臺灣博碩士論文加值系統

隨著科技的快速發展，3D相機不再只是傳統的雙眼相機，能夠估測景深資訊的景深估測相機也慢慢地發展起來，所估測出來景深資訊的精準度也慢慢地提升，獲得景深資訊也就更加的方便。
由於同一個非平面的物件在景深圖中會具有不同的景深值，在傳統的二維連通區域標記演算法會造成過度分割物件的問題。因此我們提出了三維連通區域標記演算法，利用景深圖的值來切片成256個階層的二元影像，除了使用二維的連通標記考慮上方座標及左邊座標是否有值外，額外再考慮前一個切片的二元影像的上方座標及左邊座標是否有值。
本研究透過Kinect攝影機取得彩色視訊和景深視訊，利用在景深影像中同一物件的景深值會有相近數值的特性，搭配我們所提出的三維連通區域標記演算法將所有的Frame的物件都分割出來，接著從景深視訊的第一張Frame選取感興趣的物件，再利用感興趣的物件與下一張Frame的所有物件進行比對，來偵測出要追蹤的物件。
最後，經過實驗結果顯示實驗結果顯示我們提出的三維連通法所切割出來的物件數量和實際的物件數量相同，解決同一個物件過度分割的問題。我們所提出的物件追蹤的方法與改變偵測法進行比較，經實驗結果可以證明本研究所提出的方法能夠更精確的偵測出追蹤物件。

With the rapidly development of technology, 3D camera is not just the traditional binocular camera. It is able to estimate depth information to depth estimation camera also slowly developed. Estimating the accuracy of depth information is also slowly improved. Getting depth information will be more convenient.
Over-segmentation occurs in the traditional 2D connected component labeling algorithm because a nonplanar object exhibits various depth values in a depth map. This study proposed a 3D connected-component labeling algorithm that slices images into 256 levels of binary images on the basis of depth values. In addition to the up and left coordinates of the current segment, the proposed algorithm considers coordinates of the previous segment.
This study uses Kinect camera to obtain color video and depth video. The depth value of the same object in the depth image has a similar value. Using our proposed 3D connected-component labeling algorithm to segment objects of all frames. A target object was selected from the first frame of the depth video and was compared with all objects in the subsequent frame to detect the tracked object.
Finally, the experimental results revealed that the proposed 3D connected-component labeling algorithm can segment objects according to the actual number of objects in an image, thereby resolving the problem of over-segmentation. The experimental results generated by the proposed object tracking method show that it can detect tracked objects more accurately than a change detection method.

誌謝 I
摘要 II
ABSTRACT III
目錄 IV
圖目錄 VIII
表目錄 X

第一章 緒論 1
1-1 前言 1
1-2 研究目的 2

第二章 研究背景 3
2-1 立體圖像技術 3
2-2 Kinect攝影機 3
2-3 景深圖 5
2-4 透視轉換(Perspective Transformations) 6
2-5 濾波器 8
2-5-1 中值濾波器(Median Filter) 9
2-5-2 雙邊濾波器(Bilateral Filter) 10
2-6 直方圖(Histogram) 12
2-7 門檻值(Thresholding) 12
2-8 二維連通區域標記(Connected-Component Labeling ,CCL) 13
2-8-1 二維連通區域標記 13
2-8-2 二維連通區域標記演算法 14
2-9 輪廓偵測 15
2-10 改變偵測法(Change Detection Method) 16
2-11 OpenCV(Open Source Computer Vision Library) 17

第三章 研究方法 19
3-1 景深圖 21
3-2 透視轉換 22
3-3 濾波器 24
3-4 填補雜訊 24
3-5 三維連通區域標記 25
3-6 圖像相減比對 33
3-7 輪廓偵測 34
3-8 與彩色/景深影像匹配 35

第四章 研究結果 36
4-1 三維連通區域標記演算法 36
4-2 物件追蹤 38
4-2-1 透視轉換 39
4-2-2 濾波器 40
4-2-3 填補雜訊 41
4-2-4 三維連通區域標記 42
4-2-5 物件分割 43
4-2-6 選取感興趣物件 46
4-2-7 物件追蹤 46
4-2-8 輪廓偵測 47
4-2-9 實驗結果 47
4-3 與改變偵測法比較 48

第五章 結論與未來工作 50
5-1 結論 50
5-2未來工作 50

參考文獻 51
附件 53
附件1 物件追蹤第一組 53
附件1-1 彩色視訊截取單張影像 53
附件1-2 景深視訊截取單張影像 53
附件1-3 景深影像透視校正 54
附件1-4 彩色影像裁剪大小 55
附件1-5 景深影像裁剪大小 55
附件1-6 中值濾波 56
附件1-7 雙邊濾波 57
附件1-8 填補雜訊 57
附件1-9 三維連通區域標記 58
附件1-10 物件分割 59
附件1-11 物件追蹤 67
附件1-12 輪廓偵測 67
附件1-13 彩色視訊物件追蹤結果 68
附件1-14 景深視訊物件追蹤結果 68
附件2 物件追蹤第二組 69
附件2-1 彩色視訊截取單張影像 69
附件2-2 景深視訊截取單張影像 69
附件2-3 景深影像透視校正 70
附件2-4 彩色影像裁剪大小 70
附件2-5 景深影像裁剪大小 71
附件2-6 中值濾波 72
附件2-7 雙邊濾波 72
附件2-8 填補雜訊 73
附件2-9 三維連通區域標記 74
附件2-10 物件分割 74
附件2-11 物件追蹤 84
附件2-12 輪廓偵測 85
附件2-13 彩色視訊物件追蹤結果 85
附件2-14 景深視訊物件追蹤結果 86
附件3 物件追蹤第三組 87
附件3-1 彩色視訊截取單張影像 87
附件3-2 景深視訊截取單張影像 87
附件3-3 景深影像透視轉換 88
附件3-4 彩色影像裁剪大小 88
附件3-5 景深影像裁剪大小 89
附件3-6 中值濾波 89
附件3-7 雙邊濾波 90
附件3-8 填補雜訊 90
附件3-9 三維連通區域標記 91
附件3-10 物件分割 91
附件3-11 物件追蹤 97
附件3-12 輪廓偵測 98
附件3-13 彩色視訊物件追蹤結果 98
附件3-14 景深視訊物件追蹤結果 99

[1] A. K. Jain, Fundamentals of Digital Image Processing, Englewood Cliffs, NJ: Prentice Hall, 1989.
[2] C. Fehn, “A 3D-TV approach using depth-image-based rendering (DIBR), in Proc. Vis., Imaging, and Image Process. (VIIP), Benalmádena, Spain, pp. 482–487, 2003.
[3] C. Fehn, “Depth-Image-Based-Rendering (DIBR), Compression and Transmission for a New Approach on 3D-TV,” in Proceeding of International Society for Optical Engineering Conf. Stereoscopic Displays and Virtual Reality Systems XI, vol. 5291, San Jose, CA, USA, pp. 93-104, 2004.
[4] C. Tomasi, R. Manduchi, "Bilateral filtering for gray and color images," in Proc. 6th IEEE Conf. Computer Vision, pp. 839-846, 1998.
[5] C. Kim and J. N. Hwang, “Fast and Automatic Video Object Segmentation and Tracking for Content-Based Applications,” IEEE Transactions Circuits And Systems for Video Technology, vol. 12, no. 2, pp. 122-129, 2002.
[6] D. Bradley and G. Roth, “Adaptive thresholding using the integral image,” Journal of Graphics Tools, vol. 12, no.2, pp.13-21, 2007.
[7] E. Meijering, "A chronology of interpolation: from ancient astronomy to modern signal and image processing," Proceedings of the IEEE, vol. 90, no. 3, pp. 319-342, 2002.
[8] F. Blais, M. Picard, and G. Godin, “Accurate 3d acquisition of freely moving objects,” in Proceedings of the 2nd International Symposium on 3D Data Processing, Visualization and Transmission(3DPVT), Thessaloniki, Greece, pp.422-429, 2004.
[9] G. Bradski, A. Kaehler, Learning OpenCV, Computer Vision with the OpenCV Library, O’Reilly Media, 2008.
[10] H.-I. Kim, S.-H. Lee, and Y.-M. Ro “Multispectral Texture Features from Visible and Near-Infrared Synthetic Face Images for Face Recognition,” IEEE International Symposium on Multimedia, Miami, FL, USA, pp. 593-596, 2015.
[11] J. W. Tukey, “Nonlinear (nonsuperposable) method for smoothing data,” Proceedings of Congress Record EASCON, Washington DC, pp.673-681, 1974.
[12] J. W. Tukey, Exploratory Data Analysis, Reading, MA: Addison-Wesley, 1977.
[13] Kinect Ads: You Are the Controller. Microsoft. October 21, 2010. Retrieved January 25, 2011.
[14] L.Busin, N. Vandehbroucke, L. Macaire, and J-G. Postaire, “Color Space Selection for Unsupervised Color Image Segmentation by Histogram Multi-Thresholding,” IEEE Conference on, Image Processing, ICIP’04., vol. 1, Singapore, pp. 203-206, 2004.
[15] L. He, Y. Chao, and K. Suzuki, “A run-based two-scan labeling algorithm," IEEE Transactions on Image Processing, vol. 17, no. 5, pp. 749-756, 2008.
[16] R. Szeliski, D. Scharstein, H. Hirschmüller, Middlebury Stereo Datasets, [online]. Available: http://vision.middlebury.edu/stereo/data/
[17] R. N. Devich and F. M. Weinhaus, “Image perspective transformations”, Proceedings of SPIE, United States, Vol. 238, pp.322-332, 1980.
[18] R. Hartley and A. Zisserman, Multiple View Geometry in Computer Vision, 2nd ed, New York: Cambridge university press, 2003.
[19] S. E. Umbaugh, Computer Imaging: Digital Image Analysis and Processing, Boca Raton, FL: CRC Press, 2005.
[20] S. Dotenco, M. Dalsass, L. Winkler, T. Wurzner, C. Brabec, A. Maier, and F. Gallwitz, “Automatic Detection and Analysis of Photovoltaic Modules in Aerial Infrared Imagery,” IEEE Winter Conference on, Applications of Computer Vision (WACV), Lake Placid, NY, USA, pp. 1-9, 2016.
[21] S. Suzuki, K. Abe, "Topological structural analysis of digitized binary images by border following," Computer Vision, Graphics, and Image Processing, vol. 30, no. 1, pp. 32-46, 1985.