Please use this identifier to cite or link to this item:
標題: 使用邊緣特徵改善CAMShift的物件追蹤方法之效能
Improving the Effectiveness of CAMSHIFT Based on Object Edges
作者: 林易增
Lin, Yi-Tzeng
關鍵字: CAMSHIFT;CAMSHIFT;Object tracking;PCA;物件追蹤;主成分分析
出版社: 資訊科學與工程學系所
引用: [1] 陳勝勇,劉盛,基於OpenCV的計算機視覺技術實現.北京:科學出版社,2008. [2] 劉瑞禎,于仕琪,OpenCV教程-基礎篇.北京:北京航空航天大學出版社,2007. [3] 吳俊霖,陳彥良,“一個不同曝光時間影像序列之強健特徵導向影像定位法,”資訊科學應用期刊,第3卷,第1期, Jul.2007. [4] 唐政元,吳怡樂,王文宏,林澍新,“使用SIFT做物件辨識,”2008數位科技與創新管理國際研討會,Jue.2008. [5] 張人祐,“人型機器人之模仿能力探討與實作,”碩士論文,中興大學資訊科學與工程所,2008. [6] 張全豐,“利用PCA與CHMM於國語數字之語音辨識,”碩士論文, 中興大學應用數學所,2007. [7] G.R. Bradski, “Computer Vision Face Tracking For Use in a Perceptual User Interface,” Intel Technology Journal Q2 ‘98 [8] J.F. Canny, “A computational approach to edge detection,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. PAMI-8, no. 6, pp. 679-698, 1986. [9] D. Comaniciu, V. Ramesh and P. Meer,“Real-Time Tracking of Non-Rigid Objects using Mean Shift,” BEST PAPER AWARD, IEEE Conf. Computer Vision and Pattern Recognition (CVPR''00), Hilton Head Island, South Carolina, Vol. 2, pp. 142-149, 2000. [10] R.O. Duda, P. E. Hart, D. G. Stork, “Pattern Classification,” Wiley November 2000. [11] G. Gini, A. Marchi, “Indoor Robot Navigation with Single Camera Vision,” Proc. Pattern Recognition in Information Systems, Alicante, Spain, Apr.2002. [12] H. Hotelling, “Analysis of a Complex of Statistical Variables into Principal Components,” Journal of Educational Psychology, 24, 1933, pp. 417 -441, pp. 498-520. [13] H.P. Huang, C.T. Lin, “Multi-CAMSHIFT for Multi-View Faces Tracking and Recognition,” 2006 IEEE Intl. Conf. on Robotics and Biomemetics, Kunming, China, Dec.17-21, 2006. [14] D. G. Lowe, “Distinctive Image Features from Scale-Invariant Keypoints”, International Journal of Computer Vision, v.60 n.2, p.91-110, November, 2004. [15] D.G. Lowe, “Object recognition from local scale-invariant features,” International Conference on Computer Vision, Corfu, Greece, pp. 1150-1157, Sep. 1999. [16] Y. Nagumo, A. Ohya, “Human following behavior of an autonomous mobile robot using light-emitting device,” 10th IEEE Int. Workshop on Robot and Human Communication, Bordeaux and Paris, France, pp. 225–230, Sep. 2001. [17] K. Pearson, “On Lines and Planes of Closest Fit to Systems of Points in Space,” Philosophical Magazine, 6(2) , pp. 559 -572, 1901. [18] L.I. Smith, “A Tutorial on Principal Components Analysis,” private communication, Department of Computer Science, University of Otago, New Zealand, 2002. [19] M. Turk and A. Penland, “Face recognition using eigenfaces,” Proceedings of International Conference on Pattern Recognition, pp. 586-591, 1991. [20] P. Wu, L. Kong, X. Li, K. Fu, “A hybrid algorithm combined color feature and keypoints for object detection,” Industrial Electronics and Applications, 2008. ICIEA 2008. 3rd IEEE Conference, pp.1408-1412. 3-5 June 2008. [21] Y. Yoo, T.S. Park, “A Moving Object Detection Algorithm for Smart Cameras,” Computer Vision and Pattern Recognition Workshops, 2008. CVPRW ''08. IEEE Computer Society Conference, pp.1-8, 23-28 Jun. 2008 [22] Wikipedia, “HSL and HSV,”
CAMSHIFT(Continuously Adaptive Mean Shift)是一種常被使用的物件追蹤方法,單純使用顏色來當唯一的辨識特徵。它有一定的效果,且複雜度低,不需要大量的運算。但是不論是物件、背景太過複雜,或是物件快速移動的時候,常常會失去它的效果。
本文提取並增加了物件的邊緣特徵,來加強CAMSHIFT的效果。且為了降低比對的成本,先將物件邊緣特徵使用主成分分析 (Principal Component Analysis, PCA)計算過後,得到另一個資料量較少的新特徵植。之後先使用邊緣特徵粗估物件的位置之後,再使用CAMSHIFT做第二次的辨識。合併兩種方法讓辨識、追蹤的正確率提高,並且不會付出太多計算的代價。並且經由實驗驗證效果的確可行。

Object tracking is a widely-used skill in lots of advanced applications, such as robots vision and surveillance systems where object tracking is adopted as preliminary identification.
Many tracking methods are proposed. CAMSHIFT (Continuously Adaptive Mean Shift), using color as the only identification feature, is very popular. It can achieve certain result. Moreover, it is not complicated, and does not require a great amount of computation. However, the result is usually not satisfactory when the object or the background is too complex, or when the object is moving too quickly.
In this thesis, we improve the effectiveness of CAMSHIFT by adding another feature, edges of the object. In order to reduce the cost of computation, the first step is to use PCA (Principal Component Analysis) to screen the data and reduce the amount of data. PCA data matching is used to find the object in the screen. And then we use CAMSHIFT to do the second matching. The experimental results show that, by combining the two methods, edge feature and CAMSHIFT, we improve the accuracy of object identification and tracking at an affordable cost.
其他識別: U0005-1307200916560500
Appears in Collections:資訊科學與工程學系所

Show full item record

Google ScholarTM


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.