Vision based autonomous underwater vehicle navigation: underwater cable tracking

Abstract

In this paper, a vision based autonomous underwater vehicle (AUV) navigation system is proposed and its performance is demonstrated by an autonomous underwater cable tracking (AUCT) mission at Lake Biwa using the test bed AUV "Twin-Burger" available at the University of Tokyo. The proposed system for the AUCT consist of three levels of controllers; namely, the higher level controller, the middle level controller and the lower level controller to navigate the AUV. A single CCD camera and an acoustic sensor are used to determine the relative position of the underwater cable with respect to the AUV in 3 dimensional space. Visual data provide two dimensions and acoustic data provide the third dimension. An image filtering technique to reduce some undesirable features which are unique to the underwater images, is used based on the Laplacian of the Gaussian (LoG) operator. A real-time algorithm is developed to determine the position of the cable in the image plane by cascading an estimator governed by AUV dynamics, with the Hough transformation technique. At the Lake Biwa mission, the proposed system performed very well other than in situations where the cable was covered by waterweeds for a long distance. The proposed system can be used as a solution to the problem of underwater positioning especially in pre-determined areas.