衛載SAR影像方位參數之求解，是以檔頭資料所提供之衛星位置為初始值帶入平差模式進行求解，並加入數值地形模型進行建立物空間與像空間之關係，以求取地塊影像，本研究即以此地塊影像進行匹配。 本研究重點首次引入額外一張地塊影像作為輔助資訊來進行匹配；於本實驗中新加入之地塊影像，定義為另一目標窗地塊影像，而兩張目標窗地塊影像必需完成基準轉換。在匹配模式中，使用粗差偵測中的迭代資料搜評技術，逐次將含有雜訊過大或高頻特徵之觀測量從觀測資料組中剔除，以提高匹配精度。 本文將使用最小二乘匹配法探討三張地塊影像匹配（二張目標窗及一張搜尋窗匹配）之可行性，並和二張地塊影像匹配成果作分析比較，且嘗試於三張地塊影像匹配使用最佳不變二次無偏差估計式進行方差分量估計，對匹配影像之觀測量之權值進行較合理的給予與研究。 The orientation parameters are solved from the satellite header files information and adjustment model. Then add digital terrain model to develop relations between object space and image space. In this study, we use groundel area images to match them with one another. Object surface element is the basic element of object space matching, also called groundel. The size is pixels rate project to object space. Groundel images are derived from object space position to get image coordinates and gray level. The methodology has to develop the mathematical correlation between object space and image space. We provide another groundel area image as additional information for image matching. Next, match groundel area images base on the least square matching. To improve a precise matching, we use the iterated data snooping（IDS）－use the normalized residual to detect and delete the blunder of observation at each iteration. Finally, we use the best invariant quadratic unbiased estimation（BIQUE）to provide a method to estimate the variance components and to get reasonable weight.
