航空攝影測量技術發展的最終目標為對地面進行快速且精確的測量，然而，傳統攝像測量在連續影像的拼接技術上仍然有相當的限制。過去利用現場布置地面控制點來進行影像間的點對點拼接，常消耗大量人物力，加上人類視覺上的限制，航測影像拼接精度與效率的提升成為拼接技術的關鍵因素。SIFT（Scale Invariant Feature Transform）演算法為利用電腦視覺進行特徵提取的演算法，其特徵點具有抗影像尺寸、角度及亮度改變的特性，並提供大量非視覺性的特徵點，是一高穩定性與多量性的特徵點提取演算法。SIFT演算法利用多尺度空間的角度切入來提取特徵點，對於資訊量豐富且影像幅度較大的航測影像而言，特徵提取運算時間較長，所以本研究提出間隔取樣法（IGDS）降低影像尺寸與相對資訊量以提高運算效率，並且將所取得特徵點進行相鄰影像間的特徵匹配，配合RANSAC除錯機制挑選出正確且具代表性的特徵點對，最後利用Hugin全景影像拼接軟體進行航帶連續影像拼接以獲得拼接全景圖。
研究結果顯示在利用間隔取樣法縮小影像尺寸3倍後，特徵位置差異值在次像元以內，並且降低了一半的運算時間，相對於最鄰近內插與立方迴旋內插的影像尺寸縮小法而言，間隔取樣法可以在不失位置精度下有效的提升特徵提取速率。而SIFT特徵點匹配門檻值在0.4至0.6間可以依照影像類型取得最大正確與最少錯誤匹配特徵點數量，並配合RANSAC除錯機制有效的挑選出最佳的匹配特徵點數量與位置。在影像拼接方面，Hugin全景影像拼接軟體可以有效的利用大量的特徵點，進行幾何校正和色彩調整，獲得具有一致性的全景出圖，並輸出解析度變動性低且具量測意義的連續航帶拼接影像。

The ultimate goal of the development of aerial photogrammery is to acquire rapidly and accurately the ground measurements. However, traditional photogrammetric technologies, particularly in the continuous digital images stitching technique, is still very limited. In the past, the ground control points were used as the references for the image registration, however, it is very time and resource consuming, as well as human visual capability constraint. Accuracy and efficiency are two key factors which need to be enhanced to meet the practical requirement for aerial imageries stitching. The SIFT (Sale Invariant Feature Transform) algorithm was used in the computer vision to perform feature extraction in good condition. The extracted SIFT features are invariant to image scale, rotation, noise and change in illumination, and it is a robust and abundant feature extraction algorithm. SIFT algorithm extracts feature points from multi-scale space. For a large scale aerial image containing huge amount of image contents, it will spend a lot of time to extract features from imagery. Therefore, this study proposes a new method, called Inter-Grid Down-Sampling (IGDS) method, to reduce the image size and relative amount of image information to improve the computing efficiency. The correspondent extracted features are matched in the adjacent images with additional RANSAC outlier removal procedure to select correct and characteristic feature points. Finally the Hugin-Panorama Photo Stitching software is used to stitch all the continuous photogrammetric images for producing a panorama imagery of all flight lines.
The experiment results indicate that sub-pixel accuracy for extracted feature points can be obtained when the down-sampling factor 3 was selected for the IGDS method, and it only needs half of the computing time. Compared to the Nearest-Neighbor Interpolation and Cubic Interpolation methods to reduce the image size, the IGDS method can increase more feature extraction efficiency without scarifying the location accuracy. When threshold value for SIFT was set between 0.4 to 0.6, we can achieve the largest correct matching rate. In addition, the RANSAC outlier removal procedure can effectively select the best matching feature points both in numbers and locations. For image stitching, the Hugin-panorama photo stitching software can effectively be used to match feature points and do geometric correction and color adjustment to obtain a consistent panorama imagery. Finally, the proposed method in this study can derive a low-variant in resolution and measurements significance for a stitching image from continuous aerial images.

中文摘要 I
Abstract II
誌謝 IV
目錄 VI
圖目錄 X
表目錄 XVI
第一章 前言 1
1-1 研究動機與目的 1
1-2 文獻回顧 3
1-3 研究方法 8
1-4 研究流程架構 9
1-5 論文架構 11
第二章 基於SIFT演算法的快速特徵提取 13
2-1 特徵點提取 13
2-1-1 尺度空間極值的求取 14
2-1-2 DoG影像極值 17
2-2 特徵點位置確定17
2-2-1 過濾低對比度特徵點18
2-2-2 過濾邊緣處特徵點 18
2-3 特徵點梯度方向確定 20
2-4 特徵點描述子21
2-5 SIFT特徵點匹配 22
2-6 匹配除錯 23
2-7 應用SIFT方法於拼接大型影像的方法 25
2-7-1 間隔取樣法 27
2-7-2 影像縮小尺寸方式比較28
2-7-3 影像尺寸縮小可行性精度評估分析 29
2-7-4 影像分析 30
第三章 影像拼接 39
3-1 航帶鄰近元件辨識 40
3-2 影像幾何轉換 42
3-2-1 影像扭曲校正 43
3-2-2 透視投影校正46
3-2-3 自動全景拉直50
3-3 影像色彩強度混合 52
3-3-1 光暈與曝光調整 52
3-3-2 多頻段色彩混合 57
第四章 實驗成果與討論 62
4-1 實驗區資訊62
4-2 實驗程序 63
4-3 航帶編輯 65
4-4 影像灰階程序65
4-5 航拍影像尺寸縮小差異分析 66
4-5-1 影像尺寸縮小分析結果（一）67
4-5-2 影像降階分析結果（二） 71
4-6 影像匹配測詴 76
4-7 航帶影像拼接運算時間分析 88
4-8 出圖焦距測詴 90
4-9 航帶拼接實驗 95
4-9-1 多影像完整航帶拼接實驗 96
4-9-2 多變化航拍影像拼接實驗99
4-9-3 大型影像拼接實驗101
4-9-4 影像拼接錯誤補強實驗 103
第五章 結論與建議 109
5-1 結論109
5-2 建議112
參考文獻 114
附錄一Hugin軟體操作說明 119
附錄二程式列表 125

李良輝（1991），計算視覺方法於SPOT衛星影像之幾何特性分析，國立中央大學大氣物理研究所博士論文。
李寒、牛紀楨、郭禾（2007），基於特徵點的全自動無縫圖像拼接方法，機算機工程與設計 第28卷 第9期 第2083-2085頁。
朱廣新（2007），基於特徵點匹配的圖像拼接及醫學應用，南京理工大學計算機科學與技術學院碩士論文。
汪華琴（2007），基於特徵點的圖像拼接方法研究，華中師範大學計算機運用技術碩士論文。
何維信（1995），航空攝影測量學，國立編譯館主編，大中國圖書公司印行。
陳良健、林士哲（2006），空載三線式掃描儀影像立體定位之研究，航測及遙測學刊 第11卷 第4期 第381-402 頁。
陳金徽（2006），數位影像輻射品質之客觀指標評估，國立成功大學地球科學研究所碩士論文。
林事乾、杜明鴻（2005），影像變形專題實做，國立台灣大學資訊工程學系專題，[線上].
http://www.csie.ntu.edu.tw/~r93944019/932/VFX/project1/#m3。
高超、張鑫、王云麗、王暉（2007），一種基於SIFT特徵的航拍影像序列自動拼接方法，計算機應用 第27卷 第11期 第2789-2792頁。
章國鋒、秦學英、董子龍、華煒、鮑虎軍（2006），面相增強視頻的基於結構和運動恢復的攝相機定標，Chinese Journal of Computers 第29卷 第12期 第2104-2111頁。
張庭榮（2008），SIFT 演算法於立體對影像匹配與影像檢索應用之研究，國立高雄應用科技大學土木工程與防災科技研究所碩士
論文。
楊子賢（2002），從視訊串流合成彷真的臉部模型，國立成功大學資訊工程系碩士論文。
劉學東、劉波、程學軍、陳鵬霄（2004），山區流域航空攝影關鍵技術探討，長江科學院院報 第21卷 第3期 第54-56頁。
Andrew Mihal(2004), enblend/Enfuse, [Online], Available: http://enblend.sourceforge.net/index.htm.
Autopano-SIFT(2005), making panoramas fun, [Online], Available: http://user.cs.tu-berlin.de/~nowozin/autopano-sift/.
Beis, J. S. and Lowe, D. G( 2003), Shape indexing using approximate
nearest-neighbor search in high-dimensional spaces, Conference on Computer Vision and Pattern Recognition, pp1000--1006.
Brown, M. and Lowe, D. (2003), Recognizing panoramas. In Ninth International Conference on Computer Vision (ICCV’03), pages 1218–1225, Nice, France.
Burt, P. and Adelson, E.(1983), A multiresolution spline with application to image mosaics. ACM Transactions on Graphics, No.2(4),pp.217-236.
Burt, P.and Adelson, E.(1983),The Laplacian Pyramid as a Compact Image Code. IEEE Transactions on Communications.
Canny, J.（1986）.A Computational Approach to Edge Detection. IEEE Computer Society Washington, DC, USA, pp.679-698.
David G. Lowe,(1999), Object Recognition from Local Scale-Invariant Features. International Conference on Computer Vision, Corfu, Greece , pp. 1150-1157.
David G. Lowe,(2004), Distinctive image features from
scale-invariant Key points. International Journal of Computer Vision, No.60, pp. 91-110.
David G. Lowe,（2006）, Automatic Panoramic Image Stitching using Invariant Features. International Journal of Computer Vision, No. 74(1), pp.59–73.
Fishchler M A.(1981), Random sample consensus: a paradigm for model fitting with application to image analysis and automated cartography. Communication Association Machine, No.24(6), pp.381-395.
Harris, C. and Stephens, M.(1988), A combined corner and edge detector. In Fourth Alvey Vision Conference, Manchester, UK, pp.147-151.
Helmut Dersch(2001), Panorama Tools, [Online], Available: http://www.all-in-one.ee/~dersch/.
Helmut Dersch(2007), Panorama Tools- Open Source Software for Immersive Imaging. International VR Photography Conference Berkeley 2007, [Online], Available: http://webuser.fh-furtwangen.de/~dersch/IVRPA.pdf.
hugin - Panorama photo stitcher, [Online], Available: http://hugin.sourceforge.net/.
Koenderink, J.J. (1984),The structure of image. Biological Cyberneyics, No.50, pp.363-396.
Lindeberg T(1994), Scale-space theory: A basic tool for analyzing structures at different scales, Journal of Applied Statistics, No.21(2):224-270,1994.2(60):91-110.
Lindeberg T.(1998), Feature detection with automatic scale selection.
International Journal of Computer Vision, No.30(2),pp.79-116.
Media Wiki. Pano Tools, [Online], Available: http://wiki.panotools.org/.
Mikolajczyk, K.(2002), Detection of local features invariant to affine transformations, Ph.D. thesis, Institut National Polytechnique de Grenoble, France.
Mikolajczyk, K. and Schmind, C.(2002), An affine invariant interest point detector. In European Conference on Computer Vision(ECCV), Copenhagen, Denmark, pp.128-142.
Pablo d’Angelo.(2007), Radiometric alignment and vignetting calibration. The 5th International Conference on Computer Vision Systems.
Qian Du, Nareenart Raksuntorn, Adnan Orduyilmaz, and Lori M. Bruce（2008）, Automatic Registration and Mosaicking for Airborne Multispectral Image Sequences. Photogrammetric Engineering & Remote Sensing Vol. 74, No. 2, pp. 169–181
Shmuel Peleg, Moshe Ben-Ezra(1999), Stereo Panorama with a Single Camera, Proc. IEEE Conf. Computer Vision and Pattern Recognition, CVPR’99, pp. 395-401.
Sidney F. Ray (2002), Applied photographic optics, Third Edition.
Szeliski, R. and Kang, S.(1995), Direct methods for visual scene reconstruction. In IEEE Workshop on Representations of Visual Scenes. Cambridge, MA, pp.26-33.
Thaddeus Beier, Shawn Neely(1992), Feature-Based Image Metamorphosis, Computer Graohics, ACM SIGGRPH Conference 1992, Vol.26, No.2, pp.35-42.
Thorsten M. Buzug, Dietrich Holz, Jens Bongartz, Matthias Kohl-Bareis, Ulrich Hartmann and Simone Weber(2007), An Algorithm for Automatic Stitching of CR X-ray Images. Advances in Medical Engineering, No.114, pp.193-198.
Triggs, W., McLauchlan, P., Hartley, R., and Fitzgibbon, A.(1999), Bundle adjustment: A modern synthesis. In Vision Algorithms: Theory and Practice, No.1883 in LNCS. Springer-Verlag. Corfu, Greece, pp.298-273.
Xiang Fan, Shun-ren Xia(2007), Feature Based Automatic Stitching of Microscopic Images. Advanced Intelligent Computing Theories and Applications. With Aspects of Contemporary Intelligent Computing Techniques, No.2, pp.791-800.
Wikipedia, Flight Dynamics, [Online], Available: http://en.wikipedia.org/wiki/Pitch_(flight).