Detection of Markers Using Deep Learning for Docking of Autonomous Underwater Vehicle

Autonomous Underwater Vehicle (AUV) has limited energy capacity due to it being an embedded system. To overcome this limitation, the AUV can home into a docking station to recharge its battery. Several research has been conducted on the docking of AUV using vision. In some literatures, docking would fail if the target placed at the docking station is missing or disoriented from the camera view. This study proposes a deep learning system to detect the target markers to solve the disoriented view issues. The proposed system comprises of two phases which are training and testing. In training phase, there are region proposal, labeling data, developing convolutional neural network architecture, and network training. In testing phase, the trained network will be fed by various input data so as to measure the performance of the network. Result in this study shows that the system is able to locate and classify the target markers even though the view of the object of interest is disoriented. Future work may include the implementation of the developed system on real docking operation.