簡易檢索 / 詳目顯示

回結果列表
研究生： 羅家偉
Luo, Jia-Wei
論文名稱： 利用二階段模板比對方法於物件偵測之研究
Object Detection by Using Two Stage Template Matching Method
指導教授： 賴源泰
Lai, Yen-Tai
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 電機工程學系
Department of Electrical Engineering
論文出版年： 2010
畢業學年度： 98
語文別： 英文
論文頁數： 47
中文關鍵詞： 模板比對 、物件偵測 、旋轉不變
外文關鍵詞： Template matching, Object detection, Rotation invariant
相關次數： 點閱：79  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 模板比對是一項早期發展的圖形比對方式，可用來搜尋圖片中使用者感興趣的部份。此方法亦被使用在產品的品質監測，機器導航或者是用於圖片的邊緣偵測。一般的比對方式須將一張欲搜尋模板影像在目標影像上的所有可能位置進行相似度的計算，獲得最高相似度的位置即標示為偵測結果。然而當欲搜尋的模板在目標影像上具有不同的旋轉角度或是大小改變時，原始的比對方法將不再適用。
    本論文提出一個二階段的模板比對方法，第一階段結合了環形投影與多個模板的比對方式篩選出可能為正確位置的候選點。在第二階段再針對這些候選點做進一步的比對，以提高其應用在物件偵測的準確性。利用本方法進行物件偵測，相較於傳統比對方式不僅更加快速，且對於具旋轉與尺度縮放性質的物件偵測結果仍保有一定的準
    確性。

    Template matching is an early developed image matching approach which can be used to find the part of interest in the scene image. It can be used in manufacturing as a part of quality control, mobile robot navigation, or edge detection in images. The conventional matching process is measuring the similarity between the template image and the sub-image at every possible position on the scene image. However, if the rotation and scaling problems are involved, the conventional matching method is not suitable.
    In this thesis, a two stage template matching method is proposed. The first stage selects pixels with high possibility to be the correct match as candidate positions by using the ring projection and multi-template matching method. The second stage provides a further inspection to improve the matching precision in the real scene images. The proposed method is much faster than the conventional one and the results of object detection are invariant to rotation and scaling problems.

    Chapter 1 Introduction..................................................... 1 1.1 Motivation .................................................................................. 1 1.2 Comparison between Feature-based Matching and Correlation Based Matching .......................................................................... 2 1.3 The concept of correlation-based template matching ................ 2 1.4 Thesis organization .................................................................... 4 Chapter 2 Conventional Correlation-Based Template Matching .......................................................... 5 2.1 Introduction ................................................................................ 5 2.2 Related Works for Reducing Computational Cost ..................... 6 2.3 Scale-Invariant Approaches ....................................................... 7 2.4 Rotation-Invariant Approaches .................................................. 8 Chapter 3 Ring Projection Representation ..................... 9 3.1 Theorem ..................................................................................... 9 3.2 The Measurement of Similarity ............................................... 14 3.3 Comparison of Computational Complexity ............................. 15 Chapter 4 Two Stage Template Matching Method ....... 16 4.1.1 Stage One: Scale Estimation ................................................. 17 4.1.2 Preparation for Pattern Scale Estimation .............................. 17 4.1.3 Matching Process at Stage One............................................. 19 4.1.4 Candidate Pixels Selection .................................................... 22 4.2.1 Stage Two: Rotation Angle Detection ................................... 24 4.2.2 Rotation Tolerance of the Conventional Template Matching ................................................................ 24 4.2.3 Radial Mapping Process ....................................................... 27 4.3 Summary of the Two Stage Matching Method ........................ 30 Chapter 5 Experimental Results .................................... 32 5.1 Complexity Analysis ................................................................ 32 5.2 Experimental Results and Performance ................................... 36 5.3 Video Tracking Application ..................................................... 41 Chapter 6 Conclusion ..................................................... 44 References........................................................................ 45 LIST OF TABLES Table 5.1 Computational complexity analysis ................................................................... 35 Table 5.2 Average matching time ...................................................................................... 36 Table 5.3 Matching accuracy ............................................................................................. 41 LIST OF FIGURES Fig.1 Correlation-based Template Matching ....................................................................... 3 Fig.2.1 Image pyramid of the scene image .......................................................................... 6 Fig.2.2 Image pyramid of the template image ..................................................................... 6 Figs. 3.1 Illustration of the ring projection ........................................................................ 11 Figs 3.2 Result-1 of ring projection ................................................................................... 12 Figs 3.3 Result2 of ring projection transformation............................................................ 13 Figs 4.1 Multi-template ring projection transformation .................................................... 18 Figs 4.2 Apply ring projection to Scene image ................................................................. 19 Figs 4.3 Constructing k correlation matrices ..................................................................... 21 Fig 4.4 Match two ring projection vectors with the same number of elements ................ 21 Figs 4.6 An experiment of rotation tolerance .................................................................... 25 Fig. 4.7 Matching ration of distinctive orientation ............................................................ 26 Fig 4.8 Result of polar transformation .............................................................................. 27 Fig 4.9 Calculating the mean pixel value to obtain radial mapping vectors ..................... 28 Fig 4.10 Radial mapping vector sets ................................................................................. 29 Fig. 4.11 Flow chart of the proposed matching method .................................................... 31 Figs 5.1 Detection results of the proposed method ........................................................... 33 Figs 5.2 Detection results of the ring projection method .................................................. 34 Figs 5.3 Matching results of test Images: Book Cover ..................................................... 37 Figs 5.4 Matching results of test images: Sign1 ................................................................ 38 Figs 5.5 Matching results of test images: Sign2 ................................................................ 39 Figs 5.6 Matching results of test images: Sign .................................................................. 40 Figs 5.7 Application on video sequence 1 ......................................................................... 42 Figs 5.8 Application on video sequence 2 ......................................................................... 43

    [1] R. C. Gonzalez and R. E. Woods, Digital Image Processing, Second Edition, Prentice
    Hall, 2002.
    [2] Milan Sonka, Vaclav Hlavac, and Roger Boyle, Image processing, analysis, and
    machine vision , Chapman & Hall Computing, London, 1993.
    [3] Ryo Takei, “A New Grey-Scale Template Image Matching Algorithm Using The
    Cross-Sectional Histogram Correlation Method”, July 2003.
    [4] K. Tanaka, M. Sana. Ohara, and M.Okudaira, “A parametric template method and its
    application to robust matching,”IEEE Conference on Computer Vision and Pattern
    Recognition, Vol. 1, pp. 620-627, June 2000.
    [5] D.M. Tsai and C. H. Chiang, “Rotation-Invariant Pattern matching Using Wavelet
    Decomposition,” Pattern Recognition Letters, vol 23, pp. 191-201, 2002.
    [6] M. S. Choi, W. Y. Kim, “ A novel two stage template matching method for rotation
    and illumination invariance,” Pattern Recognition, Vol. 35,pp. 119-129, 2002.
    [7] D. M. Tsai and Y. H. Tsai, “ Rotation-Invariant Pattern Matching with Color
    Ring-Projection,” Pattern Recognition, Vol. 35, pp. 131-141, 2002.
    [8] Du-Ming Tsai and Chien-Ta Lin, “Fast normalized cross correlation for defect
    detection,” Pattern Recognition Letters, Volume 24, Issue 15, November 2003.
    [9] L. D. Stefano, S. Mattoccia, and F. Tombari, “ZNCC-Based Template Matching using
    Bounded Partial Correlation,” Pattern Recognition Letters, Vol. 26, 2005, pp.
    2129-2134.
    [10] J. MacLean and J. Tsotsos. “Fast Pattern Recognition Using Gradient-Descent Search
    in an Image Pyramid,” International Conference on Pattern Recognition (ICPR’00),
    vol. 2, pp. 2873, 2000.
    [11] R. Y. Wong, E. L. Hall, “Sequential hierarchical scene matching,” IEEE Trans.
    Comput. Vol. 27, pp. 359-366, 1978.
    [12] R.J. Prokop, A.P. Reeves, “A survey of moment-based techniques for unoccluded
    object representation and recognition”, CVGIP Graph. Models Image Process. 54 (5),
    pp.438-460, 1992.
    [13] Tang, Y. Y., Cheng, H, D., Suen, C. Y., “Transformation-ring-projection(TPR)
    algorithm and its VLSI implementation”. Int. J. Pattern Recog. Artifical Intell.
    5,25-56,1991.
    [14] Yuen, P. C., Feng, G. C., Tang, Y. Y. "Printed Chinese character similarity
    measurement using ring projection and distance transform". Int. J. pattern Recog.
    Artifical Intell. 12, 209-221, 1998.

    無法下載圖示 校內：2020-07-20公開
    校外：不公開
    電子論文尚未授權公開，紙本請查館藏目錄
    QR CODE