Remote Sensing Image Target Detection Method Based on Improved YOLOX
In view of the complex background of remote sensing images,the large size differences between different categories of targets in the images,and the low accuracy of existing models in detecting small-sized targets,YOLOX was improved to implement a remote sensing im-age target detection method.A position attention module is introduced into the backbone network to allow the model to focus on learning posi-tive sample features;a two-layer weighted feature pyramid is used to replace the existing feature fusion network,and a context decoupling de-tection head is used on the detection end to fully perform the target frame regression and target classification tasks break down.Experimental results show that the average accuracy of this model on the test set reaches 90.14％,which is 9.97％higher than the YOLOX model and 4.34％higher than the YOLOv8 model.The detection speed of a single image is only 0.019 s,with real-time detection capabilities.
万方数据
维普
