Exchange Data Augmentation for Change Detection

JunYi Duan, Yijing Wang, Xu Tang, Jingjing Ma, Yuqun Yang

Abstract: Change detection is developed to automatically identify semantic changes between remote sensing (RS) images captured at different points of time in a specific geographic location. Due to the limited availability of annotated data and the high complexity of the change detection problem, the performance of change detection models can not meet our expectations. To address this issue, many methods are proposed to solve this issue. However, ignoring the characteristics of the change detection task limits their performance. Therefore, we introduce a novel exchange data enhancement method (EDEM) strategy to generate image pairs to help the neural network to capture the temporal consistency in the data. We evaluate the proposed approach on two publicly available datasets and compare it with several state-of-the-art methods. The experimental results demonstrate that our proposed approach can effectively improve the performance of change detection models, achieving state-of-the-art performance on both datasets.