Improved YOLOv5 Network Method for Remote Sensing Image Based Ground Objects Recognition

Abstract

High resolution remote-sensing images have the characteristics of complex background environment, clustering of objects, etc., which lead to the problem of low accuracy in recognition of large ground objects such as airports, dams, golf field, etc. Based on this problem, this paper proposes a remote sensing image object detection method based on YOLOv5 network. By improving the backbone extraction network, the network structure can be deepened to get more information about large objects, the detection effect can be improved by adding attention mechanism and adding output layer to enhance feature extraction and feature fusion. The pre-training weight is obtained by transfer learning and used as the training weight of the improved YOLOv5 to speed up the network convergence. The experiment is carried out on DIOR dataset, the results show that the improved YOLOv5 network can significantly improve the accuracy of large object recognition compared with the YOLO series network and the EfficientDet model on DIOR datadet, and the mAP of the improved YOLOv5 network is 80.5%, which is 2% higher than the original YOLOv5 network.