Indoor exploration using a µUAV and a spherical geometry based visual system

Abstract

This research presents a new vision system that explores a spherical geometry and will be provide innovative solutions for tracking, surveillance, navigation and mapping with micro Unmanned Aerial Vehicle (µUAV) in unknown indoor environments. The system will be used with µUAV in indoor environment and it is composed by twenty six cameras that are arranged in order to sample different parts of the visual sphere around the µUAV. This configuration allows that some of the cameras will have overlapped field of view. This system has been designed for the purpose of recovering ego-motion and structure from multiple video images, having a distributed omnidirectional field of view. We use the spherical geometry to extend the field of view, from one single direction to a single point of perspective, but with multiple views. This manuscript will prove that spherical geometric configuration has advantages when compared to stereo cameras for the estimation of the system’s own motion and consequently the estimation of shape models from each camera. The preliminary field tests presented the theoretical potential of this system and the experimental results with the images acquired by 3 cameras.