基于多尺度融合的对象级变化检测新方法

讨论了遥感图像变化检测中的多尺度融合问题. 首先分析了高分辨率遥感图像的城市变化检测的难点和传统的变化检测方法的局限性; 针对这些难点和局限性, 提出了基于多尺度融合的对象级的变化检测框架. 该框架利用对象级的变化检测方法, 提高了变化类和非变化类的可分性. 为了提高变化检测的精度, 该框架引入了两种不同的多尺度融合策略. 在 QuickBird 高分辨率遥感图像上的实验结果验证了该方法的有效性和鲁棒性.

Object-level Change Detection Based on Multiscale Fusion

Urban change detection based on multiscale fusion is discussed in this paper.The limitations of traditional change detection methods and the difficulties of urban change detection of very high resolution images are first analyzed. To address the above limitations,a novel object-level framework based on multiseale fusion is presented for urban change detection.Taking advantages of the object-level change detection,the framework can improve the separability of changed class and unchanged class.To further improve the change detection accuracy,two different multiscale fusion strategies are applied in the framework.Experimental results obtained on QuickBird images confirm the effectiveness of the proposed approach.