弱交會幾何衛星影像之有理函數模型區域平差

RFM-Based Block Adjustment for Satellite Images with　Weakly Convergent Geometry

巫婉瑜(Wan-yu Wu)

指導教授 ： 陳良健

國立中央大學/工學院/土木工程學系/碩士(2008年)

摘要

在衛星遙測的應用中，目標區域通常涵蓋一張以上的衛星影像。因此，將正射化之影像進行鑲嵌為提供大範圍的影像資料並與其他相關資料整合之重要工作。影像正射化之首要工作為方位的重建，基於良好的交會幾何，使用區域平差可增加影像間的幾何性並達到高定位精度。然而，在許多衛星影像應用中，為求較大的影像覆蓋面積，其交會角及影像重疊區域通常很小，傳統上以三維定位為考量之區域平差模式將不適用。
有理函數模型配合以衛星方位參數求得之有理函數轉換係數的改正成果已被證實近似於嚴密幾何模型。由於具有數學模型簡單且標準化的特性，使得有理函數模型廣泛地應用在高解析衛星影像幾何校正。因此，本研究將針對弱交會幾何之衛星影像，提出一個新的有理函數模型區域平差程序，並加入數值高程模型作為航帶連結點之高程控制，藉以降低航帶連結間之相對偏移。
研究中包含三個主要工作：(1) 有理函數模型區域平差，(2) 加入數值高程模型作為高程控制，(3) 以最小二乘過濾減少地面坐標之局部性系統誤差。測試資料有兩組，分別為FORMOSAT-2 和SPOT-5影像。實驗內容將分別測試區域平差、高程控制及最小二乘過濾之成效，以及有理函數模型區域平差和直接地理定位、光束法平差之成果比較。精度評估分為兩大部份，分別為絕對精度和影像間之相對偏移量。實驗結果顯示，加入航帶連結點進行有理函數模型區域平差，能有效降低影像間之相對偏移。

關鍵字

有理函數模型 ； 區域平差 ； 高程控制 ； 最小二乘過濾

並列摘要

In remote sensing applications, the area of interest often covers two or more satellite images, thus mosaicking for orthorectified images is important. Orientation modeling is the prerequisite in orthorectification. Block adjustment could enhance the geometry consistency between the images to reach high accuracy positioning for orientation determination. Based on the favorable convergent geometry, traditional bundle adjustment is done in such a way that three dimensional positioning for tie points are employed. In many satellite image applications, it is found that either the convergent angle or the overlapping area would not be large enough. Thus the traditional block adjustment should be modified to adopt for the general situation.
RFM (Rational Function Model) has been demonstrated to achieve almost identical precision with respect to the rigorous sensor model when the RPCs (Rational Polynomial Coefficients) are derived from the satellite on-board parameters. Taking the advantages of simplicity and standardization, RFM had been widely used in the remote sensing community. Thus this investigation aims at the development of a block adjustment procedure to account for the weakly convergent geometry based on the RFM. A DEM (Digital elevation model) is included in the adjustment as an elevation control.
The major works of the proposed scheme include: (1) RFM-Based block adjustment, (2) elevation control with a DEM, and (3) least squares collocation. Satellite images including FORMOSAT-2 and SPOT-5 are used in the validation. The experiments test the effect on block adjustment, elevation control and least squares collocation. The results of RFM-Based block adjustment are compared with the direct georeferencing and bundle adjustment. The validation includes two parts: absolute accuracy and relative discrepancy. Experimental results indicate that the proposed scheme improves the geometric consistency between the image strips.

並列關鍵字

least squares collocation ； elevation control ； block adjustment ； RFM

參考文獻

張智安，2002，“EROS A衛星影像幾何改正之研究”，碩士論文，國立中央大學土木工程研究所。

劉建良，2004，“多航帶推掃式衛星方位平差及影像正射化”，碩士論文，國立中央大學土木工程研究所。

Aguilar, M.A., Aguilar, F.J., Agüera, F., and Sánchez, J.A., 2007. Geometric Accuracy Assessment of QuickBird Basic Imagery Using Different Operational Approaches, Photogrammetric Engineering & Remote Sensing, 73(12):1321-1332.

Chen, L.C., and Chang, L. Y., 1998. Three Dimensional Positioning Using SPOT Stereostrips with Sparse Control, Journal of Surveying Engineering, ASCE, 124(2):63-72.

Chen, L.C., Teo, T.A., Liu, C.L., 2006. The Geometrical Comparisons of RSM and RFM for FORMOSAT-2 Satellite Images, Photogrammetric Engineering & Remote Sensing, 69(1):59-68.