摘要 地形物元是物空間匹配的基本單元，亦稱為地元，其大小約是將像元依比例投射至地表的尺寸。地元影像產生方法是由物空間位置出發推求影像坐標並擷取影像灰值，此方法需建立物空間與像空間之數學相關，即影像方位參數。 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混合平差模式進行求解，並加入數值地形模型進行建立物空間與像空間之關係，以求取一區域之地元影像，稱為地塊影像，本研究並以此地塊影像進行匹配。 在匹配模式中，以最小二乘匹配模式為基準，附加輻射參數進行地元區塊影像匹配。並使用統計上之判別檢定以尋找影像最佳匹配參數，進行參數優化之工作，以及粗差偵測中的迭代資料搜評技術，利用標準化殘差，逐次將含有雜訊過大之觀測量從觀測資料組中剔除，以達到偵測匹配影像對中含有顯著大錯之像元對並予以剔除進而提高匹配精度。並嘗試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。 Parameter Optimization in Satellite SAR Groundel Area Image Matching Abstract Object surface element is the basic element of object space mathing or call groundel. The size is pixels rate project to object space. Groundel images are derived from object space position get image coordinates and gray level. The methodology has to build the mathematical correlation between object space and image space. The orientation parameter are obtained from the satellite header files information as initial and fixed by the mixed adjustment model. Then adding digital terrain model to build relation of object space and image space correlate. The research get groundel area image to match. Based on least squares matching and added radiant parameters to groundel area image matching. To make the parameters optimal, we use the discrimination test in statistics, and to improve the precision of matching, we use the iterated data snooping which using the normalized residual to detect and deleted the blunder of observation at each iterated. Finally, we use best invariant quadratic unbiased estimation（BIQUE）to estimate the variance components and try to get reasonable weight.
