Contaminants detection and Cleaning of Underwater Robot Vision Based on Deep Learning

At present, underwater robots are increasingly widely used in the fields of ocean exploration and resource extraction. However, the development of related technologies is not comprehensive enough. Due to the defects of visual algorithm and results, the accuracy and safety of operation under different working conditions cannot be guaranteed. Many researchers pay attention to the development and design of underwater vision technology. In this paper, based on deep learning and binocular vision technology, the underwater tar-get object detection algorithm and the underwater ranging algorithm are established to achieve visual recognition and positioning. Visual decontamination structure are designed to prevent visual occlusion in case of emergency. The accuracy and feasibility of the related techniques are verified by experiments.