Robust recognition of targets for underwater docking of autonomous underwater vehicle

Underwater docking for an autonomous underwater vehicle is important in sense that the vehicle can stop at a docking station to recharge its battery, transfer data, and can be used for launch and recovery system. To perform docking, recognizing the station through vision is important. There are few researches conducted on underwater docking using vision to recognize targets as guidance for the underwater vehicle to home towards the station. In those researches, docking is unsuccessful when one or more of the targets are not detectable. Specifically, the image processing part failed to recognize the target if the number of target taken from a captured image is not the same as the number of target in a desired image. This paper proposes a robust recognition of targets algorithm using bounding box partitioning to overcome the aforementioned problem. Result shows that the algorithm is capable to recognize the targets even if some of the targets went missing.