Abstract
Abstract. This paper provides details of both hardware and software conception and realization of a hand-held stereo embedded system for underwater imaging. The designed system can run most image processing techniques smoothly in real-time. The developed functions provide direct visual feedback on the quality of the taken images which helps taking appropriate actions accordingly in terms of movement speed and lighting conditions. The proposed functionalities can be easily customized or upgraded whereas new functions can be easily added thanks to the available supported libraries. Furthermore, by connecting the designed system to a more powerful computer, a real-time visual odometry can run on the captured images to have live navigation and site coverage map. We use a visual odometry method adapted to low computational resources systems and long autonomy. The system is tested in a real context and showed its robustness and promising further perspectives.
Highlights
Mobile systems nowadays undergo a growing need for self localization to accurately determine its absolute/relative position over time
We propose to guide the survey based on a visual odometry approach that runs on a distributed embedded system in real-time
Traditional visual odometry methods based on local bundle adjustment (BA) suffers from rotation and translation drifts that grows with time (Mouragnon et al, 2009)
Summary
Mobile systems nowadays undergo a growing need for self localization to accurately determine its absolute/relative position over time. Whereas in traditional stereo matching the search for correspondence is done along the epipolar line within certain fixed range, in our method we proceed first by computing a priori rough depth belief based on image lightness and. Our first contribution is that we benefit from the rough depth estimation to limit points correspondence search zone to reduce processing time From another side, traditional visual odometry methods based on local BA suffers from rotation and translation drifts that grows with time (Mouragnon et al, 2009). According to our knowledge there is no method that proposes an adaptive search range following a rough depth estimation from lightness in underwater imaging
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have