每年夏天都會有數個颱風侵襲台灣，伴隨著強風及豪雨，造成生命財產上的損害。而最近幾年，豪雨過後的土石流災害更是嚴重。2009年的莫拉克颱風引發土石流，造成高雄縣甲仙鄉小林村滅村，數百人遭到活埋。經濟損失高達新台幣195 億元。2001 年桃芝颱風的土石流導致148 人死亡，失蹤。土石流發生的機率是由一個區域的雨量來決定，但使用者無法確切知道土石流是否真的發生。當土石流發生，通常造成很嚴重的損失和很多人死亡。確切知道土石流的發生，這是一件很重要的事。台灣政府在土石流易發生的區域架設攝影機，以掌握土石流的動態資訊。但這些設備僅僅是記錄這些資訊。本論文嘗試利用這些影像，建立一個演算法來自動判斷土石流是否發生。
這篇論文由兩個部分組成，偵測和辨識。藉由適應性的背景混合模型(adaptive background mixture models)，我們得以知道攝影機監測的範圍內是否有物體移動。偵測到移動物後，需要辨識它是否是土石流。本論文選取三組特徵資訊，紋理資訊(texture information)，色彩直方圖(color histogram)和鄰域顏色矩直方圖(neighborhood color moment histogram)。利用這三組特徵來代表背景和土石流。背景是選取測試影片的一組背景做萃取，土石流是用其他影片們的土石流資料。接著利用粒子群聚演算法(particle swarm optimization, PSO) 來訓練最能分別代表背景和土石流的數據。再萃取測試影片的特徵，並用完全限制最小平方差(fully constrained least squares, FCLS) 來比對，看現在的移動物是比較接近背景還是土石流。最後用模糊決策法(fuzzy decision method)來統合三組特徵，自動判斷結果。

Several typhoons strike Taiwan in the summer season every year. The strong breeze and heavy rain sometimes cause casualties and damages. In recent years, the loss by debris flow becomes more serious. The debris flow caused by typhoon Morakot in 2009 buried entire village of Shiaolin in Kaohsiung and killed hundreds of people. The economical loss is about 19.5 billion NTD. The debris flow caused by typhoon Toraji in 2001 kills 148 people. Government in Taiwan announces the possibility of debris flows by the amount of rain in the region. However we do not actually know the occurrence of debris flow, which is very important information. To monitor the debris flows, Government in Taiwan builds some charge-coupled device (CCD) camera in the region which the debris flow occurs often. However these systems just record the video without any processing. In this study, we attempt to build an automatic detect and recognize algorithm for debris flow from the video.
This study is composed of the detection and the recognition. The moving objects are first detected by adaptive background mixture models. And then these detected objects are recognized by the following procedure. Three types feature vectors is used to represent the debris flow and the background, which are texture features, color histogram and neighborhood color moment histogram. The feature vectors of background and debris flow are extracted from the background of testing video and other debris flow videos respectively. Particle swarm optimization (PSO) is then applied to select the best feature vectors of debris flow and background. We classify the feature vectors of testing video into the composition percentages of debris flow and background by fully constrained least squares (FCLS). Finally, the decision of debris flow is made automatically by fuzzy decision method.

[1] C. Stauffer and W. E. L. Grimson, “Adaptive background mixture models for real-time tracking,” in Computer Vision and Pattern Recognition, 1999. IEEE Computer Society Conference on., 1999, vol. 2.
[2] Chen Q., Tai X. Y., and Ba T. E., “Gastroscope image retrieval based on histogram of neighborhood color moments,” Computer Engineering and Applactions, vol. 11, no. 44, pp. 205-207, 2008.
[3] J. Kennedy and R. Eberhart, “Particle swarm optimization,” in Neural Networks, 1995. Proceedings., IEEE International Conference on, 1995, vol. 4, p. 1942–1948.
[4] D. C. Heinz, “Fully constrained least squares linear spectral mixture analysis method for material quantification in hyperspectral imagery,” Geoscience and Remote Sensing, IEEE Transactions on, vol. 39, no. 3, p. 529–545, 2001.
[5] W.-J. Wang, Fuzzy Introduction. Taiwan: Chuan Hwa, 2005.
[6] M. Fathy and M. Y. Siyal, “An image detection technique based on morphological edge detection and background differencing for real-time traffic analysis,” Pattern Recognition Letters, vol. 16, p. 1321–1330, Dec. 1995.
[7] A. J. Lipton, H. Fujiyoshi, and R. S. Patil, “Moving target classification and tracking from real-time video,” pp. 8-14, Oct. 1998.
[8] D. J. Dailey, F. W. Cathey, and S. Pumrin, “An algorithm to estimate mean traffic speed using uncalibrated cameras,” Intelligent Transportation Systems, IEEE Transactions on, vol. 1, no. 2, p. 98–107, 2000.
[9] B. K. P. Horn and B. G. Schunck, “Determining optical flow,” Artificial Intelligence, vol. 17, no. 1-3, pp. 185-203, Aug. 1981.
[10] J. L. Barron, D. J. Fleet, and S. S. Beauchemin, “Performance of optical flow techniques,” International Journal of Computer Vision, vol. 12, no. 1, pp. 43-77, 1994.
[11] P. Soille, Morphological Image Analysis: Principles and Applications. Springer-Verlag Telos, 1999.
[12] R. M. Haralick and L. G. Shapiro, Computer and Robot Vision, Vol. 1. Addison-Wesley, 1991.
[13] R. M. Haralick, S. R. Sternberg, and X. Zhuang, “Image analysis using mathematical morphology,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 9, p. 532–550, Jul. 1987.
[14] E. R. Dougherty, An Introduction to Morphological Image Processing. SPIE-International Society for Optical Engine, 1992.
[15] M. Sonka, V. Hlavac, and R. Boyle, “Image processing, analysis, and machine vision second edition,” International Thomson, 1999.
[16] R. M. Haralick, K. Shanmugam, and I. Dinstein, “Textural features for image classification,” Systems, Man and Cybernetics, IEEE Transactions on, vol. 3, no. 6, pp. 610-621, 1973.
[17] M. Partio, B. Cramariuc, M. Gabbouj, and A. Visa, “Rock texture retrieval using gray level co-occurrence matrix,” in Proc. of 5th Nordic Signal Processing Symposium, 2002, vol. 75.
[18] I. Christoyianni, E. Dermatas, and G. Kokkinakis, “Neural classification of abnormal tissue in digital mammography using statistical features of the texture,” in Electronics, Circuits and Systems, 1999. Proceedings of ICECS’99. The 6th IEEE International Conference on, 1999, vol. 1, p. 117–120.
[19] R. M. Haralick, “Statistical and structural approaches to texture,” Proceedings of the IEEE, vol. 67, no. 5, pp. 786-804, 1979.
[20] C.-F. Lee, S.-Y. Chang, and Z.-P. Lin, “Infrared Thermal Image Applied on Debris Flow Monitoring,” Journal of Chinese Soil and Water Conservation, vol. 35, no. 3, pp. 261-274, 2004.
[21] F. W. Billmeyer Jr, “Principles of color technology,” University of California, Los Angeles, 1981.
[22] R. C. Gonzalez, R. E. Woods, and S. L. Eddins, Digital image processing using MATLAB. Pearson Education India, 2004.
[23] R. B. Singer and T. B. McCord, “Mars-Large scale mixing of bright and dark surface materials and implications for analysis of spectral reflectance,” in Lunar and Planetary Science Conference Proceedings, 1979, vol. 10, pp. 1835-1848.
[24] Y. E. Shimabukuro and J. A. Smith, “The least-squares mixing models to generate fraction images derived from remote sensing multispectral data,” Geoscience and Remote Sensing, IEEE Transactions on, vol. 29, no. 1, pp. 16-20, 1991.
[25] C. I. Chang, X. L. Zhao, M. L. G. Althouse, and J. J. Pan, “Least squares subspace projection approach to mixed pixel classification for hyperspectral images,” Geoscience and Remote Sensing, IEEE Transactions on, vol. 36, no. 3, pp. 898-912, 1998.
[26] J. J. Settle and N. A. Drake, “Linear mixing and the estimation of ground cover proportions,” International Journal of Remote Sensing, vol. 14, no. 6, pp. 1159-1177, 1993.
[27] Te-Ming Tu, Chin-Hsing Chen, and Chein-I Chang, “A posteriori least squares orthogonal subspace projection approachto desired signature extraction and detection,” Geoscience and Remote Sensing, IEEE Transactions on, vol. 35, no. 1, pp. 127-139, Jan. 1997.
[28] C. I. Chang and D. Heinz, “Subpixel spectral detection for remotely sensed images,” IEEE Trans. Geosci. Remote Sensing, vol. 38, no. 3, p. 1144–1159, 2000.
[29] K. H. Haskell and R. J. Hanson, “An algorithm for linear least squares problems with equality and nononegativity constraints generalized,” Math. Prog., vol. 21, p. 98–118, 1981.