历史解译知识引导下组合遥感图谱特征的变化检测方法

发布时间：2018-12-12 18:18

【摘要】：高空间分辨率遥感影像为地表变化监测提供了大量细节信息,这使得基于高分辨率影像的变化检测技术成为当前遥感领域的研究热点之一。本文提出了一种历史解译知识引导下组合遥感图谱特征的变化检测方法。首先,通过分割前后时相的组合影像构建空间位置一致的对象,并在提取对象光谱和纹理特征后,引入前期土地覆盖专题图指导2类图谱特征相似度的DS证据融合;然后,利用其历史存档图斑所属区域的优势地类标签指示不同特征相似度的证据差异融合;最后,基于GMM(Gaussian Mixture Mode)模型的二值化方法提取最终的变化区域。实验结果表明,该方法能充分利用历史解译知识指导不同时相高分辨率影像对象特征相似度组合,一定程度上提高了变化检测正确率。
[Abstract]:High spatial resolution remote sensing images provide a great deal of detail information for surface change monitoring, which makes the change detection technology based on high resolution images become one of the research hotspots in the field of remote sensing. In this paper, a change detection method based on historical interpretation knowledge combined with remote sensing map is proposed. Firstly, the objects with the same spatial position are constructed by the combined images before and after segmentation. After extracting the spectral and texture features of the objects, the DS evidence fusion of the similarity between the two types of atlas is guided by the land cover thematic map. Then, the dominant class labels of the historical archived region are used to indicate the evidence fusion of the similarity of different features. Finally, the binarization method based on GMM (Gaussian Mixture Mode) model is used to extract the final change region. Experimental results show that this method can fully use historical interpretation knowledge to guide the combination of feature similarity of different phase high resolution images and improve the accuracy of change detection to some extent.
【作者单位】：长安大学理学院数学与信息科学系;浙江省海洋大数据挖掘与应用重点实验室;
【基金】：浙江省海洋大数据挖掘与应用重点实验室开放课题(OBDMA201508) 国家高分辨率对地观测系统重大专项(03-Y30B06-9001-13/15-01) 中国科学院重点部署项目(KZZD-EW-07-01) 长安大学中央高校基本科研业务费专项资金基础研究项目(310812161010) 国家自然科学基金项目(11261044)
【分类号】：P237

【相似文献】