High-Resolution Remote Sensing Change Detection Based on Inverse Correction and Density Peak Clustering

Zhi Li,Zhenming Peng,Siying Cao,Lunhai Luo

DOI: https://doi.org/10.1109/igarss52108.2023.10282762

2023-01-01

Abstract:Remote sensing image change detection (CD) is used to identify ground surface changes by analyzing multi-temporal images. Scholars have summarized and extended numerous useful methods for high-resolution remote sensing images. Most of the existing approaches use clustering and correction to detect change regions. However, some methods cannot detect change areas accurately due to bountiful clutters and complex information in high-resolution remote sensing images. To overcome these limitations, we propose a new detection approach using image-inverse correction and density peak clustering: Firstly, the multi-temporal images are corrected by the relative geometry and intensity. Then, the positive and inverse phase difference images are calculated.Moreover, the V channel is extracted in the HSV color space. Subsequently, applying image-inverse correction enhances the intensity of change regions and suppresses unchanged regions. Furthermore, density peak clustering is implemented to remove clutters. Finally, the processed positive-phase differential images are fused with the inverse-phase differential images to obtain the final change detection results. This approach establishes an accurate and efficient module to detect the change areas, which can accurately detect the change regions in the multi-temporal remote sensing image. The experiments show that the proposed research has stronger detection accuracy than the other six advanced change detection algorithms.