, a two-step detection algorithm based on the threshold segmentation and morphological open operation has been developed for identify non-fluorescent labeled particles under microscope, which could separate micro-contact targets. Moreover, we have proposed a novel correlation algorithm that can exploit the strengths of the global shortest path algorithm and Hungarian algorithm, which updating online in real time and considering the occlusion among particles. The proposed approach could achieve the temporal optimal match and spatial optimal solution by utilizing the multi-frame information. Moreover, the proposed method could realize the tracking of occluded particles tracking, and outperform the single global shortest path algorithm and Hungarian algorithm. The proposed method was successfully applied to six real image sequences with the maximum number of particles per frame ranging from 23 to 55, as well as a synthetic and fluorescent labeled sequence. The results of the contrast experiments demonstrated that the proposed algorithm is practical and can realize real-time tracking.
更多查看译文
关键词
Microscopic image, multiple particles tracking, target occlusion, global data association
AI 理解论文
溯源树
样例
![](https://originalfileserver.aminer.cn/sys/aminer/pubs/mrt_preview.jpeg)
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要