数値標高モデルを用いたランドサットTM画像の幾何補正の最適化

抄録

In the conventional scheme of geometric correction of satellite imagery, a number of ground control points (GCPs)are used to determine the geometric transform equation statistically. The selection of GCPs on the map is usually arbitrary, and the identification and measurement of GCPs on the satellite image need intensive manual labor unless they are determined automatically. In this paper, we propose a new geometric correction method, which directly determine the optimum geometric transform equation without selecting the GCPs. It is noted that satellite data over rugged terrain is highly correlated with direct solar irradiance, which can be calculated by using digital elevation model and solar position. The correlation between the simulated irradiance image and the ortho-rectifieds atellite image is used to evaluate the geometric transform equation and to optimize itsparameters. The method is applied to system corrected Landsat TM data with geo-reference parameters provided by NASDA. It is shown that optimization on two parameters (scene center displacement) is sufficient for relatively narrow area (18km by 18km) with subpixel accuracy.