Abstract

The limited underwater observation scenarios pose great challenges to the problem of object recognition from the low-resolution underwater images. This paper proposes a framework to explicitly learn the discriminative features from relatively low resolution images, by resorting to deep learning approaches and super-resolution method. Firstly, the framework tackles the problem of limited discriminative information of low resolution images by a single-image super resolution method. Then state-of-the-art deep learning approaches are employed to learn recognition models for the special underwater fish recognition task. The proposed framework can be effectively implemented for real-time underwater object recognition on autonomous underwater vehicles. To verify the effectiveness of our method, experiments on a public underwater image dataset of fishes are carried out. The results show that our framework achieves promising results for fish recognition on underwater image datasets.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call