Abstract
Using visual sensors for detecting regions of interest in underwater environments is fundamental for many robotic applications. Particularly, for an autonomous exploration task, an underwater vehicle must be guided towards features that are of interest. If the relevant features can be seen from the distance, then smooth control movements of the vehicle are feasible in order to position itself close enough with the final goal of gathering visual quality images. However, it is a challenging task for a robotic system to achieve stable tracking of the same regions since marine environments are unstructured and highly dynamic and usually have poor visibility. In this paper, a framework that robustly detects and tracks regions of interest in real time is presented. We use the chromatic channels of a perceptual uniform color space to detect relevant regions and adapt a visual attention scheme to underwater scenes. For the tracking, we associate with each relevant point superpixel descriptors which are invariant to changes in illumination and shape. The field experiment results have demonstrated that our approach is robust when tested on different visibility conditions and depths in underwater explorations.
Highlights
Visual tracking of relevant regions in scenes with poor visibility is an important problem in robotic vision research
We compare the relevant regions detected by our approach (AVA) and the Neuromorphic Visual Toolkit (NVT) [12]
A comparison between the regions considered as relevant by a group of people and by the proposed approach is presented. The purpose of this experiment is to show that our visual attention algorithm is able to detect regions that have the potential to draw the attention of a human
Summary
Visual tracking of relevant regions in scenes with poor visibility is an important problem in robotic vision research. The main challenge is to transfer to the robotic agent the ability of recognizing what regions are of interest for monitoring and to keep those regions on view for a certain period of time to be able to obtain useful visual data for its posterior analysis As these targets or regions of interest may be located far from the vehicle, they need to be detected from the distance. Since underwater environments are highly unstructured and constantly changing environments, one of the main problems that still remains open is the accurate estimation of the robot’s position and orientation This makes the detection and tracking of visual cues difficult. Poor visibility conditions underwater affect the perception of color This is due to the attenuation of light, water conditions, distance to objects, depth, and other factors [5].
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
