Multi-level Change Detection in Urban Areas Using ZY-3 Multi-temporal Stereo Imagery

Due to the rapid process of urbanization, there is an urgent need for accurate and detailed change detection in urban areas. Ziyuan-3 (ZY-3), as China's first civilian high spatial resolution satellite, carrying the three-line array camera, enables the acquisition of in-track stereo images. The characteristics of multi-angle imaging and high-spatial resolution are fully exploited to explore subtle urban change information. In this study, a multi-level (pixel-grid-block) change detection framework using stereo images was conducted to detect information related to land cover transitions, changed hotspots, and multi-temporal landscape. The experimental results reveal some interesting and informative findings: (1) our proposed multi-level method is effective in detecting change details, with kappa coefficient reaching 0.81 at the pixel level and correctness being 95% at the grid level; (2) the block-level landscape analysis indicate that there was greater fragmentation and spatial heterogeneity of building landscape during the urbanization process; (3) the mean nearest neighbor distance between building patches decreases by about 1.0 m during 2012 and 2013.