Towards Underwater Object Recognition Based on Supervised Learning
Zhengyu Chen, Tongtong Zhao, Na Cheng, Xudong Sun, Xianping Fu · 2018
Underwater robots play a significant role in exploring the underwater world. In recent years, underwater robots still can't recognize the underwater objects accurately. In order to find a solution to the problem of underwater robot recognition, we put forward a framework. There are three parts in our framework. First, a color correction algorithm is used to compensate color casts and produce natural color corrected images. Second, we employ Super-Resolution Generative Adversarial Network to enhance the underwater images. There are two parts in Super-Resolution Generative Adversarial Network. One is the modified generate network$G$, and the other is the discriminator network$D$. We modify the generate network$G$on basis of ResNet. Third, we employ object recognition algorithm to process the enhanced images for detecting and recognizing the underwater object. The experimental results show that the proposed framework can achieve good results in underwater object recognition.