Semi-automatic interactive modeling for model-based camera tracking

Abstract

Object modeling is an important task as a pre-processing of model-based camera tracking. In this paper, we present a semi-automatic interactive modeling method to create 3D object models with simple primitives such as points and lines. Our approach is based on feature-based modeling that uses 3D point clouds of target scenes. Instead of using graphic modeling tools, the modeling is intuitively performed by user participation in video sequences. Finally, we analyze sensitivity of camera tracking according to modeling errors using Cramer-Rao Lower Bound.