臺灣博碩士論文加值系統

近幾年來，卷積類神經網路已被廣泛地應用於很多影像資訊科技領域，例如影像處理暨分析、影像辨識、虛擬與增強實境、智慧型視訊監控系統等應用。為了保障生命、財產的安全，在許多開放式場所或隱密巷子皆會安裝監控攝影機來嚇阻與攝錄犯罪行為，然而一般安控中心通常負責很多的監控畫面，若是透過人員來監看，容易造成安控人員視覺上的疲勞而無法長時間專注於監控畫面，也容易耗費大量的人力資源，若能自動偵測監控畫面出現的異常事件(例如: 搶劫、鬥毆、逗留、追逐) 並即時通報給執法人員以儘速進行處置，也可減少人力成本，並可有效降低犯罪率全。目前大多數視訊監控系統的異常偵測主要是以人群突然或快速聚集或疏散為目標事件，對於毆打、搶劫、攻擊、逗留、追逐等非正常行為的偵測則較為少見。本文為了有效偵測上述異常事件，提出了一種基於時空域的特徵神經網路之異常事件偵測方法，主要是先對攝取視訊進行影像縮減取樣以提高後續處理的執行速度，再對已處理後的影像序列提取前景移動區域; 接著將前述移動區域輸入至異常事件偵測模組以進行判斷是否為異常，該異常事件偵測模組是由時空域的特徵神經網路所建構，藉由時空域的特徵資訊來做異常與正常分類以輸出判斷結果，該分類處理是透過閾值與誤差值來決定正常與異常的類別。實驗測試是採用公開數據集與自拍數據集以進行評估，實驗結果顯示本方法對於異常事件偵測比其他深度學習方法有較高的偵測率與大約對等的偵測時間。

In recent years, convolutional neural network (CNN) has been used in many image information technology fields, such as image processing and analysis, image recognition, virtual reality & augmented reality, and intelligent video monitoring system. For ensuring safety of both life and property, the surveillance cameras are installed in many important open spaces or alleys in order to deter and record criminal behavior. In general, there are many surveillance screens to be monitored in the security control center and this usually causes the visual fatigue so as to be unable to focus on the monitoring screens for a long time. In addition, it is also to consume a lot of human resources. Currently, the abnormality detection in most video surveillance systems is mainly based on the sudden or rapid gathering or fleeing of people, but the detection of activities, such as beatings, robberies, attacks, stays, and chasing, is relatively rare. To effectively detect those abnormal activities, this thesis proposes abnormal events detection method based on the spatiotemporal feature neural networks. In the proposed method, the captured image sequence is first down-sampled in order to speed-up the later processing. Then, the moving areas are extracted and inputted into the abnormal event detection module to determine whether it is abnormal, where the abnormal event detection module is constructed by the feature neural network in the space-time domain. The spatiotemporal feature information of those moving areas is classified for determining abnormal or normal event. The classification can generate the categories of normal and abnormal through threshold and error values. By using both public and self-capturing data set for evaluation, the experimental results show that the proposed method has higher detection rate and the approximate detection time than other deep-learning approach of using CNN scheme.

摘要 III
ABSTRACT III
致謝 V
目錄 III
圖目錄 VIX
表目錄 IX
第一章 緒論 1
1.1 研究背景 1
1.2 本系統簡介與架構 2
1.3 論文大綱 3
第二章 習知處理技術與知識 4
2.1 縮減取樣 4
2.1.1 最近相鄰內插法 4
2.1.2 雙線性內插法 5
2.1.3 雙立方內插法 7
2.2 畫面差異性 8
2.2.1 背景相減法 8
2.2.2 影像相鄰相減法 9
2.3 光流法 9
2.3.1 稠密光流 11
2.3.2 稀疏光流 11
2.4 二值化圖像 13
2.4.1 固定式二值化 13
2.4.2 迭代法二值化 14
2.5 形態學 15
2.5.1 膨脹 15
2.5.2 侵蝕 16
2.5.3 斷開、閉合 17
2.6 類神經網路(Neural Network) 18
2.6.1 卷積層(convolutional layer) 18
2.6.2 遞迴神經網路(Recurrent Neural Networks, RNN) 22
2.6.3 長短期記憶網路(Long Short-Term Memory, LSTM) 24
2.7 激勵函數(Activation Function) 27
2.7.1 雙彎曲函數(Sigmoid) 28
2.7.2 雙曲正切函數(Tanh) 29
2.7.3 線性整流函數(Rectified Linear Unit, ReLU) 29
2.7.4 Softmax函數 30
第三章 相關方法探討 31
3.1 Abnormal events detection based on trajectory clustering 32
3.2 Global abnormal events detection in crowded scenes using context location and motionrich spatio-temporal volumes 35
3.3 Abnormal Event Detection in Indoor Video using Feature Coding 41
3.4 Abnormal Event Detection Using Convolutional Neural Networks and 1-Class SVM classifier 46
3.5 Learning Temporal Regularity in Video Sequences 50
3.6 Sparse Coding Guided Spatiotemporal Feature Learning for Abnormal Event Detection in Large Videos 54
3.7 Abnormal Event Detection in Videos using Spatiotemporal Autoencoder 57
第四章 系統流程與方法 60
4.1 影像預處理 63
4.1.1 縮減取樣 64
4.1.2 前景目標切割 65
4.1.3 光流法預處理 66
4.2 異常偵測模組 68
4.2.1 自動編碼 70
4.2.2 空間域特徵向量 72
4.2.3 特徵向量轉置卷積 75
4.2.4 時間域特徵向量 76
4.2.5 空間域特徵與時間域特徵向量融合 80
4.2.6 數據集 81
4.2.7 重建誤差 84
4.2.8 閾值 85
4.2.9 訓練 86
4.2.10 計算事件次數 87
第五章 實驗結果 88
5.1 實驗環境與設備 88
5.2 實驗結果 90
5.3 實驗結果與分析 98
第六章 結論與未來方向 103
6.1 結論 103
6.2 未來方向 104
參考文獻 105

[1] R. Keys, "Cubic convolution interpolation for digital image processing," vol. 29, no. 6, pp. 1153-1160, Dec 1981.
[2] D. Meyer ; J. Denzler ; H. Niemann, "Model Model based Extraction of Articulated Objects in Image Sequences for Gait Analysis," in Proceedings of International Conference on Image Processing, ICIP, Santa Barbara, CA, USA, pp. 78-81, Oct 1997.
[3] D. Gutchess ; M. Trajkovics ; E. Cohen-Solal ; D. Lyons ; A.K. Jain, "A background model initialization algorithm for video surveillance," in Proceedings Eighth IEEE International Conference on Computer Vision. ICCV 2001, Vancouver, BC, Canada, Canada, pp. 733-740, Jul 2001.
[4] Bruce D.Lucas, Takeo Kanad, "An iterative image registration technique with an application to stereo vision," in IJCAI'81 Proceedings of the 7th international joint conference on Artificial intelligence, Vancouver, BC, Canada, pp. 674-679, Aug 1981.
[5] Geoffrey E. Hinton ; Simon Osindero ; Yee-Whye Teh, "A Fast Learning Algorithm for Deep Belief Nets," Neural Computation, vol. 18, no. 7, pp. 1527-1554, Jul 2006.
[6] Ilya Sutskever ; Geoffrey E. Hinton, "Deep, Narrow Sigmoid Belief Networks Are Universal Approximators," Neural Computation, vol. 20, no. 11, pp. 2629 - 2636, Nov 2008.
[7] K. Fukishima, "Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position," Biological Cybernetics, vol. 36, no. 4, pp. 193-202, Apr 1980.
[8] Y. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard and L. D. Jackel, "Backpropagation Applied to Handwritten Zip Code Recognition," Neural Computation, vol. 1, no. 4, pp. 541-551, Dec 1989.
[9] A. Krizhevsky, I. Sutskever and G. E. Hinton, "ImageNet Classification with Deep Convolutional Neural Networks," in Proceedings of the 25th International Conference on Neural Information Processing Systems, Lake Tahoe, Nevada, pp.1097-1105, Dec 2012.
[10] Tomáš Mikolov ; Stefan Kombrink ; Lukáš Burget ; Jan Černocký ; Sanjeev Khudanpur, "Extensions of recurrent neural network language model," in 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), Prague, Czech Republic, pp.5528-5531, Jul 2011.
[11] Tomáš Mikolov ; Anoop Deoras ; Daniel Povey ; Lukáš Burget ; Jan Černocký, "Strategies for training large scale neural network language models," in 2011 IEEE Workshop on Automatic Speech Recognition & Understanding, Waikoloa, HI, USA, pp.196-201, Dec 2011.
[12] V. Mnih, N. Heess, A. Graves and K. Kavukcuoglu, "Recurrent Models of Visual Attention," in Proceedings of the 27th International Conference on Neural Information Processing Systems, MIT Press Cambridge, MA, USA, pp.1057-1063, Jun 2014.
[13] Sepp Hochreiter ; Jürgen Schmidhuber, "Long Short-Term Memory," Neural Computation, vol. 9, no. 8, pp. 1735-1780, Nov 1997.
[14] Najla Bouarada Ghrab ; Emna Fendri ; Mohamed Hammami, "Abnormal events detection based on trajectory clustering," in 2016 13th International Conference on Computer Graphics, Imaging and Visualization (CGiV), Beni Mellal, Morocco, pp.301-306, Apr 2016.
[15] N. Patil ; Prabir Kumar Biswas, "Global abnormal events detection in crowded scenes using context location and motion-rich spatio-temporal volumes," IET Image Processing, vol. 12, no. 4, pp. 596-604, Mar 2018.
[16] Mona Izadi ; Zohreh Azimifar ; Gholam-Hossein Jowkar, "Abnormal Event Detection in Indoor Video using Feature Coding," in 2017 Artificial Intelligence and Signal Processing Conference (AISP), Shiraz, Iran, pp.151-155, Oct 2017.
[17] P. Dollar ; V. Rabaud ; G. Cottrell ; S. Belongie, "Behavior recognition via sparse spatio-temporal features," in 2005 IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance, Beijing, China, pp.65-72, Oct 2005.
[18] Samir Bouindour ; Mohamad Mazen Hittawe , "Abnormal Event Detection Using Convolutional Neural Networks and 1-Class SVM classifier," in 8th International Conference on Imaging for Crime Detection and Prevention (ICDP 2017), Madrid, Spain, pp.1-6, Dec 2017.
[19] Mahmudul Hasan ; Jonghyun Choi ; Jan Neumann ; Amit K. Roy-Chowdhury ; Larry S. Davis, "Learning Temporal Regularity in Video Sequences," in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, pp.733-742, Jun 2016.
[20] Wenqing Chu ; Hongyang Xue ; Chengwei Yao ; Deng Cai, "Sparse Coding Guided Spatiotemporal Feature Learning for Abnormal Event Detection in Large Videos," IEEE Transactions on Multimedia, vol. 21, no. 1, pp. 246-255, Jan 2019.
[21] Yong Shean Chong,and Yong Haur Tay, "Abnormal event detection in videos using spatiotemporal autoencoder.," in 14th International Symposium, ISNN 2017, Sapporo, Hakodate, and Muroran, Hokkaido, Japan, pp.189-196,Jun 2017.
[22] Cewu Lu, Jianping Shi, Jiaya Jia, "Abnormal Event Detection at 150 FPS in Matlab," in 2013 IEEE International Conference on Computer Vision, Sydney, NSW, Australia, pp.2720-2727, Dec 2013.
[23] Amit Adam, Ehud Rivlin, Ilan Shimshoni and Daviv Reinitz, "Robust real-time unusual event detection using multiple fixedlocation monitors," IEEE Transactions on Pattern Analysis and Machine Intelligence, pp. 555 - 560, 21 Jan 2008.
[24] V. Mahadevan, W. Li, V. Bhalodia, and N. Vasconcelos, "Anomaly detection in crowded scenes," in 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, CA, USA, pp.1975–1981, June 2010.