Multi-sensor fusion for autonomous underwater cable tracking

In this paper, a multi-sensor fusion technique is proposed for autonomous underwater cable tracking. The work presented here is an extension of the vision based tracking system proposed earlier [1,2,3,4]. Two practical problems, encountered in vision based tracking, are considered and a solution based on a sensor fusion technique is proposed. The two practical problems considered are; 1) situation when the cable is totally invisible in the image; and 2) situation when there are many similar cables appearing in the image. An experiment is conducted by setting an underwater cable with real world situations, especially with the above mentioned cases, and an Autonomous Underwater Vehicle (AUV), the "Twin-Burger 2" is used to track the cable. The experimental results demonstrate that the proposed method is quite stable in handling the practical problems related to vision based tracking systems.