Motion Stereo including Tracking Stability Analysis for Eye-in-Hand Systems

Abstract

This paper proposes a precise and fast approach to solve stereo correspondence problem in monocular eye-in-hand systems. Basic stereo methods use two static viewpoints. Some stereo methods use multiple static viewpoints. These methods improve accuracy performance, but take a lot of time in the monocular eye-in-hand systems. For a precise and fast measurement, our approach uses two static viewpoints and multiple dynamic viewpoints. Using the dynamic viewpoints, we take two steps. 1) propose two simple movement models of the eye-in-hand systems, 2) define tracking stability of each feature points in the models.