由於載具飛行高度與影像跨幅的關係，在進行衛載SAR影像的後方交會以反演飛行軌跡時，主要問題為交會角過小所造成的誤差，且迭代過程中初始值之給定亦為交會成果優劣之關鍵之一。本研究以ERS-1影像檔頭資料所提供之衛星預估位置以計算初始值，進行平差求解。由於衛星位置構建於地心直角坐標系（WGS84）下，為符合演算模式之假設，且考慮演算之有效位數及符合台灣地區之坐標架構，研究中定義一工作坐標系，並將所有坐標皆轉換至工作坐標系中操作。目的在於找出載具軌跡多項式，以得到較精確的方位參數。 匹配模式中，需由物空間位置出發推求影像坐標以擷取灰值進行匹配計算。然而在雷達影像中，由物空間三維坐標出發計算影像坐標係為二元二次求解問題，為簡化計算，本研究提出地形物元求解方法，以得到各地形物元所相對應的影像坐標殘差，以解決二元二次方程式求解之問題，求得影像坐標，以擷取影像灰值，進行影像匹配。本研究並將最小二乘匹配模式作更進一步的研究與討論，成功提升衛載SAR影像的匹配精度。 Because of the small ratio of ERS-1 satellite orbit height to image swath, the main problem of solving fly-path by resection is the error caused by the small intersection angle. Besides, the initial approximation of unknowns for iteration are the other key point under consideration. In this paper, we make use of the metadata of ERS-1 images to obtain the information of satellite position and then to computer better initial approximation. The coordinates of satellite position are originally located in ECEF (WGS84) system. In order to improve the computational accuracy and to be consistent with present coordinate system of Taiwan, we defined a coordinate system in our study. All coordinates in other systems are transformed to this system. In the matching model, the problem of coordinate transformation from object space to image space can be formulated by second-order polynomial with two parameters. In order to solve the problem and consider the statistic character of the orientation parameters, we present a “terrain object element” concept. It works as control points with little weight. By introducing the object elements into resection adjustment, we can get residuals and corresponding image coordinates that will be used in match algorithm. We improved Least Square Matching model, and got the better result of SAR matching experiment.
