UAV small target detection algorithm based on an improved YOLOv5s model

Abstract

The targets of UAV target detection are usually small targets, and the backgrounds are complex. In this work, aiming at the problem that small targets are easy to be missed or misdetected during the UAV detection, an improved YOLOv5s_MSES target detection algorithm based on YOLOv5s is proposed. First of all, to solve the problem of UAV’s difficulty in detecting small targets, the detection layer is ameliorated into the small target detection layer STD, which makes the model more easily detect the small targets. Then, the multi-scale feature fusion module is added to improve the detection accuracy of the small targets. Furthermore, by combining multi-scale module and attention module, a new connection method is proposed to retain the large scale of feature information. Finally, in contrast with some existent methods, the experimental results of VisDrone2019 UAV target detection dataset show that our proposed YOLOv5s_MSES can achieve the better detection effect, and more effectively complete the small target detection task for UAV aerial photography images.