摘要
针对日益增多的多传感器不同分辨率的遥感数字图像数据,提出其综合利用和尺度转换的有效方法。其中,作为尺度上推(scaling-up) 技术,本研究选择IHS和小波分解变换两种数据融合方法,应用Landsat ETM+数据和IRS-1C数据进行了实例研究。发现: IHS方法操作简单,对两种待融合的图像的像元大小和像元数(bit数) 的依赖不是很高,而对原始图像光谱信息的依赖很高,因此适用于不同传感器之间的数据融合上,其缺点就是不能够更多的保留原始多光谱数据的光谱信息。小波分解变换(WD) 方法可以很好地保留多光谱传感器的所有光谱信息,但是在重采样中对两种待融合的图像的像元配准的要求很高,因此比较适合于对同一传感器的多光谱数据与全色光谱数据的融合上。作为尺度下退(scaling-down) 技术,通过探讨将札幌野外实验站点数据镶嵌到高时间分辨率遥感图像的问题,提出像元级尺度扩展的具体操作方法和对NDVI图像进行扩展时出现混合像元的识别方法。应用AVHRR, MODIS, ETM+和ASTER的NDVI图像数据进行了像元尺度的扩展研究。得出: 对不同传感器的不同分辨率的数据之间进行像元级的尺度扩展后,能够较好地识别出不同地物的分布边界,较好的分辨混合像元。
In order to acquire high resolution, data fusion technique can be used to combine multiple data from different sensors. This study practices two methods of data fusion: IHS transformation method and wavelet-based method. The result showed that IHS transformation method was a relatively simple one to be used, but it can not remain all information except for three bands of RGB. However, the wavelet-based method is relatively complicated and it can get high resolution images in all bands. As an approach to scale down the resolution of images, a so-called pixel level data scaring model was used in this study. Comparisons were made from data acquired by four multi-spectral sensors (Landsat/ETM+, Terra/ASTER, Terra/MODIS, and NOAA/AVHRR) over Kushiro Marsh in Hokkaido, Japan, on September 26, 2001. To reveal the effect of the sensors' spatial resolution, simulated data are generated from the higher spatial resolution (small size pixel) data to match the lower spatial resolution (larger size pixel) data. The result shows that the Terra/ASTER images can be effectively down-scaled to the resolution of Landsat/ETM+. However, it is rarely effective to scale down both Landsat/ETM+ and Terra/ASTER images to the resolution of MODIS and AVHRR.
出处
《地理学报》
EI
CSCD
北大核心
2004年第1期101-110,共10页
Acta Geographica Sinica
基金
日本北海道道厅科学技术振兴会的合作研究资助项目(03-01-03 103-021017)~~