近年來，微軟推出的Kinect利用內附的深度感測器，計算出人體的骨架，稱為深度骨架偵測，讓體感偵測的方式有更多元的變化，相關的研究不斷地提出與跟進，而動作的偵測與分析亦是探討的研究之一。
    透過深度影像體感偵測，在進行部分動作時會無法偵測到，使得偵測動作時會發生誤判情形。例如:腳交叉、躺姿。為了改善此誤判情形，本論文透過影像顏色區塊的重心來達成動作的偵測，確保追蹤的部分為身體的骨架，來達到顏色追蹤，以解決誤判和追蹤的問題，達到骨架的正確追蹤。

In recent years, using of Microsoft's Kinect depth sensor to calculate the human skeleton is called depth skeleton detection. So the way of somatosensory detection can more diverse. Related researches are constantly raised and follow-up, and the action of the detection and analysis is also discussed.
    There are some motions cannot be detected by the skeleton from depth image, for example, foot-cross and lying will cause judgment failed.
    In order to improve this misjudgment, this thesis uses the centroid of the color regions to achieve the motion detection. Ensuring the tracking point is the skeleton of the body to achieve the color tracking and solving the problem of misjudgment and tracking. Let the skeleton be tracked correctly.

目錄
第一章 緒論 1
1.1 研究背景 1
1.2 研究動機及研究目的 1
1.3 論文架構 2
第二章 相關研究 3
2.1 影像格式(Image Format) 3
2.1.1 HSV色彩空間(HSV Color Space) 3
2.1.2 YCbCr色彩空間(YCbCr Color Space) 5
2.2 影像切割(Image Segmentation) 6
2.2.1 Canny Edge Detection 6
2.2.2 Flood Fill 7
2.3 影像追蹤(Image Tracking) 9
第三章 研究方法 11
3.1 系統架構 11
3.2 Training 12
3.2.1 定義目標顏色 12
3.2.2 顏色特徵 13
3.2.3 色彩直條圖(Image Histogram) 13
3.2.4 Training Data 15
3.3 Initialization 16
3.3.1 設定偵測區域 17
3.3.2 減少影像雜訊 17
3.4 Color Detection 19
3.4.1 調整對比(Contrast Adjustment) 19
3.4.2 遮罩(Mask) 20
3.4.3 顏色偵測(Color Detection) 23
3.5 Region Growing 25
3.6 Color Tracking 27
第四章 實驗數據 31
4.1 實驗環境說明 31
4.2 Training data數據 31
4.3 調整對比成效 33
4.4 Region Growing效果 34
4.5 Tracking軌跡呈現 36
第五章 結論與未來方向 39
5.1 結論 39
5.2 未來方向 39
參考文獻 40
附錄–英文論文 41

圖目錄
圖1.1 腳交叉時的誤判 2
圖2.1 HSV色輪[2] 4
圖2.2 YCbCr色彩空間圖[3] 6
圖2.3 Canny edge detection實作圖 7
圖2.4 Flood fill實作概念圖 8
圖2.5 Camshift追蹤效果[6] 10
圖3.1 系統架構圖 12
圖3.2 色彩直條圖 14
圖3.3 HSV色彩直條圖轉換 15
圖3.4 Training圖 16
圖3.5 人體骨架偵測深度圖 17
圖3.6 ISO值比較圖 18
圖3.7 高斯模糊化圖示 19
圖3.8 直條圖均衡化比較圖 20
圖3.9 遮罩成果圖 21
圖3.10 飽和度和明度調整 22
圖3.11 顏色偵測流程 23
圖3.12 二值化轉換圖 24
圖3.13 Region Growing示意圖 27
圖3.14 追蹤目標重心 28
圖3.15 系統流程圖 30
圖4.1 Training過程 32
圖4.2 對比調整圖 34
圖4.3 Region Growing效果圖 35
圖4.4 Tracking初始化動作 36
圖4.5 腳誤判情形與顏色區塊追蹤 36
圖4.6 行走時tracking過程 38

表目錄
表4.1 實作環境規格表 31
表4.2 Training結果呈現表 33

[1]Booth Paul A., “An Introduction To Human-Computer       
   Interaction,” Psychology Press, 1989.
[2]Raphael Gonzalez, Richard E. Woods , “Digital Image 
   Processing,”2nd edition, Prentice Hall Press, ISBN 0-201-
   18075-8, p.295, 2002.
[3]XU Yan, CHEN Xiao-Wei, “Skin Pick-up Patterns for Face 
   Detection,” Computer Simulation, pp.194-199, 2006.
[4]John Canny, “A Computational Approach to Edge Detection,” 
   IEEE Transactions On Pattern Analysis And Machine 
   Intelligence, pp.679-698, NOVEMBER 1986.
[5]Hoon Kang, Seung Hwan Lee, Jayong Lee, ”Image 
   segmentation based on fuzzy flood fill mean shift 
   algorithm,” Fuzzy Information Processing Society 
   (NAFIPS), 2010 Annual Meeting of the North American, 
   pp.1- 6, 2010.
[6]Bradski G.R., “Computer Vision Face Tracking for Use in A 
   Perceptual User Interface,” Intel Technology Journal, 2nd 
   Quarter, pp.13-27, 1998.
[7]Comaniciu D., Meer P., “Mean Shift: A Robust Approach 
   Towards Feature Space Analysis,” IEEE Transactions on 
   Pattern Analysis and Machine Intelligence, pp. 603-619, 
   2002.
[8]Gedraite E.S., Hadad M., ”Investigation on the effect of 
   a Gaussian Blur in image filtering and segmentation
   ,” ELMAR, 2011 Proceedings, pp.393- 396, 2011.
[9]Ibrahim, H., Kong N.S.P, ”Image sharpening using sub-
   regions histogram equalization,” IEEE Transactions on 
   Consumer Electronics, pp.891-895, 2009.
[10]Tilton J.C., Tarabalka Y., Montesano P.M., Gofman E., 
   ”Best Merge Region-Growing Segmentation With Integrated 
   Nonadjacent Region Object Aggregation,” IEEE Transactions 
   on Geoscience and Remote Sensing , pp.4454-4467, 2012.