An Improved U2-Net Method for Road Extraction in High-resolution Remote Sensing Images
Currently,there is a common problem in deep learning-based road extraction methods,which is the tendency to ignore the detailed features of images.To address this issue,this paper proposes a novel approach for extracting road information from high-resolution remote sensing images by enhancing the U2-N et algorithm.This method incorporates the modules of convolutional attention mechanism and self-attention mechanism into the original U2-Net model,which not only increases the global semantic information of images but also preserves sufficient spatial features,achieving effective fusion of different types of features.Experimental results on two road datasets,which are DeepGlobe and CHN6-CUG,demonstrate that the proposed method has stronger feature extraction and anti-interference capabilities,and overall performance is better than other similar research achievements,enabling more effective extraction of roads from high-resolution remote sensing images.
deep learningroad extractionsemantic segmentationU2-Netdual attention mechanism
万方数据
