Monocular vision-based motion capture system: A performance model

Abstract

We propose a performance model for a monocular vision-based motion capture system. Such a system can use uniquely patterned augmented reality (AR) markers worn on the body. Two key factors in evaluating such a system are tracking accuracy and blurring effects. Past work involving AR markers has emphasized other factors such as detection rate or pixel error. In cases where accuracy has been studied, it has been done only for specific systems. In contrast, our model is more general and can accommodate different types of cameras and marker sizes. Our model can also simulate a marker worn on a moving limb. Preliminary experiments show that our model has the potential to accurately predict real-world performance.