Visual summaries for low-bandwidth semantic mapping with autonomous underwater vehicles

Abstract

A fundamental problem in autonomous underwater robotics is the high latency between the capture of image data and the time at which operators are able to gain a visual understanding of the survey environment. Typical missions can generate imagery at rates orders of magnitude greater than highly compressed images can be transmitted acoustically, delaying that understanding until after the robot has been recovered and the data analyzed. We present modifications to state-of-the-art online visual summary techniques that enable an autonomous robot to select representative images to be compressed and transmitted acoustically to the surface ship. These transmitted images then serve as the basis for a semantic map which, combined with scalar navigation data and classification masks, can provide an operator with a visual understanding of the survey environment while a mission is still underway.