Object Modelling with Superquadrics: Fusing Data from Camera and Laser Radar (Objektanpassning med Superquadrics: Foerena Data fran Kamera och Laser Radar)

摘要

The 3D shape of a command vehicle is modelled by a superquadric, a generalized quadric surface with non-integer exponents. The model is fitted simultaneously to range data from a laser range scanner and silhouette points extracted from a video picture of the vehicle. A good fitting, even to rather scattered range data, is achieved using some a priori knowledge about the object shape and orientation. The recovered model orientation is checked against the vehicle's path through four sequential positions. It is clear that superquadrics is a useful class of models for model based signal analysis, when data are noisy and scattered.