臺灣博碩士論文加值系統

隨著電腦技術的越來越發達，人們對生命安全的也越來越重視。所以安全監控系統的研究也成為一個重要的議題，其中對於物件追蹤的準確性就是一個重要的關鍵。在處理物件追蹤的時候主要可以分為兩個步驟：首先就是要正確的萃取出畫面中我們所關心的物件，接著才能對這些物件做進一步的追蹤。對於物件的萃取，Kim 等人在2005年提出了一套背景建立技術，這個技術可以在變動的背景中找出我們需要的物件。不過有時候一些背景也會被當作物件偵測出來。在物件追蹤的時候也常常會遭遇到複雜的物件交會情形，這也是造成難以建立一個完善的追蹤系統的主要原因。
在本篇論文中，我們首先對Kim的背景模型做了一些適當的修正並加以使用這些修正可以增加物件偵測時的完整性。接著我們加入了一個使用邊緣模型，藉由這個模型我們可以消除一些錯誤的幻影，另外這些邊緣也給追蹤的演算法提供了良好的特徵。在處理物件的追蹤時我們主要將可能發生的情況分成五大類，針對這五種不同的情況我們設計各自不同的演算法，藉由各個演算法的執行，我們最後可以得到一個準確的追蹤結果。
在實驗結果中，與一般單純的使用背景模型來做檢測的結果相比，再另外加入了一個邊緣模型之後，證明可以有效的降低幻影的發生。而我們的追蹤系統也成功的解決了一些追蹤的時候可能發生的困難。物件分裂的時候我們可以正確的區分不同的物件，或是物件過小的情況下，藉由我們提供了一個以邊緣為基礎的meanshift演算法，讓這些過小的物件仍然可以被順利的追蹤下去。另外我們的演算法對於低取樣頻率的影片也可以提供正確的追蹤結果。我們的系統在一般的電腦上可以達到每秒10到15張畫面的處理速度，這個速度已經可以滿足即時運算的要求。

As the computer technology becomes more and more convenient, people also pay more attention to security of life. Study of surveillance system becomes an important issue nowadays, where an accurate object tracking is one of key topics. When tracking objects, there are two main steps. First, object detection, which can locate the foreground objects of interest from the video, is needed. Second, we have to propose a good system that can track objects with several kinds of interaction. For object detection, Kim et al. proposed a Codebook background modeling algorithm which can detect objects in a varying background. However, sometimes ghost objects are detected. For object tracking, there are usually complicated interactions of multiple objects, causing difficulty.
In this thesis, we utilize and make some modifications on Codebook model so that it can provide a more complete result for detecting objects. Then an edge-based model is combined with Codebook model that can reduce some ghost objects during object detection. We classify all possible events into five different situations, and each situation is processed by a specific algorithm. A robust tracking result is obtained from the object tracking system.
In the experiment, by the combination of Codebook model and edge-base model, it can effectively reduce occurrence of ghost objects. Our proposed tracking algorithm also solves some possible difficulties. When dealing with splitting, objects can be distinguished into correct targets. By using a proposed edge-based meanshift algorithm, objects, which are very small, can be kept tracking successfully. Our system also produces precise result for low frame rate video. The proposed system works at 10 - 15 fps and achieves real-time operation.

Abstract.....................................i
Table of Contents...........................ii
List of Figures..............................v
List of Tables.............................vii
Chapter1 Introduction........................1
1.1 Introduction of Object Tracking..........1
1.2 Motivation...............................3
1.3 System Overview..........................3
1.4 Thesis Organization......................6
Chapter2 Related Works.......................7
2.1 Object Detection.........................7
2.2 Object Tracking..........................9
Chapter3 Background Modeling and Object Detection.........13
3.1 Codebook Background Modeling..........................14
3.1.1 Background Modeling and Detection...................14
3.1.2 Layered Modeling and Detection - Model Maintenance..19
3.2 Some Modifications of Codebook Modeling...............20
3.2.1 A Modified Updating for Brightness Range of Codeword20
3.2.2 Background Subtraction with Spatial Continuity......21
3.3 Foreground Edge Detection.............................23
3.3.1 Sobel Operator......................................23
3.2.2 Edge-based Background Modeling and Subtraction......24
3.4 Morphological Operation...............................25
3.5 Blobs Connection......................................26
3.6 Color Feature of Measurements Extraction..............27
Chapter4 Object Tracking..................................29
4.1 Representation of Target..............................30
4.2 Correspondence and Classifier.........................30
4.3 The One-to-One Process................................32
4.4 The Merging Process...................................33
4.5 The Splitting Process.................................34
4.5.1 Feature Similarity Algorithm........................35
4.6 The New-coming Process................................36
4.7 The Missing Process...................................37
4.7.1 Edge-based Meanshift Algorithm......................38
4.8 Confidence of Tracking Targets........................39
Chapter5 Experimental Results and Discussion..............41
5.1 Object Detection......................................41
5.1.1 Background Subtraction and Updating of Codebook Model.....................................................41
5.1.2 The Ghost Objects and Results with Combination of Edge Detection............................................42
5.1.3 More Detection Results..............................44
5.2 Object Tracking.......................................45
5.2.1 Tracking Results....................................45
5.2.2 Edge-based Meanshift Algorithm......................48
5.3 Performance Analysis..................................49
Chapter6 Conclusion and Future Works......................52
6.1 Conclusion............................................52
6.2 Future Works..........................................53
Reference.................................................54

[1] M. Valera and S. Velastin, ” Intelligent Distributed Surveillance Systems: a Review”, IEE Proceedings Vision, Image & Signal Processing, vol. 152 Issue 2, pp.192-204, 2005
[2] S. Joo and Q. Zheng, “A Temporal Variance-Based Moving Target Detector”, International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance, 2005.
[3] C. Stauffer and W.Grimson, “Adaptive Background Mixture Models for Real-Time Tracking”, International Conference on Computer Vision and Pattern Recognition, pp. 246-252, 1999.
[4] K. Kim, T. H. Chalidabhongse, D Harwood and L. Davis, ”Real-time Foreground–Background Segmentation Using Codebook Model”, Real-Time Imaging, pp. 172-185, 2005.
[5] M. Heikkila and M. Pietikainen, “A Texture-Based Method for Modeling the Background and Detecting Moving Objects”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 28,, pp. 657-652, 2006.
[6] D. Meyer, J. Denzler, and H. Niemann, “Model Based Extraction of Articulated
Objects in Image Sequences for Gait Analysis,” International Conference on Image Processing, vol. 3, pp. 78–81, 1997.
[7] N. Dalal and B. Triggs, “Histograms of Oriented Gradients for Human Detection”, International Conference on Computer Vision and pattern Recognition, vol.1,.pp.886-893, 2005.
[8] N. Dalal, B. Triggs, and C. Schmid, “Human Detection Using Oriented Histograms of Flow and Appearance”, European Conference on Computer Vision, pp.428-441, 2006.
[9] P. Viola and M. Jones, ”Robust Real-Time Face Detection”, International Journal of Computer vol. 57 issue2, pp137-154, 2004.
[10] W. Hu, T. Tan, L. Wang, and S. Maybank, ” A Survey on Visual Surveillance of Object Motion and Behaviors”, IEEE Transaction on Systems, Man, and Cybernetics- Part C, pp334-352, 2004.
[11] C. R. Wren, A. Azarbayejani, T. Darrell, and A. P. Pentland, “Pfinder: Real-Time Tracking of The Human Body,” IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. 19, pp. 780–785, 1997.
[12] A. Baumberg and D. C. Hogg, “Learning Deformable Models for Tracking The Human Body,” in Motion-Based Recognition, M. Shah and R. Jain, Eds. Norwell, MA: Kluwer, pp. 39–60, 1996.
[13] H. Fujiyoshi and A. J. Lipton, ”Real-Time Human Motion Analysis by Image Skeletonization”, IEEE Workshop on Applications of Computer Vision, pp.15-21, 1998.
[14] S. Y. Chien, W.K. Chan, D.C. Cherng, and J.Y. Chang, “Human Object Tracking Algorithm with Human Color Structure Descriptor for Video Surveillance Systems”, International Conference on Multimedia and Expo., pp.2097-2100 , 2006.
[15] T. Yang, S. Z.Li, Q. Pan, and J. Li, “Real-time Multiple Objects Tracking with Occlusion Handling in Dynamic Scenes”, International Conference on Computer Vision and Pattern Recognition, vol. 1, pp970-975, 2005
[16] B. Wu and R. Nevatia, “Tracking of Multiple, Partially Occluded Humans based on Static Body Part Detection”, International Conference on Computer Vision and Pattern Recognition, vol. 1, pp.951-958, 2006.
[17] V. Takala and M. Pietikainen, “Multi-Object Tracking Using Color, Texture and Motion”, International Conference on Computer Vision and Pattern Recognition, pp1-7, 2007.
[18] H. Lim, V.I. Morariu, O. I. Camps, and Mario Sznaier, “Dynamic Appearance Modeling for Human Tracking”, International Conference on Computer Vision and Pattern Recognition, vol.1, pp751-757, 2006.
[19] M. N. Siddiqui, B. Yousaf, “Bare Bones Strategy for Human Detection and Tracking” IEEE Symposium on Computational Intelligence in Image and Signal Processing, pp30-35, 2007.
[20] K. Bai and W. Liu, “Improved Object Tracking with Particle Filter and Mean Shift”, International Conference on Automation and Logistics, pp.431-435, 2007.
[21] S. V. U. Ha and J. W. Jeon, ”Combine Kalman Filter and Particle Filter to Improve Color Tracking Algorithm”, International Conference on Control, Automation and Systems, pp.558-561, 2007.
[22] A. Yilmaz, O. Javed, and M. Shah, “Object Tracking: A Survey”, ACM Computing Surveys, Vol. 38, No. 4, Article 13, pp1-45, 2006.
[23] K. P. William, “Digital Image Processing: PIKS Inside 3rd Edition”, John Wiley & Sons, Inc, pp428, 2001.
[24] T. Yang, S. Z. Li, Q. Pan, and J. Li, “Real-time Multiple Objects Tracking with Occlusion Handling in Dynamic Scenes”, International Conference on Computer Vision and Pattern Recognition, vol.1, pp970-975, 2005.
[25] IEEE International Workshops on Performance Evaluation of Tracking and Surveillance, ftp://pets2001.cs.rdg.ac.uk.
[26] 劉瑞禎, 于仕琪, ” OpenCV教程:基础篇”, 北京航空航天大学出版社, 2007