基于D-S证据理论的遥感影像融合技术研究

Research on fusion technology of remote sensing image based on D-S theory

多源遥感影像融合是一种将来自多个传感器所拍摄的同一区域图像进行智能合成的一项技术。目前,遥感影像融合研究大多数都停留在像素级融合方面,很少考虑到异质影像融合后新像元的物理意义丢失。因此,提出基于D-S证据理论的热红外高光谱和可见遥感影像决策级融合方法,首先采用最大似然监督分类方法分别对热红外高光谱遥感影像和可见光遥感影像分类,并对分类结果进行评价,然后利用D-S证据理论对热红外高光谱影像和可见光遥感影像分类信息实现决策级融合,实验结果表明:使用D-S证据理论融合后的图像分类精度较融合之前改善效果非常明显,说明该方法在异质遥感影像融合中有很强的理论和实际意义。

The fusion of multisource remote sensing image is a technique to intelligently synthesize images taken by multiple sensors in the same region.At present the study on the present pixel-level fusion is still the main content of the remote sensing image fusion,and the new pixel is proposed without physical significance after heterogeneous image fusion,where the fused image serves as a disservice to its subsequent application.In this paper,the decision-level fusion is developed based on Dempster-Shafer(DS)theory to fuse the thermal infrared hyperspectral image and optical image.Firstly,the thermal infrared hyperspectral image and optical image are classified respectively by employing Maximum Likelihood Classification.Secondly,the precision evaluation is carried out for the classified images.Finally,the decision-level fusion is implemented based on D-S theory to fuse the classified images.The experiment result shows that the precision of the classified image fused by D-S theory is higher than the pre-fusion classified image,and the method has a strong theoretical and practical significance in the area of heterogeneous remote sensing images fusion.