Tolerance analysis of model-based camera tracking by modeling errors

Abstract

Accuracy of 3-D models is important in model-based camera tracking. However, their precise modeling requires delicate procedures. On the other hand, end-users are difficult to perceive tracking errors, which are caused by a certain level of modeling errors. Therefore, human perceptual errors can be different from real tracking errors. In this paper, we analyze the tolerance of modeling errors by comparing computational matching errors of model-based camera tracking with its human perceptual matching errors through user evaluations. We also discuss permissible ranges of modeling errors for reliable tracking without precise modeling.