Real-time panoramic system for underwater cleaning robot

Abstract

The underwater cleaning robot works in an unknown environment, where knowing the real-time surroundings and cleaning situation is necessary. This paper proposes a novel approach to realize the panoramic navigation of the underwater cleaning robot. First, we select several feature points manually to calculate the images' relative location. Second, we solve the transformation matrix from the original view to the front view. Third, we build a matrix library marked by the robot's attitude. Fourth, the panoramic navigation system makes the real-time image projection and stitching based on the attitude of the robot. We use the fade-out fusion to eliminate the image stitching gap, which is time-efficient. This panoramic system uses the pre-set parameters to make the real-time image process, instead of doing the feature detection or matching, which could increase the efficiency significantly.