Subspace and motion segmentation via local subspace estimation

Abstract

Subspace segmentation and clustering of high dimensional data drawn from a union of subspaces are important with practical robot vision applications, such as smart airborne video surveillance. This paper presents a clustering algorithm for high dimensional data that comes from a union of lower dimensional subspaces of equal and known dimensions. Rigid motion segmentation is a special case of this more general subspace segmentation problem. The algorithm matches a local subspace for each trajectory vector and estimates the relationships between trajectories. It is reliable in the presence of noise, and it has been experimentally verified by the Hopkins 155 Dataset.