Error Analysis of a Real-Time Vision-Based Pointing System

Abstract

In this paper, an efficient error analysis of a real-time vision-based pointing system is proposed. We use two cameras to implement the pointing system according to a simplified 3D reconstruction scheme which is based on image feature extraction, homography, and 3D geometry. Similar to other 3D reconstruction approaches, reconstruction errors also exist in the proposed pointing system. We investigate the relation between image noise and the ultimate reconstruction errors, and develop efficient methods to find the worst-case error range in the latter with respect to a fixed magnitude of the former due to each camera. Experimental results show that the proposed approach can find the error range satisfactorily. Thus, users of similar pointing systems can get a more robust pointing result by identifying a special pointer position, or possibly a special pair of cameras, that will result in minimal range of pointing error.