摘要
多时相影像对应波段中地物波谱“准”不变特征点(Pseudo-Invariant Features,PIFs)的选取是相对辐射校正的前提,采用目视解译选取主观性较强,在一定程度上影响了校正精度。利用绍兴试验区两时相TM影像,通过样本点间差异阈值,主成份分析和回归分析方法控制选取影像对应波段间的PIFs,使得校正样本点的选取具有客观性。在此基础上获取增益和偏移量并对两时相影像进行相对辐射校正,取得了较好的校正效果,从而有利于土地利用/覆盖变化监测。
The first step of the Relative Radiometric Normalization （RRN） of muhitemporal images is to select the Pseudo-Invariant Features （PIFs）. The selection of PIFs by visual interpretation is subjective and the radiometric resolution of the image is reduced. In practice, two dates of images in Shaoxing area are normalized radiometrically, which selects candidate PIFs objectively with technique including difference threshold, principal component analysis, and regression analysis procedures. Based on gains and offsets calculated, the image in 1984 is normalized radiometrically to the reference image in 1997. The technique for RRN is effective and the normalization radiometrically images are suitable for land use/cover change detection.
出处
《测绘学报》
EI
CSCD
北大核心
2006年第2期122-127,共6页
Acta Geodaetica et Cartographica Sinica
基金
浙江省国土资源遥感综合调查项目(20001500012021)
关键词
相对辐射校正
多时相影像
线性相关
主成份分析
阈值
relative radiometric correction
muhitemporal images
linear correlation
principal component analysis
threshold
