Object detection and tracking for autonomous underwater robots using weighted template matching

Underwater environment has a noisy medium and limited light source, so underwater vision has disadvantages of the limited detection range and the poor visibility. However it is still attractive in close range detections, especially for navigation. Thus, in this paper, vision-based object detection (template matching) and tracking (mean shift tracking) techniques for underwater robots using artificial objects have been studied. Also, we propose a novel weighted correlation coefficient using the feature-based and color-based approaches to enhance the performance of template matching in various illumination conditions. The average color information is incorporated into template matching using original and texturized images to robustly calculate correlation coefficients. And the objects are recognized using multiple template-based selection approach. Finally, the experiments in a test pool have been conducted to demonstrate the performance of the proposed techniques using an underwater robot platform yShark made by KORDI.