Regional Proposal Based Underwater Object Tracking

In recent years, object tracking has developed rap-idly, in which Siamese network-based trackers have occupied an important position in this field. Most algorithms are designed for images taken in the normal air environments; fewer methods are for underwater situations. Water always can bring distraction and reflection of the light, which can blur images and distort colors. A regional proposal based Siamese network combined with under-water image enhancement is proposed for the underwater tracking problem. The proposed tracking method is evaluated on some natural underwater images, and achieve better results than general tracking methods.