據統計資料顯示，住院病人或機構式照護中心的住民皆可能在日常活動中發生跌倒，導致跌倒事件發生的原因包含生理健康、使用藥物、環境因素等，其中年長者的跌倒事件更是不可忽視的問題，因為跌倒意外可能造成輕微擦傷，嚴重者可能導致骨折甚至有生命的危險。若有一套即時跌倒偵測系統，在跌倒事件發生時能立即向醫護人員或照護人員發出警訊，便能在最短時間提供緊急處理以降低傷害。而電腦視覺提供了可行的解決方法，可以透過影像來分析個人的行為以及偵測某些異常事件，例如跌倒。
因此，本論文希望以機器視覺為基礎開發一套智慧型行人跌倒偵測系統。透過裝設於環境中的攝影機取得影像資訊，首先經由高斯混合模型建立背景影像，以背景相減法擷取出前景影像，此時的前景影像包含了許多雜訊，並且移動物件有破碎不連續的情形，利用影像形態學修補物件，以連通物件標記法去除面積較小之物件達到消除雜訊的目的，即可得到有效的前景影像。跌倒偵測部分以移動物件的邊界框長寬比、擬合橢圓長軸角度及中心點縱向速度作為學習演算法的輸入特徵，經由推論結果判斷是否有跌倒事件發生。本論文之學習演算法包含倒傳遞類神經網路及自適應增強演算法兩種，並分別計算不同輸入特徵個數的跌倒偵測準確率。

Statistics shows that inpatients or residents of long-term care facilities may fall from daily activities. Most of these falls are associated with identifiable risk factors (e.g. weakness, unsteady gait, medications use and environment). Fall accidents can cause abrasions, serious broken bones, or even death. Therefore, falls in older people cannot be ignored. A real time fall detection system should be developed, which can alarm nurses in real time once a fall event occurs. Computer vision provides a promising solution, and utilizes image recognition technology to analyze personal behavior and detect certain abnormal events such as falls.
The thesis develops an intelligent pedestrian fall detection system based on machine vision. The image sequences are got from a camera which is installed in the indoor environment. Firstly, the background model is built up by Gaussian mixture model (GMM), and the foreground is extracted by background subtraction. Morphological operations are then used to repair damage to the image and connected-component labeling method is used to eliminate of noise. From the motion object of foreground, the aspect ratio of bounding box, the orientation of the ellipse and the longitudinal velocity of center point are extracted as input features of the learning algorithm. Finally, fall detection is accomplished according to the classification results of learning algorithm with back propagation neural network and adaptive boosting algorithm.

論文審定書 i
誌 謝 ii
中文摘要 iii
Abstract iv
目 錄 v
圖 次 vii
表 次 xi
第一章 緒 論 1
1-1 研究動機 1
1-2 文獻回顧 3
1-3 論文貢獻 8
1-4 章節介紹 8
第二章 系統概述 9
2-1 系統平台 9
2-1-1硬體設備與作業系統 9
2-1-2軟體開發套件 10
2-2 系統架構流程 12
第三章 系統實現 14
3-1 影像前景擷取 14
3-1-1 以高斯混合模型建立背景影像 14
3-1-2 影像強化 22
3-2 特徵萃取 28
3-3 具學習演算法之跌倒偵測 30
3-3-1 倒傳遞類神經網路[28] 31
3-3-2自適應增強演算法(AdaBoost) 35
第四章 實驗結果 39
4-1 實驗場景 39
4-2 影像前景擷取 43
4-3 特徵萃取 45
4-4 跌倒偵測 48
4-4-1 不同影像特徵 49
4-4-2不同攝影機視角 51
4-4-3兩種演算法不同特徵個數比較 53
第五章 結論與未來展望 59
5-1 結論 59
5-2 未來展望 59
參考文獻 60

[1] 「中華民國2012年至2060年人口推計」簡報。行政院經建會人力規劃處。
[2] 「65歲以上國人進住長期照顧及安養機構比率1.6%」國情統計通報(第238號)。行政院主計總處綜合統計處。
[3] 王淑慧, “機構式長期照護住民之跌倒危險性及相關因素探討,” 碩士論文, 嘉南藥理科技大學醫療資訊管理研究所, 2009.
[4] S. Deandrea, F. Bravi, F. Turati, E. Lucenteforte, C. L. Vecchia, and E. Negri, “Risk factors for falls in older people in nursing homes and hospitals. A systematic review and meta-analysis,” Archives of Gerontology and Geriatrics, vol. 56, pp. 407 – 415, 2013.
[5] 「台灣病人安全通報系統 2010年年報」國情統計通報(第238號)。行政院衛生署。
[6] M. Mubashir, L. Shaoa, and L. Seed, “A survey on fall detection: Principles and approaches,” Neurocomputing, vol. 100, pp. 144 – 152, 2013.
[7] J. Wang, Z. Zhang, B. Li, S. Lee, and R. S. Sherratt, “An enhanced fall detection system for elderly person monitoring using consumer home networks,” IEEE Transactions on Consumer Electronics, vol. 60, pp. 23 – 29, 2014.
[8] N. Fourty, D. Guiraud, P. Fraisse, G. Perolle, I. Etxeberria, and T. Val, “Embedded system used for classifying motor activities of elderly and disabled people,” Computers & Industrial Engineering, vol. 57, pp. 419 – 432, 2009.
[9] W. C. Cheng and D. M. Jhan, “Triaxial accelerometer-based fall detection method using a self-constructing cascade-AdaBoost-SVM Classifier,” IEEE journal of biomedical and health informatics, vol. 17, no. 2, pp. 411 – 419, 2013.
[10] A. Yazar, F. Keskin, B. U. Töreyin, and A. E. Çetin, “Fall detection using single-tree complex wavelet transform,” Pattern Recognition Letters, vol. 34, pp. 1945 – 1952, 2013.
[11] Y. Li, K. C. Ho, and M. Popescu, “Efficient source separation algorithms for acoustic fall detection using a Microsoft Kinect,” IEEE Transactions on Biomedical Engineering, vol. 61,no. 3, pp. 745 – 755, 2014.
[12] B. U. Töreyin, Y. Dedeoğlu, and A. E. Çetin, “HMM based falling person detection using both audio and video,” IEEE Conference on Signal Processing and Communications Applications, 2006, pp. 1 – 4.
[13] L. Alhimale, H. Zedan, and A. A. Bayatti, “The implementation of an intelligent and video-based fall detection system using a neural network,” Applied Soft Computing, vol. 18, pp. 59 – 69, 2014.
[14] M. Yu, Y. Yu, A. Rhuma, S. M. R. Naqvi, L. Wang, and J. A. Chambers, “An online one class support vector machine-based person-specific fall detection system for monitoring an elderly individual in a room environment,” IEEE Journal of Biomedical and Health Informatics, vol. 17, pp. 1002 – 1014, 2013.
[15] S. W. Yang and S. K. Lin, “Fall detection for multiple pedestrians using depth image processing technique,” Computer Methods and Programs in Biomedicine, vol. 114, pp. 172 – 182, 2014.
[16] C. Rougier, J. Meunier, A. St-Arnaud, and J. Rousseau, “Robust video surveillance for fall detection based on human shape deformation,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 21, no.5, pp. 611 – 622, 2011.
[17] A. Sengto and T. Leauhatong, “Human falling detection algorithm using back propagation neural network,” Biomedical Engineering International Conference (BMEiCON), 2012, pp. 1– 5.
[18] 羅技C920網路攝影機：http://www.logitech.com/zh-tw/product/hd-pro-webcam-c920?crid=34
[19] G. Bradski and A. Kaehler, Learning OpenCV, O'Reilly Media, Inc., 2008.
[20] K. Kim, T. H. Chalidabhongse, D. Harwood, and L. Davis, “Real-time foreground–background segmentation using codebook model,” Real-Time Imaging, vol. 11, pp. 172 – 185, 2005.
[21] C. Stauffer and W. E. L. Grimson, “Adaptive background mixture models for real-time tracking,” IEEE Conference on Computer Vision and Pattern Recognition, vol. 2, 1999, pp. 22 – 29.
[22] D. Reynolds, “Gaussian mixture models,” Encyclopedia of Biometric Recognition, pp. 1– 5, 2008.
[23] Z. Zivkovic, “Improved adaptive Gaussian mixture model for background subtraction,” International Conference on Pattern Recognition (ICRP), vol. 2, 2004, pp. 28– 31.
[24] 求是科技, “Visual C++ 數位影像處理技術大全,” 文魁資訊, 2008.
[25] R. M. Haralick and L. G. Shapiro, Computer and Robot Vision, Addison-Wesley Longman Publishing Co., Inc. Boston, MA, USA, vol. 1, 1992.
[26] 影像處理：Component Labeling (標號)
http://mermerism.blogspot.tw/2014/05/component-labeling.html
[27] C. Rougier, J. Meunier, A. St-Arnaud, and J. Rousseau, “Fall detection from human shape and motion history using video surveillance,” International Conference on Advanced Information Networking and Applications Workshops(AINAW), vol. 2, 2007, pp. 875 – 880.
[28] 王進德, “類神經網路與模糊控制理論入門與應用,” 全華出版社, 2011.
[29] D. E. Rumelhart and J. L. McLelland, Parallel distributed processing: explorations in the microstructure of cognition, MIT Press Cambridge, MA, USA, 1986.
[30] Y. Freund and R. E. Schapire, “A decision-theoretic generalization of on-line learning and an application to boosting,” Journal of computer and system sciences, vol. 5, pp. 119– 139, 1997.
[31] W. Iba and P. Langley, “Induction of one-level decision trees,” International Conference on Machine Learning, 1992, pp.233-240.
[32] E. Auvinet, C. Rougier, J. Meunier, A. St-Arnaud, and J. Rousseau, “Multiple cameras fall dataset,” Technical report 1350, DIRO - Université de Montréal, July 2010.