隨著近年攝影機的發展，使得數位化及高解析度的監視系統日漸普及，也對保全系統造成革命性的衝擊，根據研究報告指出，在英國有四百多萬的監視攝影機分布於城市街道，人們可以藉由將攝影機擺在關鍵的地點，藉此對該區域做日以繼夜的監控。因此，數位監視攝影系統在現今的保全上扮演著愈來愈重要的角色，越來越多的監視攝影機被配置於各種場合以確保人員及財產的安全。然而，大量的數位監視攝影資訊管理並不容易，需要花費大量的人力來監視監控畫面，負責監視監控畫面的人員在長時間專注盯著螢幕的情況之下也會造成的體力與精神的衰減與考驗，往往導致安全維護效率的降低。
在本論文中，我們提出一個高效率的影片檢索系統，透過此系統，使用者感興趣的物件(Region of Interest, ROI)可以藉由電腦視覺中的物件偵測與物件辨識在長時間的影片中快速且有效率的被擷取出來，並且快速得索引與瀏覽。其利用了高斯混合模型(Gaussian Mixture Model, GMM)作為物件偵測的方法，為了讓使用者有效率的找出符合搜尋條件的物件，本文提出低複雜度範圍樹(Range Tree)來做為搜索的方法，其透過影片摘要(Video Abstraction)的技術，搜索到的物件將在短時間中播放，藉此達到快速檢索與瀏覽進而節省時間的目的。透過本系統，前處理物件擷取的部分FPS可以達到32，而搜索的的部分，時間複雜度則是從O(N)下降為O(logD-1N)。

With the development of surveillance, digital and high resolution surveillance become more and more popular, impacting security system. It is reported that in the UK alone there are 4.2 million security cameras covering city streets. More and more digital surveillances are placed in everywhere for safety and security. Therefore digital surveillance system plays indispensable role in security today. However, the great amount of video captured form digital surveillance is difficult to manage and retrieve. 
In this study, we propose an efficient video retrieval technique. With the system, the Region of Interest (ROI) could be extracted in long video effectively and user could browse it with quick and easy way. According to the characteristic of the object in foreground distribution for the real-world video sequence, this work employs Gaussian Mixture Model (GMM) is used for object detection. In order to let users, a new video synopsis search approach, low-complexity rang tree algorithm, is proposed to improve the search the objects matching the conditions effectively in this work. With the time and space redundancy-reducing technique of video synopsis, the objects searched by users could be display in very short time. Therefore wanted objects and events could be found out and displayed quickly without wasting time watching the part without ROI. For the test video sequences, it can have an accuracy rate of 95%  and achieve 32 Frame Per Second (FPS) of online phase in processing speed and time complexity of searching decrease from O(N) to O(logD-1N).

中文摘要	I
英文摘要	II
內文目錄	III
圖表目錄	V

第一章  緒論	1
1.1 研究背景與動機	1
1.2 論文架構	3
第二章  相關技術	4
2.1物件偵測與追蹤系統	4
    2.1.1物件偵測	5
    2.1.2物件追蹤	8
2.2影片摘要相關研究	10
    2.2.1靜態摘要	11
    2.2.2動態摘要	12
    2.2.3其他	13

2.3影片檢索相關研究	15
2.4總結	17
第三章  事件偵測流程	19
3.1偵測前景物件	19
3.2色相轉換	20
3.3離散小波轉換	25
3.4高斯混合模型	28
3.5形態學補償	31
3.6標籤化	35
第四章  影片摘要檢索系統	40
4.1 概述	40
4.2 精簡影片摘要	40
4.3 物件檢索基於改良式範圍樹	47
第五章  實驗結果	61
第六章  結論	67
參考文獻	68

圖目錄

圖2.1高斯背景模型偵測法	7
圖2.2自適應性的快速移動估測流程	10
圖2.3影片摘要	15
圖2.4影片檢索系統	16
圖3.1 online phase流程示意圖	20
圖3.2色相轉換示意圖	21
圖3.3彩色影像轉換為灰階影像	21
圖3.4 RGB色彩方塊	22
圖3.5加色模式	22
圖3.6 YUV的色差信號與顏色對應關係	23
圖3.7以SMDWT降解析度示意圖	27
圖3.8高斯混合模型物件偵測流程	30
圖3.9膨脹運算示意圖	32
圖3.10侵蝕運算示意圖	33
圖3.11影像經由斷開運算之結果	35
圖3.12標籤化時掃描影像的方向	36
圖3.13標籤化時所檢查的近鄰點	37
圖3.14給予前景物件暫時的標籤值	38
圖3.15透過群組編號給定最後的標籤值	39
圖4.1影片摘要示意圖	41
圖4.2產生影片摘要流程圖	42
圖4.3閃頻效應	43
圖4.4閃頻效應產生的重疊	44
圖4.5 Dead Zone產生與物件貼合	45
圖4.6連續物件流	47
圖4.7檢索系統圖	49
圖4.8 二維樹狀結構	50
圖4.9 二維範圍樹的搜索示意圖	52
圖4.10節點所包含的子集合陣列	53
圖4.11搜尋效能圖	55
圖4.12特徵點範圍搜索	56
圖4.13搜尋方式	57
圖4.14系統操作流程	60
圖5.1影片摘要實驗結果截圖	63
圖5.2搜索效能曲線圖	65
圖5.3搜索比對	66























表目錄

表4.1樹狀結構的效能	50
表4.2改良範圍樹的效能	53
表5.1系統物件偵測的效能	56
表5.2系統影片摘要的壓縮效能	57
表5.3系統搜尋效能	59

[1]	J. Nam and A.H. Tewfik, “Video abstract of video,” IEEE Workshop on Multimedia Signal Processing , pp.117-122, 1999.
[2]	C. Pal and N. Jojic, “Interactive montages of sprites for indexing and summarizing security video,” IEEE Computer Society Conference on  Computer Vision and Pattern Recognition, vol. 2, pp. 1192, June 2005.
[3]	M. Furini, “Fast play: a novel feature for digital consumer video devices,” IEEE Transactions on Consumer Electronics, vol. 54, no. 2, pp. 513-520, May 2008.
[4]	M. A. Smith and T. Kanade, “Video skimming and characterization through the combination of image and language understanding techniques,” IEEE Proc. Computer Vision and Pattern Recognition, pp. 775-781, Jun. 1997.
[5]	Y. Zhuang, Y. Rui, T.-S. Huang, and S. Mehrotra, “Adaptive key frame extraction using unsupervised clustering,” IEEE Proc. Image Processing, vol.1, pp. 866-870, Oct. 1998.
[6]	C. Kim and J. N. Hwang, “Object-based video abstraction for video surveillance systems,” IEEE Transactions on Circuits and Systems for Video Technology, vol.12, no.12, pp. 1128- 1138, Dec. 2002.
[7]	D. Farin, W. Effelsberg and P.H.N. de With, “Robust clustering-based video-summarization with integration of domain-knowledge,” IEEE International Conference on Multimedia and Expo, vol.1, no., pp. 89- 92, 2002.
[8]	A. Rav-Acha, Y. Pritch, and S. Peleg, “Making a long video short: dynamic video synopsis,” IEEE Conference on Computer Vision and Pattern Recognition, pp. 435-441, June 2006.
[9]	Y. Pritch, A. Rav-Acha, and S. Peleg, “Nonchronological video synopsis and indexing,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 30, no. 11, pp. 1971-1984, Nov. 2008.
[10]	C.-H. Lin and Y.-C. Chen, “A LDWT-based surveillance video synopsis system using color-based SVM,” IPPR Conference on Computer Vision, Graphics, and Images Processing, Aug. 2011.
[11]	 J. Ren, P. Astheimer, D.D. Feng, “Real-time moving object detection under complex background,” International Symposium on Image and Signal Processing and Analysis, vol. 2, pp. 662-667, Sept. 2003.
[12]	 D. Fan, M. Cao and C. Lv, “An updating method of self-adaptive background for moving objects detection in video,” International Conference on Audio, Language and Image Processing, pp.1497-1501, July 2008.
[13]	 W. Zhang, M.-J. Wu, G. Wang and H. Yin, “An adaptive computational model for salient object detection,” IEEE Transactions on Multimedia, vol. 12, no. 4, pp. 300-316, June 2010.
[14]	 K. Chamnongthai, “Efficient particle filter using non-stationary gaussian based model,” International Conference on Electrical Engineering/Electronics, Computer, Telecommunications and Information Technology, pp.468-471, May. 2011.
[15]	 O. D. Nouar, G. Ali, and C. Raphael, “Improved object tracking with CamShift algorithm,” IEEE International Conference on Acoustics, Speech and Signal Processing, vol. 2, pp. II-657-II-660, May. 2006.
[16]	 C. Zhang and Y. Qiao, “An improved CamShift algorithm for target tracking in video surveillance,” IT&T Conference, 2009.
[17]	 C.-H. Hsia, Y.-J. Liou, Y.-J. Dai, and J.-S. Chiang, “Adaptive motion estimation based on CamShift algorithm for moving object tracking,” International Conference on Service and Interactive Robotics, pp. 316-321, Nov. 2011.
[18]	X. Zhu, A. K. Elmagarmid, X. Xue, L. Wu, A. C. Catlin, “InsightVideo: toward hierarchical video content organization for efficient browsing, summarization and retrieval,” IEEE Transactions on Multimedia, vol.7, no.4, pp. 648- 666, Aug. 2005.
[19]	 B. Sugandi, H. Kim, J. K. Tan, and S. Ishikawa, “Tracking of moving objects by using a low resolution image,” International Conference on Innovative Computing, Information and Control, pp. 408-408, Sept. 2007.
[20]	 C.-H. Hsia, J.-M. Guo, and J.-S. Chiang, “Improved Low-complexity algorithm for 2-D integer lifting-based discrete wavelet transform using symmetric mask-based scheme,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 19, pp. 1202-1208, Aug. 2009.
[21]	 F.-H. Cheng, and Y.-L. Chen, “Real time multiple objects tracking and identification based on discrete wavelet transform,” Pattern Recognition, vol. 39, no. 3, pp. 1126-1139, Jun. 2006.
[22]	 S.-D. Jean, C.-M. Liu, C.-C. Chang, and Z. Chen, “A new algorithm and its VLSI architecture design for connected component labeling,” International Symposium on Circuits and Systems, vol. 2, pp. 565-568, May 1994.
[23]	 W.-K. Chan and S.-Y. Chien, “Subword parallel architecture for connected componentlabeling and morphological operations,” IEEE Asia Pacific Conference on Circuits and Systems, pp. 936-939, Dec. 2006.
[24]	 K. Suzuki, I. Horiba, and N. Sugie, “Linear-time connected-component labeling based on sequential local operations,” Computer Vision and Image Understanding, vol. 89, pp. 1-23, Jan. 2003.
[25]	 H. Hedberg, F. Kristensen, and V. Owall, “Implementation of a labeling algorithm based on contour tracing with feature extraction,” International Symposium on Circuits and Systems, pp. 1101-1104, May 2007.
[26]	 L. He, Y. Chao and K. Suzuki, “A run-based two-scan labeling algorithm,” IEEE Transactions on Image Processing, vol. 17, pp. 749-756, May 2008.
[27]	 D. G. Lowe, “Distinctive Image features from scale-invariant keypoints,” International Journal of Computer Vision, vol. 60, no. 2, pp. 91-110, Nov 2004.
[28]	 C.-H. Hsia and J.-S. Chiang, “Real-time moving objects detection and tracking with direct LL-mask band scheme,” International Journal of Innovative Computing, Information and Control, vol. 8, no. 7(A), pp. 4451-4468, Jul. 2012.