基于区域特征的高分辨率遥感影像变化检测研究

传统像素级变化检测往往忽略邻近有意义的整片区域的空间、纹理、结构等信息,对高分辨率遥感影像具有很大的局限性。本文利用面向对象的思想,提出了一种基于区域特征的特定目标变化检测方法。该方法的技术流程包括:数据预处理;同质区域获取;区域特征选择;同名区域搜索;区域特征比较;变化检测精度评价及变化显示。利用提出的方法对伊朗2003年地震前后的巴姆古城标志性建筑进行检测,总体精度达到89.73%。

High resolution images change detection based on region feature

Traditional pixel-based change detection methods often ignore the spatial,texture and structure information of regions of interest,which is a huge challenge for high resolution images change detection.The region-based approach proposed in this paper takes advantage of the object-orient idea to detect object-specific changes.There are six steps:images pre-processing,homogenous region acquisition,region features selection,associated region searching,region features comparison,accuracy assessment and change visualization.We applied the method studied in this paper to detect symbolic building changes during earthquake of Bam,Iran and the overall accuracy reaches to 89.73%.