- 學位論文
以嚴密幾何模式進行ALOS PRISM及PALSAR影像之方位重建
Orientation Modeling of ALOS PRISM and PALSAR Imagery using Rigorous Sensor Model
摘要
ALOS(Advanced Land Observing Satellite)為日本國家太空中心所研發之科學衛星，主要進行陸域觀測，提供地表三維重建、製圖、災害監測以及資源調查等應用。ALOS衛星同時搭載光學感測器以及雷達感測器，光學感測器為PRISM (Panchromatic Remote-sensing Instrument for Stereo Mapping)與AVNIR-2 (Advanced Visible and Near Infrared Radiometer type 2)，雷達感測器為PALSAR (Phased Array type L-band Synthetic Aperture Radar)。 本研究目的是建立光學感測器PRISM及雷達感測器PALSAR之嚴密幾何模式。PRISM幾何處理中，比較內方位自率法、外方位參數校正、及反投影像空間校正模式於像空間之精度，並以前方交會分析其三維定位精度；PALSAR幾何處理中，分別建立外方位參數校正模式及反投影像空間校正模式。 在PRISM的幾何校正中，內方位自率法使用匹配的大量控制點來求解各CCD單元的補償量，將影像進行初步校正後再進行外方位參數校正；PRISM影像之外方位參數校正，使用單張影像獨立平差以及三重疊影像整體平差，為了避免參數相關性，外方位參數分別以位置參數以及姿態角參數來進行補償，補償量以時間多項式函數描述，分析以0階、1階及2階多項式函數進行解算的成果；PALSAR影像方面，外方位參數校正僅以位置參數進行補償，以0階、1階及2階時間多項式函數進行校正後，再進行偏斜角的校正；反投影像空間法使用嚴密幾何模式，將控制點地面坐標反投影至像空間上，得到其像坐標，再以四參數、六參數與八參數的方法將投影後的像坐標與原始像坐標進行校正。此外，以前方交會來評估PRISM影像校正後之三維定位精度。 研究資料使用PRISM兩組三重疊影像以及PALSAR兩張影像，拍攝範圍為台灣北部；另外以2.5公尺SPOT-5正射影像以及40公尺數值地形模型作為控制點與檢核點之平面與高程來源。研究顯示PRISM單張影像獨立平差時，使用外方位參數配合內方位自率法可得到較好的成果，Sample與Line方向之均方根誤差(Root Mean Square Error, RMSE)可達到1.8個像元與2.7個像元；PRISM三重疊影像以外方位參數配合內方位自率法進行整體平差，控制點大於10個時精度漸趨穩定，即整體平差對控制點數量的需求較少，且Sample與Line方向的RMSE可達1.9個像元與3.4個像元。獨立平差後之PRISM影像進行前方交會，平面與高程方向之RMSE可達到6.7公尺與6.5公尺，整體平差後之PRISM影像進行前方交會，平面與高程方向之RMSE可達到7.7公尺與4.6公尺。PALSAR影像以反投影像空間模式進行校正時可得到較好的成果，Sample與Line方向之RMSE可達到2.4個像元與3.9個像元。
並列摘要
ALOS (Advanced Land Observing Satellite) is developed by Japan Aerospace Exploration Agency (JAXA). The missions of this satellite are land observation and surface reconstruction. ALOS satellite equips with optical and radar sensors, i.e., PRISM (Panchromatic Remote-sensing Instrument for the Stereo the Mapping), AVNIR-2 (Advanced the Visible and Near the Infrared Radiometer type 2), and PALSAR (Phased Array type the L-band Synthetic Aperture Radar). The purpose of this study is to establish the rigorous sensor model of the PRISM optical sensor and PALSAR radar sensor. For PRISM triple images, this research compares different mathematic models such as interior orientation parameters (IOPs) correction model, exterior orientation parameters (EOPs) compensation model, and image space back-projection compensation model. Besides, space intersection is also performed to analyze the accuracy of three-dimensional coordinates in object space. For PALSAR radar image, the mathematic models include compensation of exterior orientation parameters and image space back-projection compensation model. IOP correction model uses a large number of matched control points to determine the compensation parameters for each CCD unit. EOPs compensation model for PRISM images includes orbital and attitude parameters compensation. The zero-, first- and second-order polynomial functions are used to compensate the EOP errors. PALSAR image only correct the orbital and squint angle as scan direction of PALSAR is fixed. The image space back-projection compensation method for both PRISM and PALSAR image includes four parameters, six parameters and eight parameters transformation model. The test data include two set of PRISM triplet images and two PALSAR images. The test area is located at north Taiwan. The reference data for control and check points measurement are 2.5m SPOT-5 orthoimage and 40m digital elevation model (DEM). In the independent adjustment of PRISM image, the integration of IOP and EOP correction reached the accuracies of 1.8 pixels and 2.7 pixels in sample and line directions, respectively. In the block adjustment of triplet PRISM images, the accuracies are 1.9 pixels and 3.4 pixels in sample and line directions when 10 GCPs are employed. The space intersection for independent adjustment model shows the accuracies of 6.7m and 6.5m in horizontal and vertical direction. On the other hand, the block adjustment reaches 7.7m and 4.6m accuracies in horizontal and vertical direction. For PALSAR image, the accuracy of image-space back-projection method is better than EOP correction model. The accuracies are 2.4 pixels and 3.9 pixels in sample and line direction.
參考文獻
延伸閱讀
- Chang, C. W. (2010). 多重視覺感知階段爲基礎之高動態範圍影像重構方法. 嘉南學報（科技類）, (36), 169-179. https://doi.org/10.29540/CNABST.201012.0022
- 張智安、陳良健（2005）。「科技短文」嚴密幾何模式與有理函數模式於QuickBird 衛星影像幾何改正之比較。航測及遙測學刊，10(2)，183-190。https://doi.org/10.6574/JPRS.2005.10(2).5
- Jagdale,Sumit (2017). 基於RANSAC-P3P平面方法,進行相機外部參數校正與利用LiDAR-相機系統達到3D場景重建 [master's thesis, National Chiao Tung University]. Airiti Library. https://www.airitilibrary.com/Article/Detail?DocID=U0030-2212201712334234
- Saba, M., Carlier, E., & Gerges, N. (2013). Use of Analog Models to Simulate Flow Recession of Karstic Springs. Frontiers in Environmental Engineering, 2(3), 38-44. https://www.airitilibrary.com/Article/Detail?DocID=P20150611020-201309-201509140008-201509140008-38-44
- Tu, J., Gu, D., Wu, Y., & Yi, D. (2012). Error Modeling and Analysis for InSAR Spatial Baseline Determination of Satellite Formation Flying. Mathematical Problems in Engineering, 2012(), 359-381-025. https://doi.org/10.1155/2012/140301