A non-feature-based model for range identification

Abstract

The correspondence problem refers to the problem of ascertaining which parts of one image correspond to which parts of another image. Correspondence problem usually needs to be solved by feature extraction and feature matching first, as a prerequisite for vision-based control and estimation. To avoid this bottleneck, assuming that an object is on a plane or has a planar surface, we present a model for non-feature-based object range estimation based on the theory of mutational analysis for shapes. The proposed model will not fail even when the object is partially out of the camera view.