Iterative Edge Enhancing Framework for Building Change Detection

Abstract

The <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Building Change Detection</i> (BCD) task serves urban planning through monitoring the landuse. However, due to the complexity of remote sensing images and high foreground-background similarity, it leads to inaccurate detection of building edge regions. Existing methods deal with this problem by fusing features of different layers. But the fusing operation cannot separate details information from overall information of buildings, resulting in inaccurate detection of building edge area. To address the above challenges, we propose an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Iterative Edge Enhancing Framework</i> (IEEF). The IEEF alleviates the building edge detection difficulty by densely implementing Detail Semantic Enhancement Module in decoding part. This module takes differential features between adjacent scales to explicitly represent the building edge information. Simultaneously, to deal with the classes imbalance problem, a Density-Guided Sampling method dedicated for change detection is proposed to increase the proportion of positive samples during training. Our proposed method achieves state-of-the-art performance on the LEVIR-CD dataset and the WHU dataset, and obtains accurate changed building edges.