Range Image Matching for Object Recognition in Real Scene

Abstract

Object recognition is very important task in the field of computer vision. In this paper, we present a range image matching for the object recognition in real scene. The matching is the process of comparing a range image with a stored shape data. We first discuss that the difference of resolutions of them must be considered for the range image matching. Then, we analyze the apparent object shape change of the range image due to the measurement distance in the relation to the property of the range finder. We claim that the apparent shape change can be modelled by a smoothing process of the object surface. We propose a method of matching after shape smoothing of the stored data to cope with has a very low resolution of the range image. We also propose a generation of a surface model of the given data, considering the resolution difference of a newly obtained data of the target object. Finally, we show several experimental results which promises the feasibility of the proposal method.