摘要
卫星遥感图像数据特性分析是图像使用前重要的一步工作。利用归一化、单变量统计、类内和类间方差比值以及空间分辨率识别等数据模型分析了中巴地球资源卫星02星(CBERS-2)CCD相机数据的相对反射率波谱曲线变化特性、整幅图像的方差特性、作为分类训练区的样本地物的统计量特性和对空间地物识别的敏感性,并与LANDSAT7/ETM+图像进行对比。通过这些数据模型的应用研究,试图提供一种进行遥感图像数据特性分析的方法。
Image quality assessment is a key step to understand and make good use of remote sensing data. Data models were employed to relatively assess for image quality of the Chinese Brazil Earth Resources Satellite (CBERS 2) and ETM+ in Li jiang region, Yunnan province. They include normalized model, univariate image statistics model, inner and mutual variance model, Signal to Noise Ratio model and spatial resolution model. The normalized model can be used to convert data into same physical quantity in order to make data be comparability. The other models can be used to open out the classified capability and information of data in supervised classification. With the study on data model, a method to analyze and assess remote sensing data was tried to bring forward. The result shows that (1) the spectral profile of both CBERS 2 and ETM+ is in accord with that of standard object; (2) the whole variance in waveband 1, 2, and 3 was all larger of ETM+ than of CBERS 2 and it is reverse in waveband 4. The variance of various objects were basically all larger of ETM+ than of CBERS 2, expect waveband 4 of mountain vegetation and water; as well as waveband 1 and 3 of snow; For various objects, the skewness and the kurtosis of CBERS 2 and ETM+ are complex; (3) For mountainous vegetation, road vegetation, old city, bare soil, water and snow, inner variance was all larger CBERS 2 than ETM+, and mutual variance was larger CBERS 2 than ETM+, except water; (4) CBERS 2 and ETM+ all have good spatial resolution, according to the profile of water edge.
出处
《干旱区地理》
CSCD
北大核心
2004年第4期485-491,共7页
Arid Land Geography
基金
国家自然科学基金资助项目(40271081)
国家973项目(G2000077900)资助