臺灣博碩士論文加值系統

在許多視覺式監控系統中，通常藉由建立背景影像，再將輸入的影像與背景影像做比較，以擷取前景物 。本研究所考慮的為動態環境，亦即環境中同一位置可能會有多種不同亮度及顏色展現的背景，而且環境中的光線也會隨時間而有所變動。要建置動態環境之背景，通常採用調適性高斯混合模型（Adaptive Gaussian Mixture Model），利用多個高斯函數來描述反覆出現的多種背景值，並透過函數參數值的調適，以適應光線所產生的變化。然而，已往的背景模型都沒有把陰影當做背景的一部分，使得陰影常會被當成前景物擷取出來，造成應用上的錯誤。本研究的目的，即在於建立可調適性背景與陰影的結合模型，以便在動態環境中，利用此模型擷取出沒有陰影的前景物影像。
建立此模型的主要流程分為兩部分，首先為陰影偵測：利用陰影亮度的物理性質篩選出可能是陰影的像素點，再將這些像素點依照反射率分成多個區塊，然後計算每個區塊的真實顏色向量；由於真實顏色向量已移去光線所造成的影響，因此可直接與資料庫中的背景顏色向量比對，以判斷出哪些區塊是背景卻被陰影覆蓋的區塊。第二部分則是建立陰影的顏色分佈：若一像素點被判定在陰影區塊中，便在其背景高斯混合模型中， 以額外的高斯函數描述該像素點的陰影顏色分佈 。建立出背景與陰影的整合模型，除了可用於擷取影像中的前景物，還可以用於偵測場景中的陰影部份，以便了解場景中的光線資訊，用途相當廣泛。
關鍵字：動態環境、調適性高斯混合模型、前景擷取、陰影偵測

In this paper, we integrate shadow information into the background model of a scene in an attempt to detect both shadows and foreground objects at a time. Since shadows accompanying foreground objects are viewed as parts of the foreground objects, shadows will be extracted as well during foreground object detection. Shadows can distort object shapes and may connect multiple objects into one object. On the other hand, shadows tell the directions of light sources. In other words, shadows can be advantageous as well as disadvantageous. To begin, we use an adaptive Gaussian mixture model to describe the background of a scene. Based on this preliminary background model, we extract foreground objects and their accompanying shadows. Shadows are next separated from foreground objects through a series of intensity and color analyses. The characteristics of shadows are finally determined with the principal component analysis method and are embedded as an additional Gaussian in the background model. Experimental results demonstrated the feasibility of the proposed background model.

Keywords: Dynamic scene, Adaptive Gaussian Mixture Model, Foreground detection, Shadow detection

目 錄
附表目錄　　　　　　　　　　　 　　　　　　　　　　　vii
附圖目錄　　　　　　　　　　　　　　　 　　　　　　 viii
第一章 簡介　　　　　　　　　　　　　 　　　　1-1
1.1 研究動機與目的……………………………………………1-1
1.2 文獻探討……………………………………………………1-2
1.2.1 背景模型建立………………………………………………1-3
1.2.2 陰影偵測……………………………………………………1-4
1.3 論文架構……………………………………………………1-6

第二章 系統架構　 　　　　　　　　　　　 2-1
2.1 系統架構……………………………………………………2-1
2.2 系統流程圖…………………………………………………2-1
2.3 系統運作……………………………………………………2-2
2.2.1 背景建立與更新……………………………………………2-2
2.2.2 陰影訓練與更新……………………………………………2-4
2.2.3 前景擷取……………………………………………………2-5

第三章 調適性高斯混合模型之背景建立 3-1
3.1 參數意義……………………………………………………3-3
3.2 初始高斯模型………………………………………………3-4
3.3 更新高斯分佈模型參數……………………………………3-5
3.3.1 更新權重……………………………………………………3-5
3.2.2 更新平均值與變異數………………………………………3-6
3.4 估計背景模型………………………………………………3-8

第四章 陰影偵測 4-1
4.1 陰影模型……………………………………………………4-3
4.2 陰影偵測……………………………………………………4-6
4.2.1 Intensity test……………………………………………4-7
4.2.2 Blue ratio test…………………………………………4-8
4.2.3 Partition…………………………………………………4-10
4.2.4 Remove skylight…………………………………………4-11
4.2.5 Body color estimation and Classification…………4-12


第五章 實驗結果 5-1
5.1 背景建立隻實驗探討………………………………………5-1
5.1.1 初始階段……………………………………………………5-2
5.1.2 光影變化劇況………………………………………………5-3
5.1.3 當前景物滯留的情形………………………………………5-4
5.1.4 室外起風的場景……………………………………………5-5
5.1.5 其它結果……………………………………………………5-6
5.2 陰影偵測與前景擷取之實驗探討…………………………5-7

第六章 結論 6-1
6.1 結論…………………………………………………………6-1
6.2 未來工作……………………………………………………6-2
參考文獻 R-1

附表目錄
表5.1高斯模型參數………………………………………………… 5-1
 
附圖目錄
第二章
圖2.1系統流程圖………………………………………………………2-1
圖2.2背景影像…………………………………………………………2-3
圖2.3陰影偵測影像……………………………………………………2-4
圖2.4前景擷取影像……………………………………………………2-5

第三章
圖3.1真實環境下的顏色在RG band分佈圖…………………………3-1
圖3.2像素點建立的高斯分佈…………………………………………3-2
圖3.3高斯混合模型的機率分佈………………………………………3-2
圖3.4四個高斯分佈與權重值示意圖…………………………………3-9
圖3.5背景模型建立結果………………………………………………3-9

第四章
圖4.1點光源造成的陰影………………………………………………4-2
圖4.2區域光源造成的全陰與半陰……………………………………4-2
圖4.3光的反射…………………………………………………………4-3
圖4.4光的反射…………………………………………………………4-5
圖4.5陰影偵測流程……………………………………………………4-6
圖4.6 Example, 馬克杯投影cast shadow在柏油路上……………4-8
圖4.7蒐集路面有陰影與無陰影像素點統計結果…………………4-9
圖4.8 Blue ratio test image……………………………………4-9
圖4.9反射率分區塊圖………………………………………………4-10
圖4.10 Remove skylight image……………………………………4-11

第五章
圖5.1初始階段影像序列………………………………………………5-2
圖5.2光影變化劇烈環境………………………………………………5-3
圖5.3前景物滯留更新情形……………………………………………5-4
圖5.4室外樹葉與草搖動的場景………………………………………5-5
圖5.5不同場景之背景建立結果………………………………………5-6
圖5.6路口晴天之陰影偵測與前景擷取………………………………5-7
圖5.7一般車道晴天場景………………………………………………5-8
圖5.8路口雨天場景……………………………………………………5-8

[Bou08] T. Bouwmans, F. E. Baf, B. Vachon, “Background modeling using mixture of gaussians for foreground detection-a survey,” Recent Patents on Computer Science, vol. 1, pp. 1-19, 2008.

[Cuc01] R. Cucchiara, C. Grana, G. Neri, M. Piccardi and A. Prati, “The Sakbot System for Moving Object Detection and Tracking,” Video-Based Surveillance Systems－Computer Vision and Distributed Processing, pp. 145-157, 2001.

[Cuc01] R. Cucchiara, C. Grana, M. Piccardi, A. Prati and S. Sirotti, “Improving shadow suppression in moving object detection with HSV color information,” Proc. IEEE Intelligent Transportation Systems Conf., pp. 334-339, 2001.

[Dag93] E.L. Dagless,A.T. Ali, J.B. Cruz, “Visual road traffic monitoring and data collection ,” Proc. IEEE Vehicle Navigation and Information Systems Conf., pp. 146-149, 1993.

[Elg00] A. Elgammal, D. Harwood and L. Davis, “Non-parametric model for background subtraction,” Proc. Computer Vision—ECCV, pp 751-767, 2000.

[Hsu84] Y. Hsu, H. H. Nagel and G. Rekers, “New likelihood test methods for change detection in image sequences,” Computer Vision Image Process, vol.26, pp. 73-106, 1984.

[Hor99] T. Horprasert, D. Harwood and L.S. Davis, “A Statistic Approach for Real-time Robust Background Subtraction and Shadow Detection” Proc. IEEE Int’l Conf. Computer Vision ’99 FRAME-RATE Workshop, 1999.

[Lee04] D. Lee, “Online adaptive Gaussian mixture learning for video applications, ” Workshop on Statistical Methods for Video Processing, pp. 105-116, 2004.

[Jab00] S. Jabri, Z. Duric, H. Wechsler and A. Rosenfeld, “Detection and location of people in video images using adaptive fusion of color and edge information,” International Conference on Pattern Recognition, vol.4, pp. 627-630, 2000.

[Mar07] N. Martel-Brisson and A. Zaccarin, “Learning and Removing Cast Shadows through a Multidistribution Approach,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 29, pp. 1133-1146, 2007.

[Mik00] I. Mikic, P.Cosman, G. Kogut and M.M. Trivedi, “Moving Shadow and Object Detection in Traffic Scenes,” International Conference on Pattern Recognition, vol.1, pp. 321-324, 2000.

[Mor03] V. Morellas, I. Pavlidis and P. Tsiamyrtzis, “Deter: detection of events for threat evaluation and recognition,” Machine Vision and Applications, pp. 29-45, 2003.

[Nad04] S. Nadimi and B. Bhanu, “Physical models for moving shadow and object detection in video,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 26, pp. 1079-1087, 2004.

[Nay96] S. K. Nayar, R. M. Bolle, “Reflectance based object recognition,” International Journal of Computer Vision, vol. 17, pp. 219-240, 1996.

[Pic04] M. Piccardi, “Background subtraction techniques: a review,” IEEE International Conference on Systems, Man and Cybernetics, vol.4, pp.3099-3104, 2004.

[Pra03] A. Prati, I. Mikic, M. M. Trivedi, R. Cucchiara, “Detecting moving shadows: algorithms and evaluation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, pp. 918-923, 2003.

[Sal04] E. Salvador, A. Cavallaro and T. Ebrahimi, “Cast shadow segmentation using invariant color features,” Computer Vision and Image Understanding, pp. 238-259, 2004.

[Sat03] I. Sato, Y. Sato and K. Ikeuchi, “Illumination from Shadows,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 25, pp. 290-300, 2003.

[Sch02] O. Schreer, I. Feldmann, U. Goelz and P. Kauff, “Fast and robust shadow detection in videoconference applications,” Proc. Fourth IEEE Int’l Symp. Video Processing and Multimedia Comm., pp. 371-375, 2002.

[Sha85] S.A. Shafer, “Using Color to Separate Reflection Components,” Color
Research and Application, vol. 10, pp. 210-218, 1985.

[Shi06] A. Shimada, D. Arita, R. Taniguchi, “Dynamic control of adaptive mixture-of-gaussians background model,” IEEE International Conference on Video and Signal Based Surveillance, pp. 5, 2006.

[Sta99] C. Stauffer, W. Grimson, “Adaptive background mixture models for real-time tracking,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol.2, pp. 246-252, 1999.

[Tur07] D. Turdu, H. Erdogan, “Improved post-processing for GMM based adaptive background modeling ,” International Symposium on Computer and Information Sciences, pp. 1-6, 2007.

[Wre97] C.R. Wren, A. Azarbayejani, T. Darrell and A.P. Pentland, “Pfinder: real-time tracking of the human body,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.19, pp. 780-785, 1997.

[Yan06] S. Y. Yang, C. T. Hsu, “Background Modeling from GMM Likelihood Combined with Spatial and Color Coherency,” IEEE International Conference on Image Processing, pp. 2801-2804, 2006.

[Zan06] Q. Zang, R. Klette, “Parameter Analysis for Mixture of Gaussians Model,” CITR Technical Report 188, Auckland University, 2006.

[Zha05] Y. Zhang, Z. Liang, Z. Hou, H. Wang and M. Tan, “An adaptive mixture Gaussian background model with online background reconstruction and adjustable foreground mergence time for motion segmentation,” IEEE International Conference on Industrial Technology, pp. 23-27, 2005.

[Ziv04] Z. Zivkovic, “Improved adaptive Gaussian mixture model for background subtraction,” International Conference on Pattern Recognition, vol. 2, pp. 28-31, 2004.