A Novel Image-Based Navigation System using Stereo Visual Odometry for Indoor and Outdoor Navigation Applications

This paper presents a novel image-based navigation strategy using stereoscopic image sequences for aiding indoor or outdoor navigation systems. Initially our image-based navigation approach is designed for robust indoor navigation. To demonstrate the powerful capabilities of Visual Odometry the new image-based navigation application is evaluated with various stereo image sequences. We compare the performance of the system using different kinds of image processing options to optimize the accuracy of the estimated trajectory and in terms of keeping the computational load low. The experimental platform for personal navigation and the results from the Stereo Visual Odometry are presented. All results are built on real image sequences recorded with our experimental, personal navigation platform or the KITTI dataset showing the potential of the discussed technologies and methods. The introduced video-based dead reckoning technique can be used as a stand-alone application for relative positioning or to aid common GNSS navigation system as a complementary system designed for GNSS-denied environments. The obtained Stereo Visual Odometry system is robust enough to provide precise position estimates even in challenging situations.