Hierarchical feature grouping for multiple object segmentation and tracking

Abstract

In this paper, we propose a hierarchical feature grouping method for multiple object segmentation and tracking. The proposed method aims to segment and track objects in the object-level without prior knowledge about the scene and object. We firstly group the motion feature into region-level with the proposed region features which represent a homogeneous region in an object. Object-level groups are achieved by clustering the region-level groups based on foreground information and motion similarity. To find optimal object-level groups, we formulate energy minimization problem, design its objective functions and solve it using simulated annealing(SA). By this hiearchical feature grouping, the proposed method efficiently segments and tracks various kinds of objects without object detector, 3D model and geometry information. Experimental results on several video clips show that our approach robustly segments and tracks multiple object regardness of camera position and object-class.