Underwater attached organisms intelligent detection based on an enhanced YOLO

In today's shipping, various organisms attached to the bottom of the ship do great harm to the ship, including increasing fuel loss, hindering the operation of electronic instruments and accelerating bottom corrosion. In the process of removing the organisms attached to the bottom of the ship plays a vital role in the operation. The underwater image recognition system will carry out the visual detection of underwater attachments based on the underwater robot and assist the robot in cleaning. In this paper, we propose a underwater attached organism detection method based on an Enhanced YOLOv4. We apply the Retinex theory to enhance the underwater images to overcome the influence from turbidity and illumination. We deploy our proposed method and do experiments in the underwater attachments, e.g., starfish, barnacle, shell, sea urchin, sea rainbow, coral, etc. The experiment results show that our enhanced YOLO method achieve much better performance than the original YOLOv4 method.