A model-based monocular vision system for station keeping of an underwater vehicle

A model-based monocular vision system for station keeping of an underwater vehicle is presented. Through such system accurate and reliable 3D pose information (roll, pitch, yaw and XYZ translations) of the vehicle relative to the model-known observed target is estimated in real time. The cooperative and observed target is placed in the working space when station keeping is demanded. The performance of the system is demonstrated in a water tank by using the underwater vehicle, which 4 degrees-of-freedom are controllable. With the pose information, the vehicle can automatically move to a desired station and maintain a fixed position and orientation with the presence of disturbance. The system is suitable for the underwater vehicles equipped with manipulators or in the structural environment, and would have a good prospect in the future applications