Image Understanding Research in Complex Environments

摘要

We have undertaken a broad range of research for developing image understandingtechnique to infer 3-D shape descriptions of-the scene front sensed data which may consist of monocular images, stereo pairs, motion sequences or range data, to recognize the objects in the scene, and to keep integrated temporal descriptions in dynamic scenes. This report details recent work in generating 3-D descriptions from range data or intensity data, object, recognition using perceptual grouping and pose estimation, and representing and recognizing landmarks for indoor navigation. (KAR) P. 1.