An Insect-Inspired Active Vision Approach for Orientation Estimation with Panoramic Images

摘要

We present an insect-inspired approach to orientation estimation for panoramic images. It has been shown by Zeil et al. (2003) that relative rotation can be estimated from global image differences, which could be used by insects and robots as a visual compass. However the performance decreases gradually with the distance of the recording positions of the images. We show that an active vision approach based on local translational movements can significantly improve the orientation estimation. Tests were performed with a mobile robot equipped with a panoramic imaging system in a large entrance hall. Our approach is minimalistic insofar as it is solely based on image differences.