改进的 YOLOv4在飞机蒙皮损伤检测中的应用

针对飞机蒙皮检测中存在的小目标检测欠佳、漏检等问题,提出了 1种基于增强特征融合和 ATSS的 YO. LOv4飞机蒙皮图像目标检测算法。首先,增加用于目标预测的大尺度浅层特征层,以提高模型对小目标的检测效果；其次,增加特征融合网络层数,通过浅层与深层特征层的深度融合,丰富多尺度特征图中的特征信息；然后,通过 K-means++聚类算法对数据集的真实框聚类,获得更具代表性的先验框尺寸,以提高预测框对目标的定位准确度；最后,引入 ATSS对 YOLOv4的样本选择策略进行优化,通过自适应获取最优的 IoU阈值,实现正负样本自动划分,提升模型的检测性能。实验表明,在增加少量计算成本的情况下,算法的检测性能得到有效提升,mAP提升 7.7%,检 测的准确率达到 80%以上。

Application of Improved YOLOv4 in Aircraft Skin Damage Detection

Aiming at the problems such as poor small target detection and missing detection in aircraft skin detection, a YO.LOv4 aircraft skin image target detection algorithm based on enhanced feature fusion and ATSS is proposed. Firstly, a large-scale shallow feature layer is added for target prediction to improve the detection effect of the model for small targets. Sec.ondly, the number of feature fusion network layers is increased to enrich the feature information in multi-scale feature mapsby deep fusion of shallow and deep feature layers. Then, the K-means ++ clustering algorithm is used to cluster the real box.es of the data set to obtain a more representative prior box size so as to improve the positioning accuracy of the predictionbox for the target. Finally, ATSS is introduced to optimize the sample selection strategy of YOLOv4, and the optimal IoUthreshold is obtained by the adaptive method to realize automatic division of positive and negative samples and improve thedetection performance of the model. Experimental results show that the detection performance of the algorithm is improvedeffectively with a small increase in computational cost, the mAP is improved by 7.7%, and the detection accuracy is morethan 80%.