Limitations of Vision Guided Underwater Navigation

Underwater vehicles that are equipped with an on-board camera can utilize the captured frames to perceive the 3D structure of the ocean floor and thus perform safe underwater self-navigation. The focal length, frame rate and exposure time, which determine the frame capturing process, in association with the speed of the vehicle determine the capabilities of the vision-based navigation system. In this work, we quantify the effect of the above imaging parameters on the performance of the vision system, and the feasibility of vision-based navigation is assessed for various parameter values. Results demonstrate that a 1200 X 1600 camera with a 5mm lens and 15f/s frame rate allows the dense reconstruction of the ocean floor, thus enabling autonomous underwater navigation, provided there is enough processing power on the vehicle to perform the necessary image processing in real time.