臺灣博碩士論文加值系統

動作辨識在影像處理的研究領域上一直是很重要的研究方向，本研究主要目的在於提出一套能對人物進行追蹤且辨識出人體基本動作之系統，對於常見動作的判讀能應用於各個生活如居家照護或異常行為等進行追蹤偵測。主要研究方法結合高斯混合背景模型與動作能量圖概念並以模板匹配達到快速且準確的辨識效果，其中高斯混合背景模型是一種適應性背景相減法，能夠針對背景變化的情況得到準確地前景與背景，但在背景物件變動過快或是光影反射等現象時可能會導致誤判成前景物件，因此本研究以形態學處理針對人體輪廓的雜訊抑制與修補並在輪廓面積區域計算下，提高完整人體輪廓的提取，並在動作能量圖中與傳統方式對齊身體頭部或軀幹方式採用不同方法，本研究採用計算質心位置作為對齊人物動作輪廓的基準點，較以往需額外判斷人體部位的位置上，更能快速完成動作能量圖，最後利用事先選定好的三種動作模版交叉進行模板匹配，採用快速的一般化平方差匹配法完成辨識動作的目標，得到整體平均辨識率達約91.4%，主要誤判原因在於動作執行與一般情況相異，以及受攝者人物體型的步伐大小問題導致該動作輪廓軌跡不明顯所致。

Human action recognition has been a very important topic in computer vision and image processing. In this thesis, we present a human action recognition system using the Gaussian mixture background model, morphological processing, motion energy image and template matching. First, we detect human and extract human contour in a video according to the Gaussian mixture background model. After the human contour is extracted, we reduce the noise in human contour by using Gaussian blur. Then we use the erosion and dilation of morphological processing to connect the broken contour area .We remove the other nonhuman targets in the background by calculating to the maximum contour area. Second, in order to set up the motion energy image, we align the centroid in the motion energy image after calculating the coordinate of the centroid in human contour. Finally, according to the three templates selected in advance, we can complete the human action recognition using template matching method. Experiments by three-fold cross validation showed that the accuracy of our human action recognition is 91.4%. Most errors are due to different walking sizes between humans.

摘要 i
ABSTRACT ii
致謝 iv
目錄 v
圖目錄 vii
表目錄 x
一、緒論 1
1.1 前言 1
1.2 研究動機與目的 1
1.3 論文架構 3
二、文獻探討 4
三、研究方法 7
3.1 高斯混和背景模型 9
3.2 形態學處理 12
3.2.1 高斯濾波法 12
3.2.2 膨脹與侵蝕 14
3.2.3 最大面積判斷 17
3.3 Motion Energy Image 19
3.3.1 人體輪廓質心 20
3.3.2 影像疊合 22
3.4 模板匹配 24
四、研究結果與討論 27
4.1 系統架構與實驗資料 27
4.2 實驗結果呈現 32
4.2.1 高斯混和模型與形態學處理結果 32
4.2.2 MEI圖疊合結果 37
4.2.3 模板匹配結果 40
4.3 實驗結果分析與討論 47
4.4 實驗結果比較 56
五、結論與未來發展 57
5.1 結論 57
5.2 未來發展 58
六、參考文獻 59

[1]W. Hu, T. Tan, L. Wang, and S. Maybank, “A survey on visual surveillance of object motion and behaviors,” IEEE Transactions on Systems, Man and Cybernetics, Part C, Vol. 34 , pp. 334 -352, 2004.
[2]Z. Zhang, “Video tracking of people under severe occlusions,” M.S. thesis, Degree of Doctor of Philosophy of the University of Technology, Sydney, Australia., 2010.
[3]S. McKenna, S. Jabri, Z. Duric, A. Rosenfeld, and H. Wechsler, “Tracking groups of people,” Computer Vision and Image Understanding, Vol. 80, pp. 42-56, 2000.
[4]N. Paragios , and R. Deriche. “Geodesic active regions and level set methods for motion estimation and tracking.” Computer Vision and Image Understanding 97, pp.259-282, 2005.
[5]Y. M. Liang, “A study on video-based human action analysis by isomap learning and VLMM techniques,” Ph.D. dissertation, Dept. Comput. Sci. ,National Chiao Tung University., Hsinchu, Taiwan,2009.
[6] S. Ji, W. Xu, M. Yang, and K. Yu.” 3D convolutional neural networks for human action recognition.” Pattern Analysis and Machine Intelligence, IEEE Transactions on 35(1),pp.221-231, 2013.
[7]V. Parameswaran, and R. Chellappa, “View invariance for human action recognition, “International Journal of Computer Vision 66 (1) ,pp.83-101,2006.
[8]M. Barnachon, S. Bouakaz, B. Boufama, and E. Guillou, “Ongoing human action recognition with motion capture, “Pattern Recognition 47 ,pp.238-247,2014.
[9]L. Xia, C. Chen, and J. Aggarwal, “View invariant human action recognition using histograms of 3D joints,” IEEE CVPR Workshop on Human Activity Understanding from 3D Data, 2012.
[10]Z. Ren, J. Yuan, J. Meng, and Z. Zhang, “Robust part-based hand gesture recognition using Kinect sensor,” IEEE Trans. Multimedia, vol. 15, no. 5,pp. 1110–1120, 2013.
[11]O. Patsadu, C. Nukoolkit, and B. Watanapa, “Human gesture recognition using Kinect camera,” The proceeding of International Joint Conference on Computer Science and Software Engineering, pp. 28-32, Bangkok, Thailand,2012.
[12]A.A. Chaaraoui , and F. Florez-Revuelta. “Adaptive human action recognition with an evolving bag of key poses.” IEEE Transactions on Autonomous Mental Development, 6(2),pp. 139-152, 2014.
[13]C. Stauffer, and W.E.L Grimson, “Adaptive background mixture models for real-time tracking,” IEEE Computer Society Conference on Computer Vision and Pattern Recognition, vol.2, pp. 246-252, 1999.
[14]Z. Zivkovic, “Improved Adaptive Gaussian Mixture Model for Background Subtraction,” International Conference on Pattern Recognition , vol. 2, pp. 28- 31,2004
[15]Z. Zivkovic, and F. v. d.Heijden, “Efficient adaptive density estimation per image pixel
for the task of background subtraction, ” Pattern recognition letters, 27(7),pp. 773-780,2006.
[16]S. S. Al-amri, N.V. Kalyankar, and S.D. Khamitkar, “Deblured Gaussian Blurred Images, ” Journal of Computing, vol. 2, pp. 33- 35, 2010,
[17]“GaussianBlur,”http://docs.opencv.org/2.4/doc/tutorials/imgproc/gausian_median_blur_bilateral_filter/gausian_median_blur_bilateral_filter.html,[accessed Nov. 2016]
[18]M. Goyal, “Morphological Image Processing,” International Journal of Computer Science and Technology , vol. 2, no. 4, pp. 161-165, 2011.
[19]“Eroding and Dilating of Morphology,”
http://docs.opencv.org/2.4/doc/tutorials/imgproc/erosion_dilatation/erosion_dilatation.html,
[accessed Nov. 2016]
[20]M. Yokoyama, and T. Poggio, “A contour-based moving object detection and tracking, ” IEEE International Workshop on Visual Surveillance and Performance Evaluation of Tracking and Surveillance, pp. 271-276 , 2005.
[21]“Bigest contour area ,”
http://docs.opencv.org/2.4/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=contourarea#cv2.contourArea, [accessed Nov. 2016]
[22]J. Flusser , and T. Suk, “Rotation Moment Invariants for Recognition of Symmetric Objects,” IEEE Trans. Image Process, vol. 15, no. 12, pp.3784-3790, 2006.
[23]“Center of mass,” http://docs.opencv.org/2.4/modules/imgproc/doc/structural_analysis_and_shape_descriptors.html?highlight=cvmatchshapes#humoments, [accessed Nov. 2016]
[24]G.R. Bradski, and J.W. Davis, “Motion segmentation and pose recognition
with motion history gradients”, Machine Vision and Applications 13, pp.174-184,2002.
[25]A. Bobick , and J. Davis, “The recognition of human movement using temporal templates,” IEEE Trans. Pattern Recognit. Machine Intell, vol. 23, pp. 257-267, 2001.
[26]J. Davis. “Recognizing movement using motion histograms. ” Technical Report 487, MIT Media Laboratory, Cambridge,MA, 1999.
[27]S. Korman, D. Reichman, G. Tsur, and S. Avidan. “Fast match:Fast affine template matching. ” in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 1940-1947, 2013.
[28]T. Wu ,and A. Toet, “Speed-up Template Matching through Integral Image based Weak Classiers,” Journal of Pattern Recogn Research, vol. 9, no. 1, pp. 1-12, 2014.
[29]D.M. Gavrila. “Multi-feature hierarchical template matching using distance transforms. ” in Proc. Int. Conf. on Pattern Recognition, vol. 1, pages 439-444, 1998.
[30]“Template Matching,” http://docs.opencv.org/2.4/doc/tutorials/imgproc/histograms/template_matching/template_matching.html, [accessed Nov. 2016]
[31]“Depth-included Human Action video dataset,”
http://mclab.citi.sinica.edu.tw/dataset/dha/dha.html, [accessed May. 2016]
[32]“Centroid,” https://en.wikipedia.org/wiki/Centroid,[accessed Nov. 2016]
[33]Y.C. Lin, M.C. Hu, W.H. Cheng, Y.H. Hsieh, and H.M. Chen, “Human action recognition and retrieval using sole depth information, ”in The 20th ACM International Conference on Multimedia (MM 2012), Nara, Japan , 2012.
[34]Y.C. Lin, M.C. Hu, W.H. Cheng, Y.H. Hsieh, and H.M. Chen, “Actions speak louder than words: searching human action video based on body movement, ” in The 20th ACM International Conference on Multimedia (MM 2012) Technical Demos, Nara, Japan, 2012.