Satellite data offer unrivaled utility in monitoring and quantifying land use/cover change over time. Radio-metric consistency among collocated multitemporal imagery is difficult to maintain due to variations in sensor characteristics, atmospheric condition, solar angle, and sensor view angle that can obscure surface change detection. Ra-diometric corrections are often performed on multitemporal imagery to rectify radiometry of images in such a way as if they have been acquired at the same imaging conditions. In this paper, an automatic normalization method was developed based on regression of gray level categorization applied on unchanged pixels in multitemporal ETM+ imagery. Through image difference histogram, different threshold values were calculated and used for each band to select unchanged pixels, and unchanged pixels were categorized to three categories, dark, gray and bright according to their gray level values. Then corresponding coefficients obtained from gray level categorization were calculated and applied to produce the normalized image. The result showed this automatic normalization method had the capability in minimizing differences in imaging conditions.%在土地利用/覆盖变化监测中,通常要对多时相遥感影像进行辐射归化处理,使得影像间的成像差异减小,相同地物的光谱特征相似.本研究利用两时相ETM+影像,精确选取不变像元,采用分段线性回归,探讨一种自动辐射归化处理方法.研究方法充分考虑了影像间成像条件的差异,通过影像对应波段差值直方图剔除变化像元来有效选取不变像元,将不变像元划分为暗、灰和亮3类并分级确定校正系数进行辐射归化处理.结果分析表明该方法具有一定优势.
展开▼
