Remote sensing image target detection algorithm based on YOLOv5x
In order to solve the problem of low detection accuracy in remote sensing image target detection task due to the large number of small targets and not obvious target features,In this paper,an object detection algorithm based on improved YOLOv5x in remote sensing images is proposed.Firstly,the D-SPP module is designed in the backbone network to integrate the information without deepening the network structure,so that the characteristics of different re-ceptive fields can be effectively fused.Secondly,SIOU_Loss is used instead of CIOU_Loss as the boundary frame loss function to improve the accuracy of boundary frame positioning.Finally,add a new detection head to obtain a larger scale feature graph for target detection,and build the smallest detection head in the network with Transformer.Experi-mental results show that the average detection accuracy of the proposed algorithm on the RSOD data set reaches 91％,which is 5.4％higher than that of the original YOLOv5x algorithm.
NETL
NSTL
万方数据
