The single-stage target detection algorithm has attracted the attention of many researchers and industries due to its simple structure and high-efficiency model.Based on the existing YOLO algorithmregarding the difficulties of small size and tight arrangement of targets in remote sensing imagesa lightweight target detection method is proposed to improve the accuracy of small target detection in complex backgrounds.In this methodweighted fusion feature network is introduced to provide each layer of feature map with a weight coefficient that can be continuously learned in trainingso as to strengthen the feature fusion of deep and shallow layers.By introducing CIoU loss and model improvementthe convergence speed of the network is accelerated to meet the real-time requirements.Comparative experiments are carried out on the small target dataset based on DOTA remote sensing images.The experimental results show that the method has better detection accuracy and detection speed.