在監控影片中異常運動物體偵測是一個重大議題，透過電腦視覺分析視頻中的行人進而判斷是否異常。同時，能夠有效率的瀏覽監控影片也是一個重要的議題，透過影片濃縮縮短監控影片的時間卻保留其資訊。於是我們將兩者結合為一，設計一套異常偵測並影片濃縮系統，自動檢測影片中的異常事件與物體，並將其依照異常程度產生濃縮影片。
本論文提出一個可即時自動偵測出異常運動物體並持續學習異常模型的方法。傳統的異常偵測方法是透過事先定義異常事件規則來判斷，但真實世界的情況較為複雜，異常行為的定義會隨著時間而變化。我們則定義異常事件為較少出現的事件類型，並透過即時學習不斷更新異常事件模型。首先偵測場景上的運動物體，我們利用高斯混合模型偵測前景，再利用邊界提取(Border Proposal)演算法檢測出運動物體。再來使用核心化相關濾波器(Kernelized Correlation Filters)追蹤演算法對這些運動物體進行追蹤，最後將物件軌跡轉為區塊軌跡以便異常偵測。
在異常運動物體偵測的程序中，將收集到的軌跡資訊，利用機率樹進行學習，學習出軌跡模組，再由軌跡模型中的出現機率為依據，判斷運動物體是否為異常。針對追蹤到的物件，我們將存下影片的資訊、軌跡以及異常程度。透過這些資訊，將物件依照異常程度進行排序，並考慮物件的軌跡避免相互碰撞，以產生一支異常濃縮影片。
在實驗的部份，我們針對異常偵測進行兩種環境的分析，分別是學校廣場與學校大門。我們提出的方法可以正確偵測到異常行為的物體，平均的準確性是88.8%。在影片濃縮的實驗中，我們的方法可以產生出避免物體碰撞、異常物體在先的濃縮監控影片。

Abnormal moving objects detection is an essential issue for video surveillance. In order to judge whether the behavior of objects is abnormal. On the other hand, monitoring surveillance videos efficiently is also an important issue. The video condensation reduces the time for monitoring the video and retains its information. Therefore, we combined the two into one, designed an abnormal detection and video condensation system, automatically detected the abnormal events and objects in the video, and generated a condensation video according to the degree of abnormality.
This thesis proposes a method for automatically detecting abnormal moving objects and continuously learning abnormal models. Traditional abnormal moving objects detection aims at particular circumstances or requirement to predefine particular detection rules which the application of abnormal moving objects detection is restricted, but the real world situation is complicated, and the definition of abnormal behavior changes with time. We define abnormal events as less frequent event types and constantly update the abnormal model through learning. First, we detect moving objects on the scene. We use the Gaussian mixture model to detect the foreground, and then use the Border Proposal algorithm to detect moving objects. Then use the KCF algorithm to track these moving objects, and finally transform the object trajectory into a block trajectory for abnormal detection.
In the abnormal moving object detection, the collected trajectory information is learned by the pattern tree, and find out the pattern of trajectories. The probability of occurrence in the trajectory model is used to determine whether the moving object is abnormal.
For these tracked objects, we will save its information, trajectories and abnormal rate of the video. Through this information, the objects are sorted according to the abnormal rate, and consider the trajectory of the object to avoid collision. Finally, the system will produce an abnormal condensation video.
The experiment monitors and analyzes different environment, such as campus and school gate. The method can detect abnormal objects correctly, and the average accuracy is 88.8%. In the experiment of video condensation, our method can produce a condensation surveillance video that avoids object collisions and abnormal objects.

[1] Y. Wang, X. Zhang, M. Li, P. Jiang, and F. Wang, “A GM-HMM based abnormal pedestrian behavior detection method,” in Proceedings of the IEEE International Conference on Signal Processing, Communications and Computing, Ningbo, China, pp. 1-6, Sep 2015.
[2] H. Rajabi and M. Nahvi, “An intelligent video surveillance system for fall and anesthesia detection for elderly and patients,” in Proceedings of the International Conference on Pattern Recognition and Image Analysis, Rasht, Iran, pp. 1-6, Mar 2015.
[3] J. C. SanMiguel, J. M. Martínez, and L. Caro-Campos, “Object-size invariant anomaly detection in video-surveillance,” in Proceedings of the International Carnahan Conference on Security Technology, Madrid, Spain, pp. 1-6, Oct 2017.
[4] K. V. Beeck, K. V. Engeland, J. Vennekens, and T. Goedemé, “Abnormal behavior detection in LWIR surveillance of railway platforms,” in Proceedings of the IEEE International Conference on Advanced Video and Signal Based Surveillance, Lecce, Italy, pp. 1-6, Sep 2017.
[5] D. V. Ngo, N. T. Do, and L. A. T. Nguyen, “Anomaly detection in video surveillance: A novel approach based on sub-trajectory,” in Proceedings of the International Conference on Electronics, Information, and Communications, Da Nang, Vietnam, pp. 1-4, Jan 2016.
[6] H. Mousavi, M. Nabi, H. Kiani, A. Perina, and V. Murino, “Crowd motion monitoring using tracklet-based commotion measure,” in Proceedings of the IEEE International Conference on Image Processing, Quebec City, QC, Canada, pp. 2354-2358, Sep 2015.
[7] T. Zhang, S. Liu, C. Xu, and H. Lu, “Mining semantic context information for intelligent video surveillance of traffic scenes,” IEEE Transactions on Industrial Informatics, vol. 9, no. 1, pp. 149-160, Feb 2013.
[8] M. Afonso and J. Nascimento, “Predictive multiple motion fields for trajectory completion: Application to surveillance systems,” in Proceedings of the IEEE International Conference on Image Processing, Quebec City, QC, Canada, pp. 2547-2551, Sep 2015.
[9] C. Shen, R. Xie, and L. Zhang, "Temporal recursive pedestrian trajectory identification for multi-camera surveillance system," in Proceedings of the IEEE International Conference on Computer and Information Technology; Ubiquitous Computing and Communications; Dependable, Autonomic and Secure Computing; Pervasive Intelligence and Computing, Liverpool, UK, pp. 1158-1163, Oct 2015.
[10] C. W. Chang, M. H. Yang, C. C. Li, and K. T. Chuang, “Toward mining anomalous behavior from big moving trajectories in surveillance video,” in Proceedings of the IEEE International Conference on Automation Science and Engineering, Taipei, Taiwan, pp. 1121-1126, Aug 2014.
[11] X. Zang, G. Li, Z. Li, N. Li, and W. Wang, “An object-aware anomaly detection and localization in surveillance videos,” in Proceedings of the IEEE Second International Conference on Multimedia Big Data, Taipei, Taiwan, pp. 113-116, Apr 2016.
[12] M. Y. Choong, R. K. Y. Chin, K. B. Yeo, and K. T. K. Teo, “Trajectory clustering for behavioral pattern learning in transportation surveillance,” in Proceedings of the International Conference on Artificial Intelligence with Applications in Engineering and Technology, Kota Kinabalu, Malaysia, pp. 199-203, Dec 2014.
[13] R. Ranjith, J. Joshan Athanesious, and V. Vaidehi, “Anomaly detection using DBSCAN clustering technique for traffic video surveillance,” in Proceedings of the Advanced Computing, Chennai, India, pp. 1-6, Dec 2015.
[14] B. Cancela, A. Iglesias, M. Ortega, and M.G. Penedo, “Unsupervised trajectory modelling using temporal information via minimal paths,” in Proceedings of the Computer Vision and Pattern Recognition, Columbus, Ohio, pp. 2553-2560, Jun 2014.
[15] H. Tan, Y. Zhai, Y. Liu, and M. Zhang, “Fast anomaly detection in traffic surveillance video based on robust sparse optical flow,” in Proceedings of the Acoustics, Speech and Signal Processing, Shanghai, China, pp. 1976-1980, Mar 2016.
[16] C. Wang, X. Wu, N. Li, and Y. L. Chen, “Abnormal detection based on gait analysis,” in Proceedings of the 10th Intelligent Control and Automation, Beijing, China, pp. 4859-4864, Jul 2012.
[17] S. Coşar, G. Donatiello, V. Bogorny, C. Garate, L. O. Alvares, and F. Brémond, “Toward abnormal trajectory and event detection in video surveillance,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 27, no. 3, pp. 683-695, Mar 2017.
[18] W. Li, V. Mahadevan, and N. Vasconcelos, “Anomaly detection and localization in crowded scenes,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 36, no. 1, pp. 18-32, Jun 2013.
[19] Y. Cai, H. Wang, X. Chen, and H. Jiang, “Trajectory-based anomalous behavior detection for intelligent traffic surveillance,” IET Intelligent Transport Systems, vol. 9, no. 8, pp. 810-816, Sep 2015.
[20] C. L. Chou, C. H. Lin, T. H. Chiang, H. T. Chen, and S. Y. Lee, “Coherent event-based surveillance video synopsis using trajectory clustering,” in Proceedings of the IEEE International Conference on Multimedia and Expo Workshops, Turin, Italy, pp. 1-6, Jul 2015.
[21] W. C. Wang, P. C. Chung, C. R. Huang, and W. Y. Huang, “Event based surveillance video synopsis using trajectory kinematics descriptors,” in Proceedings of the Fifteenth IAPR International Conference on Machine Vision Applications, Nagoya, Japan, pp. 250-253, May 2017.
[22] W. Liao, Z. Tu, S. Wang, Y. Li, R. Zhong, and H. Zhong, “Compressed-domain Video Synopsis via 3D Graph Cut and Blank Frame Deletion,” in Proceedings of the on Thematic Workshops of ACM Multimedia, Mountain View, California, pp. 253-261, Oct 2017.
[23] C. Stauffer and W.E.L. Grimson, “Adaptive background mixture models for real-time tracking,” in Proceedings of the Computer Vision and Pattern Recognition, Fort Collins, Colorado, pp. 246-252, Jun 1999.
[24] S. Suzuki and K. Abe “Topological structural analysis of digitized binary images by border following,” Computer Vision, Graphics, and Image Processing, vol. 30, no. 1, pp. 32-46, Apr 1985.
[25] J. F. Henriques, R. Caseiro, P. Martins, and J. Batista, “High-Speed Tracking with Kernelized Correlation Filters,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 37, no. 3, pp. 583-596, Aug 2014.