基于改进YOLOv5的遥感图像目标检测算法OA

In order to solve the problem of low target detection accuracy caused by dense array of remote sensing images,large scale difference and complex background,an improved YOLOv5 based remote sensing image target detection algorithm is proposed.First,based on the YOLOv5 network framework,a triplet attention mechanism is added to the C3 structure of the feature fusion network to improve the feature fusion capability of the model.Secondly,a large selective kernel network is added to the backbone network and feature fusion network to adjust the large spatial receptive field and better simulate the ranging environment of various objects in the remote sensing scene.Next,the crossover ratio based on minimum points is used as a new bounding box regression method to improve the speed and precision of bounding box regression.Finally,a new non-maximum suppression algorithm is used to reduce the missed detection of dense targets.The proposed algorithm was tested on DIOR,a public remote sensing data set.The results show that the proposed algorithm has an average accuracy increase of 6.6%compared with the original YOLOv5 algorithm.Compared with other YOLO detection algorithms and their improved algorithms,the proposed algorithm has the highest detection accuracy on the small sample data set used.