- Conference date: 17–19 December 2007
- Location: Gold Coast, Queensland (Australia)
This paper leads to a novel technique for tracking and identification of zebra‐fish cells in 3D image sequences, extending graph‐based multi‐objects tracking algorithm to 3D applications. As raised in previous work of 2D graph‐based method, separated cells are modeled as vertices that connected by edges. Then the tracking work is simplified to that of vertices matching between graphs generated from consecutive frames. Graph‐based tracking is composed of three steps: graph generation, initial source vertices selection and graph saturation. To satisfy demands in this work separated cell records are segmented from original datasets using 3D level‐set algorithms. Besides, advancements are achieved in each of the step including graph regulations, multi restrictions on source vertices and enhanced flow quantifications. Those strategies make a good compensation for graph‐based multi‐objects tracking method in 2D space. Experiments are carried out in 3D datasets sampled from zebra fish, results of which shows that this enhanced method could be potentially applied to tracking of objects with diverse features.
Data & Media loading...
Article metrics loading...