250328 Enhance publishing丨Enhanced Super-Resolution-based Dual-Path Short-Term Dense Concatenate Metric Change Detection Network for Heterogeneous Remote Sensing Images

Abstract

There are problems with spatial resolution differences, spectral differences, and complex and diverse types of changes in optical heterogeneous high-resolution remote sensing images, making it more difficult to accurately and efficiently detect changes in heterogeneous high-resolution remote sensing images. A dual path short-term dense connection metric change detection network based on enhanced super-resolution for heterogeneous remote sensing images (ESR-DMSNet) is proposed to address the above issues, exploring a new path for high-precision and high-efficiency change detection of optical heterogeneous high-resolution remote sensing images.

A heterogeneous remote sensing image quality optimization network based on enhanced super-resolution (ESRNet) is proposed, which enhances edge and detail information while addressing spatial resolution differences in heterogeneous remote sensing images at the image level; A dual path short-term dense connection metric change detection network (DSMNet) was proposed to address spectral differences in heterogeneous remote sensing images at the feature level and achieve high-precision and high-efficiency change detection; Comparative analysis of four sets of homologous and heterologous remote sensing image datasets shows that the proposed method outperforms the other 12 mainstream change detection methods, with F1 scores of 79.69%, 71.01%, 95.87%, and 90.55%, respectively.

The proposed method has higher accuracy and efficiency, and the best generalization performance. When detecting large and small land features, the detection results are more consistent internally and have finer edges. The attachment is the open source code of the author's research findings.