Vision Optimization for Altitude Control and Object Tracking Control of an Autonomous Underwater Vehicle (AUV)

Abstract

Underwater vision is very different with atmospheric vision, in which the former is subjected to a dynamic and visually noisy environment. Absorption of light by the water and rippling waves caused by atmospheric wind are resulting uncertain refraction of light in the underwater environment, thus continuously causing disturbance towards the visual data collected. Therefore, it is always a challenging task to obtain reliable visual data for the control of autonomous underwater vehicle (AUV). In this paper, an AUV was developed and is tasked to perform altitude control and object (poles) tracking control in a swimming pool by merely using a forward-viewing vision camera and a convex mirror. Prior to design and development of control system for the AUV, this paper only focuses on utilizing and optimizing the visual data acquired. The processing process involves only gray-scaled image and without any common color restoration or image enhancement techniques. In fact, the image processing technique implemented for the object tracking control in this paper contains a self-optimizing algorithm, which results improvement on the object detection. The result shows that under similar challenging and dynamic underwater environment, the detection with optimization is 80% more successful than without the optimization.