A Comparative Study of Underwater Marine Products Detection based on YOLOv5 and Underwater Image Enhancement
Guanxi Huang · International Core Journal of Engineering · 2021
The application of target detection algorithms in underwater images has not been very effective due to the quality problems such as image blurring and color incongruity that often exist in underwater images. In this paper, the YOLOv5 algorithm is used as a target detection network model, which is trained by using underwater marine products dataset and combined with six underwater image enhancement recovery algorithms to enhance and recover the images before they are detected, an attempt is made to improve the deep learning based target detection method at the input side. Finally, the impact of different underwater image enhancement algorithms on the YOLOv5-based target detection algorithm is compared and summarized through experiments, and the specific image enhancement methods can effectively improve the detection performance of the algorithm under different image background environments.