飞机蒙皮损伤智能检测方法

Intelligent Detection Method of Aircraft Skin Damage

本文基于YOLO v5提出了一种飞机蒙皮损伤检测网络,采用styleGAN来生成合成图像,用于训练所提出的飞机蒙皮缺陷检测网络,通过提出的基于StyleGAN的YOLO v5模型,提高了检测小目标、捕获低灵敏度空间信息和进行全局优化的能力。通过现场拍摄和合成的蒙皮损伤照片,对本文开发的模型进行了训练、验证和测试。本文提出的检测模型的准确率和召回率分别达到92.2%和92.3%,分别比原始YOLO v5高10.7%和12.5%。基于StyleGAN的YOLO v5具有高精度和高鲁棒性,该模型可以显著提高飞机蒙皮损伤检测效率,降低误判率。

This paper proposes an aircraft skin damage detection network based on YOLO v5.It utilizes StyleGAN to generate synthetic images for training the proposed aircraft skin defect detection network.Through the proposed StyleGAN-based YOLO v5 model,the capabilities of detecting small targets,capturing low-sensitivity spatial information,and performing global optimization are enhanced.The model developed in this paper has been trained,validated,and tested using field-captured and synthesized skin damage photos.The proposed detection model achieves an accuracy rate of 92.2%and a recall rate of 92.3%,which are 10.7%and 12.5%higher than the original YOLO v5,respectively.The StyleGAN-based YOLO v5 exhibits high precision and robustness,and this model can significantly improve the efficiency of aircraft skin damage detection and reduce the misjudgment rate.