多鏡頭視訊監控系統之前景區塊偵測與位元率分配機制

Foreground Detection and Rate Allocation in Multi-Camera Surveillance System

林明毅(Ming-Yi Lin)

指導教授 ： 張寶基

國立中央大學/資訊電機學院/通訊工程學系/碩士(2006年)

摘要

在新一代的監控系統中，使用網路影像錄影主機(NVR)與網路攝影機(IP Camera)是一個未來發展的趨勢，而當多個視訊流一起在固定頻寬的通道上傳輸時，一個有效的位元率分配機制是必需的。在這篇論文中，我們提出一套前景區塊偵測機制(EBFBD)，來找出畫面中變化的區塊，並且依照區塊的數量來決定攝影機的重要性。依此重要性我們提出一套調適性位元率分配的演算法(AQRDRA)，讓重要性高的攝影機擁有較高的位元率配置，以獲得較佳的視訊品質。最後，我們開發出一套以H.264為基礎的多鏡頭視訊監控系統，讓上述之演算法可以在此平台上獲得驗證。
我們將所提之演算法在實際有效頻寬1.1Mbps下做八路攝影機的模擬實驗，結果證實所提的方法，在幾乎不影響非重點攝影機的視覺品質下，比起位元率均分法可大幅提升重點攝影機的視訊品質最高達8.7dB之多，且有助於H.264位元率控制機制更有效達成所設定之目標位元率。

關鍵字

多鏡頭視訊監控 ； 位元率分配 ； 前景區塊偵測 ； H.264視訊編碼

並列摘要

In the new generation of video surveillance system, adopting NVR (Network Video Recorder) and IP Camera will become the future trend. When the multiple video streams are transmitted together through the fixed bandwidth channel, an efficient rate allocation mechanism is necessary. In this thesis, we develop an Edge-based Foreground Block Detection (EFBD) method to find out changing (foreground) blocks and then determine the importance of cameras based on EFBD. Accordingly, we propose an Adaptive Q-R-D Rate Allocation (AQRDRA) method to allocate higher bitrate to active cameras for better visual quality. Finally, we develop a multi-camera surveillance system using H.264 codec to implement and verify our proposed methods.
The experiments are conducted under the total available bandwidth 1.1Mbps with eight cameras. The experimental results demonstrate that the proposed scheme outperforms uniformly-distributed rate allocation. Without scarifying inactive camera too much, the proposed scheme can enhance the video quality of active camera by 8.7dB at most. Moreover, our proposed method is beneficial for the H.264 rate control scheme to achieve the target rate.

並列關鍵字

multi-camera surveillance system ； rate allocation ； foreground block detection ； H.264 video coding

參考文獻

[1]W. Hu, T. Tan, L. Wang and S. Maybank, "A survey on visual surveillance of object motion and behaviors," IEEE Transactions on Systems, Man and Cybernetics Part C: Applications and Reviews, vol. 34, pp. 334-352, 2004.

[2]K. Toyama, J. Krumm, B. Brumitt and B. Meyers, "Wallflower: Principles and practice of background maintenance," in Proceedings of the 1999 7th IEEE International Conference on Computer Vision (ICCV''99), Sep 20-Sep 27 1999, 1999, pp. 255-261.

[5]R. Cucchiara, C. Grana, M. Piccardi and A. Prati, "Detecting moving objects, ghosts, and shadows in video streams," IEEE Trans. Pattern Anal. Mach. Intell., vol. 25, pp. 1337-1342, 2003.

[6]H. Lin, T. Liu and J. Chuang, "A probabilistic SVM approach for background scene initialization," in International Conference on Image Processing (ICIP''02), Sep 22-25 2002, 2002, pp. 893-896.

[7]A. Makarov, "Comparison of background extraction based intrusion detection algorithms," in Proceedings of the 1996 IEEE International Conference on Image Processing, ICIP''96. Part 1 (of 3), Sep 16-19 1996, 1996, pp. 521-524.