基于改进YOLOv5的水下目标检测算法研究

Research on Underwater Target Detection AlgorithmBased on Improved YOLOv5

针对水下场景环境复杂和图像模糊失真导致的水下目标检测精确度低的问题,提出一种基于YOLOv5的改进水下目标检测算法。将Swin Transformer集成到YOLOv5的基础骨干网络中,增强模型对水下模糊图像的处理能力,通过调整置信度损失函数,提高目标检测的准确性。经过实验验证,该改进模型在水下环境中的表现显著,平均精度(mAP)达到了77.13%,远超传统模型,提升了水下目标检测的效率。

To address the low accuracy of underwater target detection caused by complex underwater scenes,and blurred and distorted underwater images,this study proposes an improved underwater target detection algorithm based on YOLOv5.The study integrates Swin Transformer into the foundational backbone network of YOLOv5 to enhance the model s ability to handle underwater blurry images by adjusting the confidence loss function,thereby improving the accuracy of target detection.After experimental verification,the improved model exhibits remarkable performance in underwater environments,achieving an average precision(mAP)of 77.13%,far exceeding traditional models.This achievement enhances the efficiency of underwater target detection.