臺灣博碩士論文加值系統

由於地表物種的分布狀態與地球能量收支平衡及水文循環等作用息息相關，在氣候模式中是重要的初始場參數之一，因此準確且有效的地物分類，將有助於氣候預報模式等相關研究之發展。
空間相干法( spatial coherence method )一向被應用於處理紅外光波段的衛星影像，主要為區分晴天視場與陰天視場，將雲和晴空海域區隔出來，其主要優點為濾除影像中部分有雲之混合像元(mixing pixel)。混合像元在應用衛星影像於地物分類的過程中，亦為誤差的來源之一，尤其在地物分布複雜的地區。因此，本研究將根據空間相干法的優點，濾除地表覆蓋物的混合區域，求得研究試區中各純種地物的光譜特性，應用於衛星影像之地物分類。空間相干法在建立物種光譜特性的過程中，將配合監督式分類法中的最大似然法，協助制定各類物種之光譜範圍最佳標準差，亦即結合非監督式與監督式分類法的優點，改進其共同的缺點──受混合像元之影響。
研究結果顯示，空間相干法在地物分類應用上有不錯的成效，與最大似然法之分類結果相當，特別是配合高解析度的SPOT5衛星觀測資料，可將道路明確從都會區中分出，顯示此分類方法具極高之可行性與潛力，非常適合應用於地物種類複雜的地區，將可成為地物分類應用的新方法。

Because of being strongly related to the heat budget of the Earth, hydrologic cycle, and as an important input parameter for climate models, an accurate and effective algorithm for landcover classification could make a great contribution to the climate model and environmental change researches.
The spatial coherence method is originally developed to distinguish cloudy sky pixels from clear sky pixels all the time via satellite IR images, which is ascribing its key advantage of filtering out the mixing pixels (i.e. partial cloudy pixels) to get the pure values of certain surface covers (i.e. clear sky pixels or total cloudy pixels) under radiometric considerations. It is well known that the mixing pixel effect is one of main error sources in the classification tasks. Therefore, in this research the spatial coherence method is specifically applied as an alternative way in classifying landcovers, filtering out the mixing pixels to get the pure spectral characteristics of certain landcovers. This classification method seeks the assistance of the maximum likelihood method to pick up the best standard deviation, meaning that the method combines the advantages of the supervised and unsupervised classification methods.
Result shows that a comparable performance to the maximum likelihood method does could be observed. Applying this method with the high-resolution SPOT5 images, it shows that the road pixels can be competently extracted from urban areas, revealing its high practicability and capability for typical Taiwan landcover patterns, and making it become a brand-new landcover classification method.

摘 要 I
ABSTRACT I
致 謝 III
目 錄 V
表目錄 VIII
圖目錄 X
符號說明 XV
一、前 言 1
1.1 研究背景 1
1.2 研究方法回顧 2
1.3 研究動機與目的 4
二、研 究 方 法 6
2.1 地物之分類 6
2.2 空間相干法 8
2.3 高斯分布擬合 9
2.3.1斜率法 9
2.3.2 三點法 10
2.3.3 最小平方法 11
2.4 地物的反射特性與類別 12
2.5 最大似然法 13
三、研究資料與步驟 15
3.1 SPOT5衛星與資料特性 15
3.2 土地利用基本圖 16
3.3 資料前置處理 16
3.3.1 研究範圍 16
3.3.2切斷標準差門檻值 17
3.4 研究步驟 18
3.4.1 局地極大?的判斷 18
3.4.2 波譜特徵分析與頻道整併 19
3.4.3 地物歸類 20
3.4.4 物種標準差最佳化 23
四、結果分析與驗證 26
4.1 監督式分類結果 26
4.2 空間相干法分類結果 28
4.3 測試與驗證 30
五、結論與展望 32
5.1 結論 32
5.2 展望 33
參 考 文 獻 35
附表 38
附圖 55

李瑞陽，姜如憶，「應用遙測技術於水稻田判釋之研究」，地理研究，第43期，第62-64頁，民國94年11月。
徐建瑤，「由同步衛星紅外線影像之雲追蹤估算風場」，國立中央大學大氣物理研究所碩士論文，第17-19頁，中壢市，民國81年6月。
國立中央大學太空及遙測研究中心資源衛星接收站使用者手冊，國
立中央大學太空及遙測研究中心，第四版，85頁，中壢市，民國
94年5月。
曾忠一，大氣衛星遙測學，初版，第286-289頁，國立編譯館主編，渤海堂文化事業公司，台北市，民國77年。
曾忠一，大氣遙測原理與應用，初版，第91-97頁，中央氣象局，台北市，民國72年10月。
Biggs, T. W., P. S. Thenkabails, M. K. Gumma, C. A. Scott, G. R. Parthasaradhi, and H. N. Turral, 2005:”Irrigated area mapping in heterogeneous landscapes with MODIS time series, ground truth and census data, Krishna Basin, India.” J. Remote Sens., Vol. 27, pp. 4245-4266.
Bolstad, P. V., and T. M. Lillesand, 1991:” Rapid maximum likelihood classification.” Photogram. Eng. Remote Sens., Vol. 57, pp. 67-74.
Chen, X., M. T. Jabbar, X. Cai, Z. Wu, and Y. Liang, 2005: “Detection and
evaluation of vegetation change and urbanization in the central China.” IEEE, Vol.1, pp. 230.
Coakley, J. A., Jr., and F. P. Bretherton, 1982: “Cloud cover from high-resolution scanner data : Detecting and allowing for partially filled fields of view.” J. Geophys. Res., Vol.87, pp. 4917-4932.
D''Addabbo, A., G. Satalino, G. Pasquariello, and P. Blonda, 2004: “Three different unsupervised methods for change detection: an application.” IEEE, Vol. 3, pp. 20-21.
Donoghue, D. N. M., and N. Mironnet, 2002,”Development of an integrated
geographical information system prototype for coastal habitat monitoring.” Computers and Geosciences, Vol. 28, 129-141.
Giles M. Foody, and Ajay Mathur, “Toward intelligent training of
supervised image classifications: directing training data acquisition for
SVM classification,” Remote Sens. of Environ., Vol. 93, pp.108-110, 2004.
Karl, T. R., H. F. Diaz, and G. Kukla, 1988:’’ Urbanization: Its Detection and Effect in the United States Climate Record.” J. Climate, Vol. 1, pp. 1099-1123.
Kishore Das, D., Gopal Rao, K., Prakash, A, “Improvement of Effective
Spatial Resolution of Thermal Infrared Data for Urban Landuse
Classification,” IEEE/ISPRS Joint Workshop on Remote Sensing and
Data Fusion over Urban Areas, Vol. 8, pp.332, 2001.
Lim, H. S., M. Z. MatJafri, and K. Abdullah, 2003: “Evaluation of conventional digital camera scenes for thematic information extraction.” School of Physics University Saints Malaysia.
Long, W., III, and S. Sriharn, 2004: “Land cover classification of SSC image: unsupervised and supervised classification using ERDAS Imagine.” IEEE, Vol. 4, pp. 2707-2709.
Pinheiro, A., H. Carrao, and M. Caetano, 2007: “Evaluation of ASAR and optical data synergy for high resolution land cover mapping in Portugal.” IEEE, pp. 1517.
Qin, L., Q. Zheng, S. Jiang, Q. Huang, and W. Gao, 2008: “Unsupervised texture classification: Automatically discover and classify texture patterns.” J. Remote Sens, Vol. 26, pp. 333-334.
Rosenberg, D. M., 2000:” Global-scale environment effects of hydrological alterations.” BioScience, Vol. 50, pp. 746-751.
Thomas M. Lillesand, Ralph W. Kiefer, Jonathan W. Chipman, Remote
Sensing And Image Interpretation, Fifth Edition, pp. 552-573, John Wiley& Sons, U. S., 2004.
Tou, J. T., and R. C. Gonzalrz, 1974:” Pattern recognition Principles.” Addison-Wesley, London.
Weng, Q., D. Lu, and J. Schubring, 2004: ”Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies.” Remote Sens. of Environ., Vol. 89, pp. 467-483.
W. Paul Menzel, Notes on Satellite Meteorology, pp.6-1 – 6-3, December 1995.