摘要
卫星遥感影像在接收的过程中由于受到许多因素的干扰,使影像的质量受到一定程度的影响。而不同时相的影像也会因为成像时光照和大气条件的不同而出现差异。因此,在使用遥感影像之前往往需要进行必要的辐射校正。目前,主要用于辐射校正的两种基于影像的正规化模型分别是Markham和Irish等人提出的校正日照差异的正规化模型(ICM)和美国尤他州立大学遥感和地理信息系统实验室新近提出的日照大气综合校正模型(IACM/IACMt)。本文以三对不同时相的Landsat TM/ETM+影像为例,利用不变特征地物法求出各影像的相对噪音、RMSE和斜率,并用这些指标来对模型进行定量对比和评价。结果表明,当不同时相影像之间的季相、太阳天顶角和大气条件有较大差别时,经过正规化处理的遥感影像的噪音和RMSE比未经正规化处理的影像得以明显降低,且斜率更接近于1,镶嵌接缝淡化,效果更好。但如果遥感影像之间的季相一致、成像时的太阳天顶角接近,且影像晴空无云的话,则经过正规化处理的影像和未经该处理的影像之间的区别并不大。就正规化模型之间的对比而言,当影像无明显的大气影响时,ICM要好于IACM;但当受到明显的大气影响时,带有τ值的IACMt模型更好。
The date and seasonal variations usually result in different atmospheric and illumination effects on the satellite remote sensing images and thus can cause errors in remote sensing applications. Accordingly, pre - processing of multitemporal images using absolute or relative radiometric correction procedures between multitemporal remote sensing images has become critical for reducing radiometric differences. Among various methods, the opti- mum one is based solely on the digital image and requiring no in - situ field measurements during the satellite overflight. This is achieved through image normalization technique. Taking three pairs of multi - temporal Landsat The- matic Mapper/Enhanced Thematic Mapper plus images as examples, this paper quantitatively compared and evalu- ated two image normalization models using the criteria of relative noise, root mean square error （RMSE） and slope value derived from the pseudo - invariant feature method. One of the evaluated models is the commonly - used ILl- lumination Correction Model （ICM） proposed mination and Atmospheric Correction Model （ by Markham （1986） and Irish （2001）. The other model is the Illu- IACM/IACMt） recently developed by the Remote Sensing and GIS Laboratory of the Utah State University （2004）. The results of the comparison and evaluation reveal that, in general, when the seasons and sun azimuth angles of muhitemporal images are different or the images have obvious atmospheric effects, the relative noise and the RESE values of the image can be considerably reduced after image normalization, the slopes of scatter plots of the normalized images are more close to 1 when compared with those of unnormalized images, and the seam line between muhitemporal images mosaicked using normalized images is almost invisible. However, there is no significant difference between the normalized and unnormalized images when the seasons and sun azimuth angles are similar and the atmospheric condition is also alike. As far as the normaliza- tion models are concerned, the IACMt can achieve the best result when the image was seriously atmospherically stained. Nevertheless, the application of the IACMt may sometimes bring noise to the image when the image is free of cloud and atmospheric effect is not serious. This is probably due to the overestimation of the τ value used in the model.
出处
《地球信息科学》
CSCD
2008年第3期294-301,共8页
Geo-information Science
基金
国家自然科学基金资助项目(40371107)
福建省自然科学基金项目(2007J0132)
