HJ数据的LBV变换及其在面向对象分类中的应用

为了探讨环境卫星影像在分类中的应用潜力,通过对其地物光谱进行分析,计算推导出了适用于环境卫星数据的LBV变换公式,并且将变换后得到的LBV图像应用到面向对象分类中。实验结果表明:推导的针对HJ\|1B影像的LBV变换公式具有普适性,并且经过LBV变换后的影像有效地弥补了环境卫星数据光谱分辨率不高的缺点,在分割参数相同的情况下,分割效果明显好于原始影像分割结果。利用变换后的LBV图像进行面向对象分类,可以很好地提取出水体、植被、城镇和建筑用地4大类,总体分类精度达到93%,Kappa系数为0.8894,表明经LBV变换后的HJ影像在面向对象分类中具有很大的应用潜力。

LBV Transformation for HJ-1B Data and Application in Object-oriented Classification

HJ-1A/B satellite is a new one which is studied and developed by China independently.However the study on the application of HJ-1A/B satellite images is not so much.Therefore in order to improve the capabilities of HJ\|1A/B images applied in classification,in this paper,LBV transformation and object\|oriented classification methods were used to HJ-1B images.Although there have been many LBV transformation methods for different remote sensing data,but none of them can be directly used to HJ-1B multi-spectral images.So new LBV transformation equations for HJ-1B multi\|spectral images were specially proposed based on the study of the spectral characteristics of nine typical ground features.And then these equations were used to three HJ-1B images to test and verify their feasibility and generality.At last,the LBV images on 12th November 2010 of Shenzhen city were classified by object-oriented classification method.The result showed that the transformed LBV images were more vivid,which made up the lack of spectral resolution of HJ images.What’s more,the LBV images were better at segmentation than original HJ data,and the classification accuracy could be 93%,which shows that LBV transformation has a good potential in interpreting and classifying for HJ-1B images.