本文主要在探討以影像套合的方式處理中低起伏地區空載多譜掃描儀影像
的幾何校正的一些問題,以及提出其解決之方法。在本文的研究程序當中,
我們以正射化的航空照片為參考影像,空載多譜掃描儀影像為套合影像,漸
近式自動化產生控制點,適當的轉換方程式進行座標轉換,重新取樣產生改
正之正射影像。研究的重心有二,其一為轉換方程式數學模式的物理意義
之探討,以曲規函數具有的彈性曲面的物理特性,解決因局部仿射轉換造成
的三角形網邊界不連續的問題。及嘗試以數學上較嚴密的最小二乘內插法
進行座標轉換的內插,藉以比較轉換方程式對不同地區影像精度的影響。
其二為影像套合的自動化,以自動化產生控制點為主,趨勢面的過濾及影像
特徵點的匹配為輔,以疊代的方式進行座標轉換,以求得較高精度的正射影
像。實驗的結果顯示,在中低起伏的丘陵地形上,校正後影像之精度可
達1/3像元。

This paper primarily discusses the problems of geometric
rolling terrain AMSS images processed by imageThe
solutions for these problems are suggested. In this research,
we orthophotos as reference images and AMSS images as
registration Then, with progressive generation of control
points and mapping functions to perform coordinate
transformation, by the resampling, the corrected orthoimages
are generated. There are two major approaches in this
study. The first investigate the physical meanings of
mapping functions which discontinue problems, resulting from
local affine transformation, edge of triangular net with
the physical characteristics of plate provided by surfaces
spline function. Moreover, for influences of mapping
functions on the accuracy of corrected different areas, the
least square interpolation which has more mathematical
definition is applied to perform thef coordinate
tranformation. The second is the automated
registration. The major component of this procedure is the
generation of control points. The filtering of trend
feature matching are taken into account secondly. In order
to more accurate orthoimages, the coordinate transformation is
iteratively. Experimental results indicate that the
corrected images accuracy of 1/3 pixels.