Bio-inspired reconfigurable stereo vision for robotics using omnidirectional cameras

TL;DR

This paper presents a bio-inspired, reconfigurable stereo vision system for robotics that uses omnidirectional cameras and advanced algorithms to adapt to different scenarios, improving flexibility and accuracy.

Contribution

It introduces a novel reconfigurable stereo vision strategy with a deep neural network and geometrical compensation, enabling dynamic camera alignment and enhanced visual performance.

Findings

Demonstrated adaptability on a metamorphic robot with multiple configurations.

Achieved improved depth measurement accuracy with the proposed system.

Showed effective switching between wide and narrow field configurations for different tasks.

Abstract

This work introduces a novel bio-inspired reconfigurable stereo vision system for robotics, leveraging omnidirectional cameras and a novel algorithm to achieve flexible visual capabilities. Inspired by the adaptive vision of various species, our visual system addresses traditional stereo vision limitations, i.e., immutable camera alignment with narrow fields of view, by introducing a reconfigurable stereo vision system to robotics. Our key innovations include the reconfigurable stereo vision strategy that allows dynamic camera alignment, a robust depth measurement system utilizing a nonrectified geometrical method combined with a deep neural network for feature matching, and a geometrical compensation technique to enhance visual accuracy. Implemented on a metamorphic robot, this vision system demonstrates its great adaptability to various scenarios by switching its configurations of 316{\deg} monocular with 79{\deg} binocular field for fast target seeking and 242{\deg} monocular with 150{\deg} binocular field for detailed close inspection.