Research on high resolution remote sensing image change detection based on pixel level and improved U-Net network
CHEN Yidong
Abstract：To timely capture changes in land cover information,this study fully leverages the advantages of high-resolution imagery in both spectral and spatiotemporal resolution by proposing an integrated approach that combines an improved U-Net network for semantic segmentation with pixel-level change detection.Capitalizing on the strengths of pixel-level detection,the method constructs a remote sensing image change detection model by calculating spectral and textural features of the images,and employs a Random Forest classification algorithm to achieve pixel-level change detection.Furthermore,based on object-oriented remote sensing image segmentation results,change extent and false change regions are accurately extracted.Experimental results demonstrate that the proposed method achieves high performance across evaluation metrics such as precision,false detection rate,and missed detection rate,confirming its reliability and effectiveness.
Keywords：change detectionhigh-resolution imagingpixel levelimproved U-Net network
Publication Date：2025-08-25
Pages：4( 48-51 )