臺灣博碩士論文加值系統

大型群眾集會常因為管理失當，造成人群推擠而發生不幸意外，以視覺為基礎的視訊監視系統，由於監視範圍廣及安裝便利，常用來監視公共場所的群眾活動，掌握群眾動態，以維護公共安全，若能即時偵測異常群眾事件的發生，採取因應之道，可防止意外發生或避免傷害擴大。基於異常群眾事件發生時，人群慌忙躲避，群眾分佈瞬間出現明顯的變化，由於兩相鄰影像間之間隔極短，導致影像間群眾特徵的分佈呈現異常的變化，因此本論文將群眾分佈出現大幅變化時視為異常群眾事件的發生，以時間數列表示特徵值的變化與建立網格模型描述群眾分佈來加強偵測異常群眾事件的能力，來提升監視攝影機的附加價值。

Large crowd gathering always has the potentials of creating the unfortunate incidents due to crowd pushing and mismanagement, crowd management research literature shows that most of the major incidents can be prevented or minimized by a proper management strategy. If the crowd abnormal events can be detected early and the governing agency can take appropriate actions to mitigate, accidental injury can be prevented or the incident can be contained. This thesis presents time series technical approach to gather the required crowd information using fixed cameras to collect visual data and a grid model to describe the crowd distribution so the abnormal crowd patterns can be identified promptly. Crowd dominant motions and spatial outlier detection using spatial autocorrelation. This proposed method supports valuable spatial information for space planning and management, which is applicable to public spaces for safety surveillance and the layout planning of exhibitions, while the added value of surveillance cameras can also be improved.

目錄
致謝 i
中文摘要 iii
英文摘要 iv
目錄 v
圖目錄 vii
表目錄 ix
第1章 研究簡介 1
1.1概述 2
1.2研究動機 6
1.3論文結構 11
第2章 文獻探討 12
2.1應用偏差法之相關機制 13
2.1.1拓樸法 15
2.1.2光流量法 19
2.2應用偵測法之相關機制 22
第3章 以時間數列偵測異常群眾事件之研究方法 26
3.1以時間數列偵測異常群眾事件 27
第4章 實驗模擬結果 48
第5章 結論與未來展望 66
參考文獻 69
附錄 A 74
PUBLICATION LIST OF YU-CHENG WANG 79














圖目錄
圖1-1、異常群眾事件之監視影像偵測系統 9
圖1-2、數位視訊錄影機架構 12
圖1-3、網路視訊錄影機架構 13
圖1-4、德明財經科技大學-資訊學院之監控影像系統 14
圖1-5、上海跨年活動意外事件[18] 15
圖1-6、歐洲足球冠軍運動賽事意外事件[19] 15
圖1-7、大甲媽祖衝突事件[20] 16
圖1-8、監控中心[21] 17
圖1-9、影像分析應用其他相關領域 18

圖2-1、正常群眾行為的模型之一[22] 21
圖2-2、正常群眾行為的模型之二[24] 21
圖2-3、偏差法相關機制之流程圖 22
圖2-4、稀疏粒子運動場示意圖之一[22] 23
圖2-5、稀疏粒子運動場示意圖之二[22] 24
圖2-6、群眾異常事件發生的擷取影像之一[22] 25
圖2-7、群眾異常事件發生的擷取影像之二[22] 25
圖2-8、經過轉換後的清晰拓樸之一[22] 26
圖2-9、經過轉換後的清晰拓樸之二[22] 26
圖2-10、不同群眾分佈場景[25] 28
圖2-11、不同群眾分佈場景的粒子運動場[25] 29
圖2-12、不同群眾分佈濾除雜訊之後的偵測影像[25] 29
圖2-13、稀疏群眾之紋理呈現粗糙 30
圖2-14、密集群眾之紋理呈現平滑 31
圖2-15、3X3灰階點 31
圖2-16、描述紋理的特徵值 32

圖3-1、網格模型示意圖 34
圖3-2、以時間數列偵測異常群眾事件之流程圖 36
圖3-3、RGB彩色影像 37
圖3-4、灰階影像 37
圖3-5、無人群之參考背景 38
圖3-6、群眾移動之影像 38
圖3-7、群眾移動之影像 39
圖3-8、群眾移動之二值化影像 40
圖3-9、更新參考影像示意圖 41
圖3-10、網格狀態值示意圖 42
圖3-11、不同影像像框號碼區域主動像素示意圖 43
圖3-12、狀態值產生之過程 44
圖3-13、描述群眾分佈之網格模型 46
圖3-14、離散謝比雪夫衝量值[38] 46
圖3-15、左圖為正常群眾的分佈，右圖為異常事故發生時，群眾快速離開現場 48
圖3-16、經謝比雪夫衝量簡化過後的特徵值 48
圖3-17、衝量變化之時間數列示意圖 49
圖3-18、BOX-JENKINS MODEL流程圖 50

圖4-1、實驗模擬程式平台 56
圖4-2、AUTOBOX套用商用軟體平台 57
圖4-3、正常群眾分佈 57
圖4-4、異常事件發生時 58
圖4-5、測試視訊影像間謝比雪夫衝量差值的變化 58
圖4-6、滋事分子持木棍毆打及砸車 60
圖4-7、群眾滋事打架影像間衝量差值的變化 61
圖4-8、光線充足的室外場地 65
圖4-9、光線充足的室外場地(NTH = 20) 65
圖4-10、光線充足有群眾行為室外場地 66
圖4-11、光線充足有群眾行為室外場地(NTH = 20) 66
圖4-12、光線充足有異常群眾行為的室外場地 67
圖4-13、光線充足有群眾行為室外場地(NTH = 80) 68
圖4-14、下雨的夜間室外場地(NTH = 20) 69
圖4-15、下雨的夜間室外場地(NTH = 40) 70
圖4-16、群眾異常行為下雨的夜間室外場地(NTH = 20) 71
圖4-17、群眾異常行為下雨的夜間室外場地(NTH = 40) 72
圖4-18、群眾異常行為下雨的夜間室外場地(NTH = 80) 73

圖A-1、自我迴歸模式AR的自我相關函數ACF呈衰減性(DIE OFF) 85
圖A-2、其偏自我相關函數PACF呈截尾形式(CUT OFF)，僅有少數項係數 85

表目錄
表1-1、三類監視影像系統的比較 [13] 11

表4-1、使用ARMA模型偵測離群值種類及原因判讀(由AUTOBOX 軟體計算) 59
表4-2、使用AMIDAN所提方法偵測離群值種類及原因判讀 59
表4-3、群眾滋事打架之ARMA模型偵測離群值種類及原因判讀 61
表4-4、群眾滋事打架之AMIDAN所提方法偵測離群值種類及原因判讀 62
表4-5、ROC曲線下方的面積之比較 63
表4-6、光線充足的室外場地門檻值比較 67

表A-1、各模型之ACF與PACF分佈 86

參考文獻
[1]W. R. Dornberger, “V-2,” Viking Press, First Edition, Apr. 1954.
[2]V. Romesh, “Distance Education in Technological Age,” Anmol Publications Pvt. Ltd., First Edition, Jan. 2005.
[3]D. P. B. Smith, “Magtape 1,” Wikimedia Commons, Jan. 2005.
[4]G. Maas, “AMPEX 2 Zoll,” Wikimedia Commons, Sep. 2007.
[5]H. Kruegle, “CCTV Surveillance: Video Practices and Technology,” Elsevier Butterworth-Heinemann, Second Edition, May 2007.
[6]D. Ericd, “CD-R,” Wikimedia Commons, Jan. 2012.
[7]F. Alt, “Verbatim DVD-R Spindle,” Wikimedia Commons, Jul. 2005.
[8]P. Dixon, “Surveillance in America: An Encyclopedia of History, Politics, and the Law,” World Privacy Forum, First Edition, Feb. 2016.
[9]G. C. Robb, “Police Use of CCTV (Closed Circuit Television) Surveillance
Constitutional Implications and Proposed Regulations,” University of Michigan
Journal of Law Reform, vol. 13, no. 3, pp. 571-602, May 1980.
[10]B. Yesil, “Watching Ourselves,” Cultural Studies, pp. 400-416, Aug. 2006.
[11]BBC website. Available: http://news.bbc.co.uk/2/hi/americas/1865828.stm, Mar. 2002.
[12]C. Norris, M. M. Cahill and D. Wood, “Editorial. The Growth of CCTV: a global perspective on the international diffusion of video surveillance in publicly accessible space,” Surveillance & Society, pp. 110-135, Apr. 2004.
[13]林啟豐、劉邦俊、段仁豪、邱敏煌和葉鳴紳, “數位影像平台於監控系統之整合應用,” 中華技術期刊, 第2卷, 頁數 28-41, 7月 2011.
[14]17 Media website. Available: http://17.media/, Jul. 2015.
[15]Pokémon Go website. Available: http://www.pokemongo.com/, Jul. 2016.
[16]Taiwan From Above website. Available: https://www.facebook.com/Abovetai/, Sep. 2013.
[17]iThome website. Available: http://www.ithome.com.tw/node/76084, Sep. 2012.
[18] Appledaily website. http://www.appledaily.com.tw, Jan. 2015.
[19] Reuters website. http://uk.reuters.com/, Jun. 2017.
[20] Setn website. http://www.setn.com/, Mar. 2017.
[21] Taipei Police website. http://police.gov.taipei/, Apr. 2015.
[22]N. Li and Z. Zhang, “Abnormal Crowd Behavior Detection Using Topological Methods,” Proc. of International Conference on Software Engineering Artificial intelligent, Networking and Parallel/Distributed Computing, pp. 13-18, Jul. 2011.
[23]R. Mehran, A. Oyama, and M. Shah, “Abnormal Crowd Behavior Detection Using Social Force Model,” Proc. of International Conference on Computer Vision and Pattern Recognition, pp. 935-942, Jun. 2009.
[24]C. Bian, D. Chen and S. Wang, “Velocity Field Based Modeling and Simulation of Crowd in Confrontation Operations,” Proc. of International Conference on Parallel and Distributed Systems, pp. 646-651, Dec. 2010.
[25]S. Wu, B. E. Moore and M. Shah, “Chaotic Invariants of Lagrangian Particle Trajectories for Anomaly Detection in Crowded Scene,” Proc. of International Conference on Computer Vision and Pattern Recognition, pp. 2054-2060, Jun. 2010.
[26]Y. Wen, J. Du and J. M. Lee, “Abnormal Event Detection Based on Social Force Model Combined with Crowd Violent Flow,” Proc. of International Conference on Cloud Computing and Intelligence Systems, pp. 440-446, Aug. 2016.
[27]R. M. Haralick, K. Shanmugam and I. Dinstein, “Textural Features for Image Classification,” Proc. of International Conference on Systems, Man and Cybernetics, pp. 610-621, Nov. 1973.
[28]Y. Zhou, S. Yan and T. Huang, “Detecting Anomaly in Videos from Trajectory Similarity Analysis,” Proc. of International Conference on Multimedia and Expo, pp. 1087-1090, Jul. 2007.
[29]T. Li, H. Chang, M. Wang, B. Ni, R. Hong and S. Yan, “Crowd Scene Analysis: a Survey,” IEEE Transactions on Circuits and System for Video Technology, vol. 25, no. 3, pp. 367-386, Sep. 2014.
[30]C. Bian, D. Chen and S. Wang, “Velocity Field Based Modeling and Simulation of Crowd in Confrontation Operations,” Proc. of International Conference on Parallel and Distributed Systems, pp. 646-651, Dec. 2010.
[31]D. Y. Chen and P. C. Huang, “Dynamic Hman Cowd Mdeling and Its Application to Anomalous Events Detection,” Proc. of International Conference on Multimedia and Expo, pp. 1582-1587, Jul. 2010.
[32]B. G. Amidan, T. A. Ferryman and S. K. Cooley, “Data Outlier Detection Using the Chebyshev Theorem,” Proc. of International Conference on Aerospace, pp.3814-3819, Mar. 2005.
[33] W. W. S. Wei, “Time Series Analysis : Univariate and Multivariate Methods,”
Addison Wesley, Second Edition, Jan. 2006.
[34]W. L. Hsu, C. C. Hsiao, T. L. Chang and T. L. Chen, “Vision-based Monitoring Method Using Gray Relational Analysis,” IET Computer Vision, Vol. 3, pp.103-111, Sep. 2009.
[35]M. K Hu, “Visual Pattern Recognition by Moment Invariants,” IRE Transaction on Information Theory, Vol. 8, pp.179-187, Feb. 1962.
[36]W. Kim and C. Jung, “Illumination-Invariant Background Subtraction: Comparative Review, Models, and Prospects,” IEEE Access, Vol. 5, pp.8369-8384, Apr. 2017.
[37]W. L. Hsu, Y. C. Wang and C. L. Lin, “Abnormal Crowd Event Detection Based on Outlier in Time Series,” Proc. of International Conference on Machine Learning and Cybernetics, pp.359-363, Jul. 2014.
[38]R. Mukundan, S. H. Ong and P. A. Lee, “Image Analysis by Tchebichef Moments,” IEEE Transactions on Image Processing, Vol. 10, pp.1357-1364, Sep. 2001.
[39]University of Minnesota website. http://mha.cs.umn.edu/movies/, Dec. 2016.
[40]I. Chang, G. C. Tiao and C. Chen, “Estimation of Time Series Parameters in the Presence of Outliers,” Technometrics, Vol. 30, pp.193-204, May 1988.
[41]R. S. Tsay. “Outlier, Level Shifts and Variance Changes in Time Series,” Journal of Forecasting, Vol. 7, pp.1-20, Jan. 1988.
[42]Autobox website. http://www.autobox.com/cms/index.php/downloads, Feb. 2017.
[43]C. S. Tsai, Y. C. Wang and W. L. Hsu, “Crowd Abnormal Event Detection by Using Time Series Analysis,” Journal of Computers, Mar. 2017. (Accept)
[44]W. L. Hsu, Y. C. Wang and K. F. Lin, “The Path of Crowd Dominated-Motion Detection Based on Spatial Autocorrelation,” Proc. of International Conference on Machine Learning and Cybernetics, pp. 715-719, Jul. 2016.
[45]C++ Builder website. https://www.embarcadero.com/products/cbuilder, Feb. 2016.
[46]W. L. Hsu, C. C. Hsiao, Y. L. Chang and T. L. Chen, “Vision-based Monitoring Method Using Gray Relational Analysis,” IET Computer Vision, Vol. 3, pp.103-111, Sep. 2009.
[47]V. M. Llic, M. Stankovic and E. H. Mulalic, “Comments on “Generalization of Shannon–Khinchin Axioms to Nonextensive Systems and the Uniqueness Theorem for the Nonextensive Entropy”,” IEEE Transactions on Information Theory, Vol. 59, pp. 6950-6952, May 2013.
[48]IBM SPSS website. https://www.ibm.com/analytics/us/en/, Jul. 2017.
[49]H. Narasimhan and S. Agarwal, “Support Vector Algorithms for Optimizing the Partial Area under the ROC Curve,” Neural Computation, Vol. 29, pp. 1919-1963, Jun. 2017.
[50]Alex's Work website. http://alex-phd.blogspot.tw/2014/03/optical-flowopencv.html, Jul. 2017.