臺灣博碩士論文加值系統

不管在軍事上或是搜救上的目標物偵測應用，短波紅外線高光譜影像都能提供很詳細的光譜資訊。大部分的高光譜取像是透過推掃式方式對靜態場景進行掃描，而本研究則是使用短波紅外線相機搭配液晶可調式光學濾波器擷取動態場景的高光譜影像。
由於蒐集到的影像中有干涉條紋，造成空間與頻譜上的失真而無法準確偵測目標物，本研究目的是要開發校正演算法以消除影像上的干涉條紋，並提高目標偵測率。
經過一系列的實驗探討干涉條紋的形成原因，發現從硬體端排除此干擾成本極高，因此改採用演算法進行排除。我們蒐集多組參考影像，提出三種校正方法：(1)以模板為基礎的消去法，直接利用補正值校正。(2)兩點校正法，利用兩個基礎點計算出線性特徵曲線進行校正。(3)多項式擬合校正法，利用多個基礎點以多項式擬合計算出非線性的特徵曲線進行校正。我們發現最後一種方法效果最好。三種方法的技術細節和校正後的結果記載在本論文中。
在物質辨識上，我們對於部分對焦影像與全對焦影像分別進行辨識，並使用了四種頻譜像素比對方法：(1) Spectral Angle Mapper；(2) Orthogonal projection divergence；(3) Orthogonal Subspace Projection；(4) Mahalanobis distance。最後利用不同方法辨識的結果比較及分析兩種影像。

Short-wave infrared (SWIR) hyperspectral image (HSI) provides detailed spectral information regardless of target detection applications in military or search and rescue. While most HSI tasks image a static scene by a push-broom scanner, here a liquid crystal tunable filter (LCTF) is mounted in front of a broad-band SWIR camera to take snapshots of dynamic scenes.
Due to the optical defects in LCTF, the collected images have pattern of interference fringes (PIF), which causes spatial and spectral distortion and interferes the target recognition tasks. Therefore, the purpose of this study is to develop a correction algorithm to eliminate PIF in the spectral images, and improve the recognition accuracy.
We experimentally investigated the cause of PIF, and found the cost of hardware solution is very high. Therefore, we turned to develop algorithmic solutions. By collecting the reference images with PIF only, we examined three correction methods: (1) Template-based elimination, which directly subtracts the reference from the collected image; (2) Two-point correction, which uses two referenced intensities to derive a linear curve to predict the original pixel intensities; (3) Polynomial fitting, which uses multiple reference intensities to derive a nonlinear curve to predict the original pixel intensities. We found the third method is the most effective. Their technical details and corresponding experimental results are documented in the thesis.

第一章 緒論 1
1.1 研究背景 1
1.2 研究動機與目的 2
1.3 論文大綱 5
第二章 相關文獻探討 6
2.1 液晶可調式光學濾波器 6
2.2 干涉條紋之影像校正 7
第三章 實驗方法 11
3.1 實驗內容 11
3.2 實驗場景及資料取得 13
3.3 干涉條紋影像校正方法 14
3.3.1 以模板為基礎的消去法 14
3.3.2 兩點校正法 15
3.3.3 多項式擬合校正法 16
3.4 部分對焦影像與全對焦影像 17
3.5 頻譜像素比對方法 17
3.5.1 Spectral Angle Mapper (SAM) 17
3.5.2 Orthogonal Projection Divergence (OPD) 18
3.5.3 Mahalanobis Distance (MAHA) 18
3.5.4 Orthogonal Subspace Projection (OSP) 18
第四章 實驗結果 20
4.1 實驗設備與環境 20
4.2 設備與環境因素對於干涉條紋影響 21
4.3 腳本設計對於光譜影像影響 23
4.4 影像校正實驗結果 24
4.4.1 以模板為基礎的消去法 25
4.4.2 兩點校正法 29
4.4.3 多項式擬合校正法 29
4.5 光譜影像校正之物質辨識實驗結果 30
4.6 部分對焦影像與全對焦影像之物質辨識實驗結果 34
4.6.1 評估指標 36
4.6.2 頻譜像素比對方法 37
第五章 結論與未來展望 40
5.1 適用於多項式擬合校正法的拍攝作業程序總結 40
5.2 實驗結果總結與未來展望 40
參考文獻 42

[1]R. N. Lane, “The SWIR advantage,” Proc. SPIE 2555, Airborne Reconnaissance XIX, 246– 254,1995.
[2]J. Cipar, G. Anderson and T. Cooley, "Active volcano monitoring using a space-based short-wave infrared imager," 3rd Workshop on Hyperspectral Image and Signal Processing: Evolution in Remote Sensing (WHISPERS), Lisbon, 2011.
[3]G. Sicot, M. Lennon, V. Miegebielle, D. Dubucq, “Estimation of the thickness and emulsion rate of oil spilled at sea using hyperspectral remote sensing imagery in the SWIR domain,” ISPRS Geospatial Week, France, 2015.
[4]R. H. Wilson, K. P. Nadeau, F. B. Jaworski, B. J. Tromberg, and A. J. Durkin, “Review of short-wave infrared spectroscopy and imaging methods for biological tissue characterization,” Journal of Biomedical Optics 20, 030901, 2015.
[5]L. L. Randeberg, J. Hernandez-Palacios, “Hyperspectral imaging of bruises in the SWIR spectral region,” Proc. SPIE 8207, Photonic Therapeutics and Diagnostics VIII, 2012.
[6]鄭玉權，禹秉熙, “成像光譜儀分光技術概覽,” 遙感學報, 2002.
[7]J.Y. Hardeberg, F.J. Schmitt, H. Brettel, “Multispectral color image capture using a liquid crystal tunable filter,” OPTICAL ENGINEERING VOL. 41, 2002.
[8]S.C. Gebhart, R.C. Thompson, A. Mahadevan-Jansen, “Liquid-crystal tunable filter spectral imaging for brain tumor demarcation,” APPLIED OPTICS Vol. 46, 2007.
[9]W. Wang, C. Li, E.W. Tollner, G.C. Rains, R.D. Gitaitis, “A liquid crystal tunable filter based shortwave infrared spectral imaging system: Design and integration,” Computers and Electronics in Agriculture 80 , pp.126-134 ,2012.
[10]I. Aizenberg, C. Butakoff, “A windowed Gaussian notch filter for quasi-periodic noise removal, ” Image and Vision Computing, vol. 26, pp. 1347-1353, 2008
[11]S. Colin, C. Benoit, “Automatic removal of fringes from EFOSC images,” Messenger 152, 14–16, 2013.
[12]馬亮, 危峻, 黃小仙, “使用背照減薄型CCD的色散型超光譜成像光譜儀中近紅外波段干涉條紋現象的研究與校正,” 光譜學與光譜分析 Vol. 34 No. 7, 2014
[13]A. Milton, F. Barone, M. Kruer, “Influence of nonuniformity on infrared focal plane array performance,” Opt. Eng., vol. 24, pp. 855–862, 1985.
[14]P. Fillon, A. Combette, P. Tribolet, “Cooled IR detectors calibration analysis and optimization,” Proc. of SPIE, vol. 5784, pp. 343-354, 2005.
[15]Y. F. Wu, “Edge Detection Based SWIR Image Bad Pixel Detection and Correction,” National Chiao-Tung University in Partial Fulfillment of the Requirements for the Degree of Master in Electrical and Control Engineering, 2011.
[16]Allied Vision, “Goldeye Technical Manual V2.6.0,” 2014
[17]F. A. Kruse, A. B. Lefkoff, J. B. Boardman, K. B. Heidebrecht, A. T. Shapiro, P. J. Barloon, and A. F. H. Goetz, “The Spectral Image Processing System (SIPS) - Interactive Visualization and Analysis of Imaging spectrometer Data.” Remote Sensing of Environment, v. 44, p. 145 – 163, 1993.
[18]J. C. Harsanyi and C.-I Chang, “Hyperspectral image classification and dimensionality reduction: An orthogonal subspace projection approach,” IEEE Trans. Geosci. Remote Sens., vol. 32, no. 4, pp. 779–785, Jul. 1994