结合纹理信息的高分辨率遥感图像变化检测方法  

Change Detection Method for High Resolution Remote Sensing Image in Association with Textural and Spectral Information

作　　者：王东广[1] 肖鹏峰[1] 宋晓群[2] 王铁成[3] 陈刚[1] 

机构地区：[1]南京大学地理信息科学系,南京210093 [2]江苏省国土资源信息中心,南京210029 [3]江苏省测绘工程院,南京210013

出　　处：《国土资源遥感》2012年第4期76-81,共6页Remote Sensing for Land & Resources

基　　金：国土资源部公益性行业科研专项(编号:201011015-1)资助

摘　　要：高分辨率遥感图像的纹理信息同光谱信息一样能有效地用于检测变化信息,而一些基于中低分辨率遥感图像的变化检测方法多以光谱信息为研究对象,忽略了图像中的纹理信息。针对这一问题,尝试将高分辨率图像的光谱信息与纹理信息一起用于"差值主成分变化检测"方法中,一方面借助高分辨率图像间纹理信息的差异获取变化区域内部的细节信息,以弥补高分辨率图像间光谱区分度相对不足的缺点;另一方面借助纹理信息在变化区域内部的连结作用,对变化检测结果进行狭窄缺口连结、内部孔洞填充等后续处理,从而使检测结果更加完整。实验结果表明,该方法对光谱反射信息相近、但纹理信息有较大差异的变化区域具有良好的检测效果。High resolution remote sensing image can provide a lot of spectral and textural information,and both of the two kinds of information can help effectively detect the changed information.However,the traditional methods of change-detection based on medium or low spatial resolution remote sensing images only use the spectral information to extract the changed information,with the ignorance of the textural information.In this paper,both the spectral and textural features are integrated in one change-detection method to extract the changed information from high resolution remote sensing image,and the method is called difference principal component transformation.The advantages of the proposed method can be concluded in two aspects.One is that it will be easy to get the internal changed details in the large changed areas according to textural information,which can compensate for the deficiency of spectral information in high resolution images.The other is that some post-processing procedures such as connecting narrow gaps and filling holes can make the change-detection result more complete.The experimental results show that some changes that are spectrally similar but texturally different can be effectively detected after adding textural information in this change-detection method.

关 键 词：变化检测 高分辨率遥感图像 纹理信息 差值主成分变换 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]