Underwater optical image object detection based on YOLOv7 algorithm

Underwater optical imaging is a highly challenging task owing to the intricate underwater environment. This task is often plagued by issues such as image blur, color distortion, and low contrast, which pose significant obstacles to target detection tasks. Traditional target detection methods depend on manually designed features, which may not accurately characterize underwater targets and can also be impacted by factors such as target occlusion and sediment burial. This paper presents a novel baseline for underwater object detection based on the YOLOv7 algorithm, an end-to-end detection algorithm with excellent performance in terms of detection speed and accuracy. The algorithm was trained and tested on the URPC dataset, and compared with the YOLOv5 series of algorithms. The experimental results demonstrate that YOLOv7 performs better in terms of accuracy, and effectively mitigates the effects of occlusion, image blurring, and color distortion. These findings have implications for target detection tasks of underwater unmanned systems in the future.