水中的沉降顆粒可以透過定點收集的方式記錄沉積物變化的歷程。沉積物收集在透明管中如同一般的岩心採樣，回收後必需拍照以保存原有收集樣貌。
沉積物分析具有時效性，因為常溫會使沉積物變質，所以拍照時需儘快完成。拍攝全貌圖需要高度解析度（畫素）來呈現出細部沉積物紋理，所以必需利用近拍及影像拼接來記錄全貌。沉積物結構鬆散容易因震動而導致擾動，所以不論是儲存或是拍攝時都必需維持樣品保持直立狀態。基於上述諸多限制因素，因此本研究發展一套輔助機構協助拍攝，並利用控制程式協調攝影位置及拍攝條件。此設備係利用鋁架組裝成穩定的拍攝平台，可垂直移動攝影機。燈源架設於岩心樣本兩側，提供高速快門所需的光源。影像擷取需同時考量解析度、光學扭曲、照明、曝光時間等因素，因為要在最短的時間完成，並減少轉換存檔以及影像拼接的時間，因此使用攝影機提供的Area of Interest (AOI)模式擷取有效線性範圍之影像，並搭配Linear Encoder所提供的位置資訊，有效提昇影像拼接的效率。本文研究對象影像屬顆粒狀、橫向特徵較多，故採用Harris Corner Detection萃取角點特徵。應用NCC(Normalized Cross Correlation)計算相關係數最高的匹配點得到初估位移量，再利用匹配的特徵點框出特徵範圍進行NCC運算取得微調後的位移量進行拼接，得到岩心全貌影像。

The changing process of suspension particles in water can be documented by collecting them in a sediment trap. Sinking particles are stored in a transparent tube similar to the conventional coring tubes. After recovering the sediment trap sample, it is necessary to take a series of photos to preserve the appearance of original layer structure. The sediment sample in the tube tends to degrade rapidly in room temperature. As a result, the image acquisition needs to be completed as soon as possible. Limited by the optics of the camera, a single high resolution image to
preserve the sediment texture can only be done by mosaicing a series of close shots. Therefore, we need to have a moving table which keeps a fixed distance between the camera and the tube while moving smoothly along the main axis of the tube. Sediment layered structure can be easily disturbed by vibration. So the core tube needs to maintain vertical and still while taking the image. The light source is mounted on both sides of the tube,to provide high-speed shutter required.By using Area of Interest (AOI) mode of the camera, only a small linear area of the image is taken for mosaic. This method reduces the acquisition time and mosiacing time significant.
Mosaicing technology lies in the image feature extraction and feature matching success rate of particle images in the settlement of this study Is more than the granular and lateral characteristics of the image, so the use of Harris Corner detection methods, the use of the characteristics of the corner feature extraction, and image
feature point of the overlap region search. Application of the NCC (on Normalized Cross Correlation) to calculate the correlation coefficient of the highest match points to get preliminary estimates suggest that more accurate displacement, and then use the matching feature points out of the box features range during NCC's calculations do fine-tuning operations to obtain the correct displacement of the stitching, to the automated core image capture mosaic technology.

目錄
論文審定書.........................................................................................................................i
中文摘要.............................................................................................................................ii
英文摘要.............................................................................................................................iii
第一章 緒論........................................................................................................................1
1.1 研究動機與目的...................................................................................................1
1.2 文獻回顧...............................................................................................................3
1.3 論文架構...............................................................................................................6
第二章 影像處理................................................................................................................7
2.1 影像前置處理.......................................................................................................7
2.2 邊緣檢測(Edge Detection) ...............................................................................10
2.3 特徵萃取(Feature extraction) ...........................................................................12
2.4 特徵匹配...............................................................................................................17
第三章 系統建置................................................................................................................19
3.1 硬體元件...............................................................................................................19
3.2 參數分析...............................................................................................................23
3.2.1 拍攝距離 .......................................................................................................23
3.2.2 線性不扭曲範圍 ...........................................................................................24
3.2.3 動態拍攝移動速度估算 ...............................................................................27
3.2.4 重疊區域(Overlap).......................................................................................31
3.3 岩心照相系統流程...............................................................................................34
第四章 影像拼接................................................................................................................36
4.1 影像拼接流程........................................................................................................36
4.2 影像拼接測試........................................................................................................40
4.2.1 自製影像疊合測試 ........................................................................................40
4.2.2 岩心影像測試 ................................................................................................46
4.3 岩心影像拼接.........................................................................................................48
4.3.1 USGS Trap-T10KP4影像拼接....................................................................51
4.3.2 K26重力岩心影像拼接.................................................................................54
第五章 討論與建議............................................................................................................56
5.1 討論........................................................................................................................56
5.2 建議........................................................................................................................57
參考書目.............................................................................................................................60
附錄A 岩心照相系統元件規格..........................................................................................62
A.1 感光耦合元件........................................................................................................62
A.2 計數器....................................................................................................................62
A.3 馬達控制器............................................................................................................62
A.4 伺服馬達................................................................................................................62
A.5 馬達控制器電路圖................................................................................................62
圖目錄
1.1 沉降顆粒收集示意圖..............................................................................................1
1.2 PPS3/3容量250ml收集瓶於台灣西南海域高瓶峽谷口收集樣品照片............2
1.3 沉積物收集器樣式.................................................................................................4
1.4 錨碇式沉積物收集器.............................................................................................5
2.1 影像灰階化.............................................................................................................8
2.2 高斯分布.................................................................................................................9
2.3 應用高斯平滑濾波濾除雜訊，左圖為原始灰階影像，右圖為濾波後影像。10
2.4 Prewitt遮罩............................................................................................................11
2.5 影像梯度運算........................................................................................................12
2.6 λ1及λ2的特徵關係示意圖....................................................................................14
2.7 驗證Harris特徵值特性之測試影像.....................................................................15
2.8 測試影像之特徵值特性........................................................................................15
2.9 上圖為測試影像，下圖為找出角點強度(R)最高值。......................................16
2.10 角點經旋轉角度後的強度值..............................................................................17
3.1 岩心照相系統.......................................................................................................20
3.2 感光耦合元件.......................................................................................................21
3.3 Linear Encoder...................................................................................................21
3.4 伺服馬達...............................................................................................................22
3.5 馬達控制器EZSV17............................................................................................22
3.6 凸透鏡成像關係圖...............................................................................................23
3.7 Barrel Distortion ................................................................................................25
3.8 Pincushion Distortion.......................................................................................25
3.9 定義線性不扭曲範圍之測試影像......................................................................26
3.10 找出影像中特徵並加以群組化..........................................................................26
3.11 固定攝影機及拍攝物條件下得到之線性不扭曲範圍......................................28
3.12 AOI(圖來源:USER'S MANUAL FOR GigE VISION CAMERAS P.209) ......28
3.13 影像中線性不扭曲範圍之實際度量單位..........................................................29
3.14 影響平台移動速度之因素..................................................................................30
3.15 計數器依不同時間紀錄資料量量化..................................................................31
3.16 比對特徵點，手動挑選兩張圖片相似特徵點，應用標準互相關係數計算
正確匹配位置如上圖..........................................................................................32
3.17 拼接影像重疊區域相同特徵點偏移像素..........................................................33
3.18 影像擷取流程圖..................................................................................................35
4.1 相鄰兩張影像拼接流程......................................................................................37
4.2 Harris角隅偵測，將兩張影像作角點偵測，並將偵測為角點位置用藍點
標示。..................................................................................................................38
4.3 角點強度直方圖，將所得角點強度以10為間距劃分區間.............................38
4.4 高相關係數像元配對的移動方向......................................................................39
4.5 特徵配對的位移角度直方圖..............................................................................39
4.6 median flow filter過濾後的匹配像元移動方向...............................................40
4.7 原始影像..............................................................................................................41
4.8 切割影像..............................................................................................................41
4.9 將兩張測試影像作角點偵測，並將偵測為角點位置用藍點標示。..............42
4.10 兩張影像特徵萃取之特徵強度直方圖..............................................................43
4.11 兩張影像尋得角點匹配所得結果，匹配點用綠色圓圈表示。......................43
4.12 匹配特徵點所制定，左圖為目標視窗，右圖為搜尋視窗。..........................44
4.13 兩張影像標示的匹配特徵區域做標準互相關係數計算所得結果..................44
4.14 左圖為影像拼接之結果，右圖用黑色方框表示拼接區域。..........................45
4.15 右圖為拼接影像，取圖像中左上小方框為目標視窗；左圖為原始影像，
取與拼接影像相同位置及方框大小區域作為搜尋視窗..................................45
4.16 拚接影像接合處與原始影像相同像素位置做標準化互相關係數結果 .........46
4.17 岩心擷取原始影像中，其中兩張連續影像。..................................................47
4.18 影像特徵萃取，將兩張影像作Harris角點偵測，並將偵測為角點位置用
藍點標示。..........................................................................................................47
4.19 特徵匹配並刪除不匹配點後之結果。..............................................................48
4.20 影像特徵匹配......................................................................................................49
4.21 疊合影像與原始影像強度曲線比較..................................................................50
4.22 凡娜比颱風路徑及錨碇佈放站位(圖來源:中央氣象局)..................................51
4.23 T10KP4樣品高度示意圖...................................................................................52
4.24 環管拼接圖..........................................................................................................53
4.25 K26高度示意圖..................................................................................................54
4.26 環管拼接圖.........................................................................................................55
A.1 馬達控制器電路圖..............................................................................................65
表目錄
3.1 1.4m高岩心樣本取樣影像高度3cm，依不同重疊率所擷取影像數量計算
表..........................................................................................................................32
A.1 CCD規格表.........................................................................................................63
A.2 LSI-5123/A/L計數器規格表...............................................................................63
A.3 馬達控制器規格表..............................................................................................63
A.4 CR63 × 55直流伺服馬達規格表......................................................................64

參考書目
[1] 蔡憲裕，半自動化異元高解析力遙測影像加強適應性套合研究，國立中央大學土木工程研究所碩士論文，2010。
[2] 陳柏棋，海床粗糙度雷射掃描系統資料處理分析軟體之開發，國立中山大學海下技術研究所碩士論文，2006。
[3] 李振偉，水下無人載具之視覺懸停控制，國立中山大學機械與機電工程學系碩士論文，2008。
[4] 雲冠群，基於角點偵測技術應用於光達資料之建物輪廓提取，國立中山大學海洋環境及工程學系碩士論文，2008。
[5] Huh, C.-A., J.T. Liu, H.-L. Lin, and J.P. Xu , “Tidal and flood signatures of settling particles in the Gaoping submarine canyon (SW Taiwan) revealed from radionuclide and flow measurements. “ Marine Geology, 267,8-17,2009.
[6] H. Moravec, “Obstacle Avoidance and Navigation in the Real World by a Seeing Robot Rover”, “ Tech Report CMU-RI-TR-3, Carnegie-Mellon University, Robotics Institute, September 1980.
[7] C.G. Hariss and M.J. Stephens,”A combined corner and edge detector,” Proceedings of 4th Alvery Vision Conference,Manchester:The University of Sheffield Printing Unit,pp.147-151,1988.
[8] H. Ojanen, “Automatic Correction of Lens Distortion by Using Digital Image Processing ,” Proceedings of the 1995 IEEE IECON 21st International Conference on Industrial Electronics, Control, and Instrunmentation,10 July. 1999, Vol. 1 pp. 173-178, 1999.
[9] B. Rousso, S. Peleg, Finci I., and Rav-Acha A,”Universal mosaicing using pipe projection”, In Sixth International Conference on Computer Vision (ICCV'98), pages 945–952, Bombay, January 1998.
[10] M. A. Brown,”Multi-Image Matching using Invariant Features”,PhD thesis, University of British Columbia, 2005.
[11] P. McLauchlan and A. Jaenicke,”Image mosaicing using sequential bundle adjustment”,In Proc.of the British Machine Vision Conference BMVC2000, Bristol,U.K., September 2000.
[12] D. Charnley and R.J. Blissett,”Surface reconstruction from outdoor image sequences”,Image and Vision Computing,Volume 7, Issue 1, Pages10-16, 1989.
[13] N. Chiba, H. Kano, M. Higashihara, M. Yasuda and M. Osumi,”Feature-based Image Mosaicing”,Proc.Image and Vision Computing.IAPR Workshop on Machine Vision Applications, pp. 5–10, 1998.
[14] L. Gu,”An Automatic Target Image Mosaic Algorithm”, The IEEE Region 8 Eurocon 2007 Conference,pp.369-374,2007.
[15] M.J. Stephens, C.G. Harris,”3D wire-frame integration from image sequences”, Image and Vision Computing, Volume 7, Issue 1, Pages 24-30, 1989.
[16] S.-H. Cho, Y.-K. Chung, and J. Y. Lee. “Automatic image mosaic system using image feature detection and taylor series”. In Proceedings of VIIth Digital Image Computing: Techniques and Applications, Pages 549-556, 2003.
[17] Y. Yu, K. Huang, and T. Tan,”A Harris-like scale invariant feature detector”, In Proc. Asian Conf. Computer. Vis., Pages 586-595, 2009.