在本論文中，我們提出了一個以Kullback-Leibler importance estimation procedure~(KLIEP)為概念，新型的擁擠場景異常偵測方法。該方法一開始先將影片依時間和空間切割成小塊的立方塊(cube)，並且以三種特徵：梯度直方圖(histogram of oriented gradient,HOG)、光流方向(direction) 及光流大小(motion magnitude)，去表示每個不同的立方體，接下來再帶入KLIEP算出在不同特徵下，測試資料和訓練資料的權重，相較於以往之方法其需分別求出每個測試和訓練資料的機率分佈，這可以避免在處理資料時，去估測複雜之機率密度，尤其是針對高維度的資料。求出的權重代表著測試資料和訓練資料的相似程度，為一內圍分數，根據這個分數，我們就能判斷是否在各別小方塊中有無包含異常訊息。此外，為了降低訓練時間，我們也將KLIEP結合了一個新的集成學習(ensemble learning)，提出了新的異常偵測方法，能夠使用多組模型進行偵錯，不但降低訓練模型時的運行時間，也同時提高了準確度，最後針對誤判和雜訊，我們也提出了一個一致性後續處理來過濾和判別，以更進一步的提高精確度。

我們也將集成式KLIEP與一些公開的數據，透過電腦模擬作比較，以驗證所提出新方法的有效性。

This thesis proposes a novel inlier-based outlier detection scheme for anomaly detection in crowded scenes based on the Kullback-Leibler importance estimation procedure (KLIEP).

The video is first segmented into a set of local spatio-temporal cubes, from which three feature descriptors such as the histograms of oriented gradient (HOG), the motion direction descriptors, and the motion magnitude descriptors are then extracted. Thereafter, for each feature descriptor KLIEP is invoked, which computes the ratio of test and training densities, called the importance, instead of computing these two densities separately. This allows us to avoid the difficult density estimation, in particular for high dimensional data. The importance denotes an inlier score which represents the degree of similarity between the test and training data. Based on the importance, we can then identify if a local volume contains an anomaly event.

Moreover, to reduce the long training time, a new ensemble learning scheme is incorporated with KLIEP to substantially lower the computational burden and meanwhile achieve even superior detection accuracy. At last, in order to improve the accuracy, we also proposed a consistent anomaly detection to fix the false positives and noise.
Through computer simulations, the new ensemble KLIEP provides promising results compared with the main state-of-the-art methods based on several public datasets.

[1] O. Popoola and K. Wang, “Video-based abnormal human behavior recognition-A review,” IEEE Trans. Syst., Man, Cybern. C, Appl. Rev., vol. 42, no. 6, pp. 865–878, Nov. 2012.
[2] Z. Zivkovic, “Improved adaptive gausian mixture model for background subtraction,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 28–31, 2004.
[3] V. M. Kettnaker, “Time-dependent hmms for visual intrusion detection,” in Proc. Comput. Vis. Pattern Recognit. Workshop, pp. 34, June 2003.
[4] A. S. Voulodimos, D. I. Kosmopoulos, N. D. Doulamis, and T. A. Varvarigou, “A top-down event-driven approach for concurrent activity recognition,” Multimed. Tool Appl., 69:293– 311, 2014.
[5] A. Wiliem, V. Madasu, W. Boles, and P. Yarlagadda, “Detecting uncommon trajectories,” in Proc. Digit. Image Comput. Tech. Appl., pp. 398 – 404, Dec.2008.
[6] C. Stauffer and W. Grimson, “Learning patterns of activity using real-time tracking,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 22, pp. 747 – 757, Aug. 2000.
[7] R. Mehran, A. Oyama, and M. Shah. “Abnormal crowd behavior detection using social force model,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 935–942, 2009.
[8] D. Helbing, and P. Molna I, “Social force model for pedestrian dynamics,” Physical Review E, vol. 51, pp. 4282-4286, May 1995.
[9] Y. Benezeth, P. Jodoin, V. Saligrama,and C. Rosenberger, “Abnormal events detection based on spatio-temporal cooccurences,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 2458 – 2465, 2009.
[10] K.-W. Cheng, Y.-T. Chen, and W.-H. Fang, “Abnormal crowd behavior detection and localization using maximum sub-sequence search,” in Proc. ACM/ IEEE Int’l workshop on Anal. and retrieval of tracked events and motion in
imagery stream, pp.49-58, 2013.
[11] Y. Cong, J. Yuan, and J. Liu, “Sparse reconstruction cost for abnormal event detection,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 3449-3456, 2011.
[12] A. Chan, and N. Vasconcelos, “Modeling, clustering, and segmenting video with mixtures of dynamic textures,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 30, pp. 909-926, May 2008.
[13] W. Li, V. Mahadevan, and N. Vasconcelos, “Anomaly detection and localization in crowded scenes,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 36, pp. 18 – 32, June 2013.
[14] V. Mahadevan, W. Li, V. Bhalodia, and N. Vasconcelos, “Anomaly detection in crowded scenes,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit.,
pp. 1975–1981, 2010.
[15] V. Saligrama and Z. Chen. “Video anomaly detection based on local statistical aggregates,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 2112–2119, 2012.
[16] M. J. Rohstkhari and M. D. Levine, “Online dominant and anomalous behavior detection in videos,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 2611–2618, 2013.
[17] C. C. Loy, T. Xiang, and S. Gong, “Modelling multi-object activity by gaussian processes,” in Proc. Br. Mach. Vis. Conf., pp. 1-11, 2009.
[18] O. Boiman and M. Irani. “Detecting irregularities in images and in video,” in Proc. IEEE Int’l. Conf. Comput. Vis., pp. 17 – 31, 2007.
[19] X. Cui, Q. Liu, M. Gao, and D. N. Metaxas, “Abnormal detection using interaction energy potentials,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 3161 – 3167, 2011.
[20] K. Kim, D. Lee, and I. Essa, “Gaussian process regression flow for analysis of motion trajectories,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit.,
pp. 1164-1171, 2011.
[21] A. Adam, E. Rivlin, I. Shimshoni, and D. Reinitz, “Robust real-time unusual event detection using multiple fixed-location monitors,”IEEE Trans. Pattern Anal. Mach. Intell., vol. 30, pp. 555-560, Jan. 2008.
[22] Y.-T. Chen,W.-H. Fang, and C.-Y. Lee. “Abnormal detection in crowded scenes via kernel based direct density ratio estimation.” in Proc. IEEE China Signal Process. pp , 2015
[23] N. Dalal, and B. Triggs, “Histograms of oriented gradients for human detection,” in Proc. Int’l Conf. Comput. Vis. and Pattern Recognit., pp. 886-893, 2005.
[24] B. K. Horn and B. G. Schunck, “Determining optical flow,” Artificial Intell., 17:185–203, 1981.
[25] M. Sugiyama, T. Suzuki, S. Nakajima, H. Kashima and M. Kawanabe, “Direct importance estimation for covariate shift adaptation,” Ann. Inst. Stat. Math., 60:699 – 746, Dec. 2008.
[26] S. Hido, Y. Tsuboi, H. Kashima, M. Sugiyama, and T. Kanamori, “Inlierbased outlier detection via direct density ratio estimation,” in Proc. IEEE Int’l. Conf. on Data Mining, pp. 223 – 232, 2008.
[27] Y. Kawahara and M. Sugiyama, “Change-point detection in time-series data by direct density-ratio estimation,” in Proc. SIAM Int’l Conf. on Data Mining, pp. 389-400, 2009.
[28] T. Hastie, R. Tibshirani, and J. Friedman, The Elements of Statistical Learning: Data Mining, Inference, and Prediction. Springer, 2009.
[29] L. Breiman, “Random forests,” Machine Learning, 45:5–32, 2001.
[30] P. Viola and M. J. Jones, “Robust real-time face detection,” Int’l J. Comput. Vis., 57:137–154, 2004.
[31] J. Shotton, A. Fitzgibbon, M. Cook, T. Sharp, M. Finocchio, R. Moore, A. Kipman, and A. Blake, “Real-time human pose recognition in parts from single depth images,” in Proc. Conf. Comput. Vis. Pattern Recognit., pp.1297-1304, 2011.
[32] A. Zimek, R. J. Campello, and J. Sander, “Ensembles for unsupervised outlier detection: challenges and research questions [position paper],” SIGKDD Explor., 15:11–22, June 2013.
[33] A. Lazarevic and V. Kumar, “Feature bagging for outlier detection,” in Proc. ACM SIGKDD Int’l. Conf. on Knowledge Discovery in Data mining, pp. 157– 166, 2005.
[34] S. Boyd and L. Vandenberghe, Convex Optimization. Cambridge Cambridge University Press, 2004.
[35] A. Zaharescu and R. Wildes,“Anomalous behaviour detection using spatiotemporal oriented energies, subset inclusion histogram comparison and event-driven processing,” in Proc. Eur. Conf. Comput. Vis., pp. 563–576, 2010.
[36] S. J. Blunsden and R. B. Fisher. “The behave video dataset: ground truthed video for multi-person behavior classification,” Annals of the BMVA, pp.1-11,2009.
[37] C. Lu, J. Shi, and J. Jia, “Abnormal event detection at 150 fps in matlab.” in Proc. Intl. Conf. Comput. Vis., pp. 2720-2727, 2013.
[38] J. Kim and K. Grauman, “Observe locally, infer globally: A space-time mrf for detecting abnormal activities with incremental updates,” in Proc. Conf.
Comput. Vis. Pattern Recognit., pp. 2921 – 2928, 2009.
[39] Y. Cong, J. Yuan, and J. Liu, “Abnormal event detection in crowded scenes using sparse representation,” in Proc. Conf. Comput. Vis. Pattern Recognit., pp. 1851–1864, 2013.
[40] Y. Lee and O. Mangasarian, “Rsvm: Reduced support vector machines,” in Proc. SIAM Int’l. Conf. on Data Mining, 2001.
[41] M. E. Tipping, “Sparse bayesian learning and the relevance vector machine,” Journal of Machine Learning Research, pp. 211–244, 2001.