行人偵測在電腦視覺領域中，是一項非常重要的研究。許多的應用與行人偵測相關，例如：室內外監控系統、先進駕駛輔助系統、行動機器人…等，從簡單的計算人數任務到困難的行人防撞系統。然而，要在複雜的環境中偵測行人，仍然是一項具有挑戰性的任務。因為，行人具有多變的外貌及顏色，並且受到光照條件及遮蔽的影響。所幸，隨著越來越多的特徵表示法及機器學習的演算法被導入電腦視覺的領域，行人偵測的效能及準確率得到了顯著地改善。
本篇論文提出一個基於機器視覺的即時多行人偵測與追蹤之技術，而該技術藉由具有處理遮蔽問題的追蹤方法來提升偵測率。首先，我們採用兩階段偵測的方法，將行人從拍攝的影像中偵測出來。該偵測方法結合了自適應增強（Adaptive Boosting）和支援向量機（Support Vector Machine）這兩種機器學習的演算法。在第一階段，使用基於哈爾(Haar-like)特徵的級聯式分類器，迅速地將候選行人提取出來；然後，透過方向梯度直方圖（Histogram of Oriented Gradients）分類器來驗證候選行人，以減少假陽性的數目。接著，為了提升偵測率和處理遮蔽的問題，我們利用卡爾曼濾波器來預測目標接下來的位置，並且使用模板匹配在一個範圍內，找出目標正確的位置。另外，我們提出的技術可以處理非固定式視角的單攝影機所拍攝之影像。
我們針對不同的場景進行實驗，例如：室內走廊及室外走道。另外，實驗影片中也包含各種行人被遮蔽的狀況，例如：部份遮蔽和完全遮蔽；而遮蔽物包含動、靜態場景和其他行人。我們提出的方法可以有效地偵測行人，其平均偵測率大約為75.5%，而加入追蹤演算法後，平均偵測率提升至大約96.1%。整體平均執行效率大約每秒14.8至31.3個影格。

Pedestrian Detection is a very important research in Computer Vision field. Many applications are associated with pedestrian detection, such as indoor/outdoor video surveillance system, advanced driver assistance systems (ADAS), and mobile robots. And some tasks like humans count and pedestrian collision avoidance system are also included. However, detecting pedestrian in complex environments is still a challenging task, because pedestrians have their appearance varied according to the clothes color, and they are also affected by lighting conditions and occlusion. Fortunately, since more and more feature representations and machine learning algorithms are introduced into computer vision field, the performance and accurate rate of pedestrian detection have been significantly improved.
In this thesis, a computer vision based real-time multi-pedestrian detection and tracking technique is proposed, and the technique increases the detection rate by tracking methods with occlusion handling. At first, we employ the two-stage detection method to detect pedestrians from the video sequences. The detection method combines the Adaptive Boosting with Support Vector Machine of machine learning algorithms. In the first stage, use a cascade classifier based on Haar-like features to extract candidates rapidly; and then validate those candidates through a Histogram of Oriented Gradients classifier to reduce the number of false positives. After that, for increasing the detection rate and handling occlusion problems, we utilize Kalman filter to predict the next locations of targets, then use template matching within limited region to find the correctly location of targets. In addition, the technique we proposed can handle the video sequences which are captured by monocular camera with non-fixed viewpoint.
We conducted experiments for different scenarios, such as indoor corridor or outdoor walkway. In addition, the experimental videos include variety of situations of pedestrian occlusion, such as partially or fully occlusion; and the occluded objects include scenes or other pedestrians. Our proposed method can effectively detect pedestrians, and the average detection rate is about 75.5%. After tracking algorithms are implemented, the average detection rate is increased to about 96.1%. The average performance of overall is about 14.9 to 31.3 fps (frames per second).

[1]M. Correa, G. Hermosilla, R. Verschae and J. Ruiz-del-Solar, “Human Detection and Identification by Robots Using Thermal and Visual Information in Domestic Environments,” Journal of Intelligent & Robotic Systems, vol. 66, no. 1-2, pp. 223-243, 2012.
[2]K. Yang, E. Y. Du, P. Jiang, Y. Chen, R. Sherony, and H. Takahashi, “Automatic Categorization-based Multi-stage Pedestrian Detection,” in Proceedings of 2012 15th International IEEE Conference on Intelligent Transportation Systems, Anchorage, Alaska, U.S.A., pp. 451-456, 2012.
[3]M. Enzweiler and D. M. Gavrila, “Monocular Pedestrian Detection: Survey and Experiments,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 31, no. 12, pp. 2179-2195, 2009.
[4]Q. Ye, J. Liang and J. Jiao, “Pedestrian Detection in Video Images via Error Correcting Output Code Classification of Manifold Subclasses,” IEEE Transactions on Intelligent Transportation Systems, vol. 13, no. 1, pp. 193-202, 2012.
[5]P. Dollar, C. Wojek, B. Schiele, and P. Perona, “Pedestrian Detection: An Evaluation of the State of the Art,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 34, no. 4, pp. 743-761, 2012.
[6]C. G. Keller, D. Thao, H. Fritz, A. Joos, C. Rabe, and D. M. Gavrila, ”Active Pedestrian Safety by Automatic Braking and Evasive Steering,” IEEE Transactions on Intelligent Transportation Systems, vol. 12, no. 4, pp. 1292-1304, 2011.
[7]S. Aly, L. Hassan, A. Sagheer, and H. Murase, “Partially Occluded Pedestrian Classification using Part-based Classifiers and Restricted Boltzmann Machine Model,” Proceedings of the 16th International IEEE Annual Conference on Intelligent Transportation Systems, Hague, Netherlands, pp. 1065-1070, 2013.
[8]D. Geronimo, A. M. Lopez, A. D. Sappa, and T. Graf, “Survey of Pedestrian Detection for Advanced Driver Assistance Systems,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 32, no. 7, pp. 1239-1258, 2010.
[9]P. Viola, M. J. Jones, and D. Snow, “Detecting pedestrians using patterns of motion and appearance,” International Journal of Computer Vision, vol. 63, no. 2, pp. 153-161, 2005.
[10]N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” in Proceedings of 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Diego, California, U.S.A., vol. 1, pp. 886-893, 2005.
[11]D. Geronimo, A. Sappa, A. Lopez, and D. Ponsa, “Adaptive Image Sampling and Windows Classification for On-board Pedestrian Detection,” in Proceedings of the 5th International Conference on Computer Vision Systems, Bielefeld, Germany, 2007.
[12]L. Zhang, B. Wu, and R. Nevatia, “Pedestrian detection in infrared images based on local shape features,” IEEE Conference on Computer Vision and Pattern Recognition, Minneapolis, Minnesota, U.S.A., pp. 1-8, 2007.
[13]M. Szarvas, A. Yoshizawa, M. Yamamoto, and J. Ogata, “Pedestrian detection with convolutional neural networks,” Proceedings of 2005 IEEE Conference on Intelligent Vehicles Symposium, pp. 224-229, 2005.
[14]Q. J. Wang and R. B. Zhang, “LPP-HOG: A New Local Image Descriptor for Fast Human Detection,” IEEE International Symposium on Knowledge Acquisition and Modeling Workshop, Wuhan, China, pp. 640-643, 2008.
[15]X. Wang, T. X. Han, and S. Yan, “An HOG-LBP human detector with partial occlusion handling,” 2009 IEEE 12th International Conference on Computer Vision, Kyoto, Japan, pp. 32-39, 2009.
[16]K. Yang, E.Y. Du, E.J. Delp, P. Jiang, F. Jiang, Y. Chen, R. Sherony, and H. Takahashi, “An Extreme Learning Machine-based Pedestrian Detection Method,” 2013 IEEE Conference on Intelligent Vehicles Symposium, Gold Coast, Queensland, Australia, pp. 1404-1409, 2013.
[17]P. Geismann and G. Schneider, “A Two-staged Approach to Vision-based Pedestrian Recognition Using Haar and HOG Features,” 2008 IEEE Conference on Intelligent Vehicles Symposium, Eindhoven, Netherlands, pp. 554-559, 2008.
[18]A. Mogelmose, A. Prioletti, M. M. Trivedi, A. Broggi, and T. B. Moeslund, “A two-stage part-based pedestrian detection system using monocular vision,” Proceedings of the 15th IEEE International Conference on Intelligent Transportation Systems, Anchorage, Alaska, U.S.A., pp. 73-77, 2012.
[19]S. Avidan, “Ensemble tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 29, no. 2, pp. 261-271, 2007.
[20]P. Cui, L. F. Sun, F. Wang, and S. Q. Yang “Contextual Mixture Tracking,” IEEE Transactions on Multimedia, vol. 11, no. 2, pp. 333-341, 2009.
[21]G. D. Caterina and J. J. Soraghan, “Robust complete occlusion handling in adaptive template matching target tracking,” Electronics Letters, vol. 48, no. 14, pp. 831-832, 2012.
[22]C. Zhang , J. Xu, A. Beaugendre, and S. Goto, “A KLT-Based Approach for Occlusion Handling in Human Tracking,” Picture Coding Symposium, Krakow, Poland, pp. 337-340, 2012.
[23]C. P. Papageorgiou, M. Oren, and T. Poggio, “A General Framework for Object Detection,” in Proceedings of 1998 6th International Conference on Computer Vision, Bombay, India, pp. 555-562, 1998.
[24]P. Viola and M. Jones, “Robust Real-time Object Detection,” International Journal of Computer Vision, 2001.
[25]Y. Freund and R. E. Schapire, “A Decision-Theoretic Generalization of On-Line Learning and an Application to Boosting,” Journal of Computer and System Sciences, Orlando, Florida, U.S.A., vol. 55, no. 1, pp. 119-139, 1997.
[26]W. T. Freeman and M. Roth, “Orientation histograms for hand gesture recognition,” Mitsubishi Electric Research Laboratory Report, 1994.
[27]W.T. Freeman, K. Tanaka, J. Ohta, and K. Kyuma, “Computer vision for computer games,” Proceedings of the 2nd International Conference on Automatic Face and Gesture Recognition, Killington, Vermont, U.S.A., pp. 100-105, 1996.
[28]D. G. Lowe , “Distinctive Image Features from Scale-Invariant Keypoints,” International Journal of Computer Vision, Kluwer Academic Publishers Hingham, Massachusetts, U.S.A., vol. 60, no. 2, pp. 91-110, 2004.
[29]S. Belongie, J. Malik, and J. Puzicha, “Matching Shapes,” in Proceedings of 8th IEEE International Conference on Computer Vision, Vancouver, British Columbia, Canada, vol. 1, pp. 454-461, 2001.
[30]C. Cortes and V. Vapnik, “Support-Vector Networks,” Machine Learning, Kluwer Academic Publishers Hingham, Massachusetts, U.S.A., vol. 20, no. 3, pp. 273-297, 1995.
[31]“Support vector machine,“ [Online] Available: http://en.wikipedia.org/wiki/Support_vector_machine (accessed on September 30, 2014).
[32]I. El-Naqa, Y. Yang, M. N. Wernick, N. P. Galatsanos, and R. M. Nishikawa, “A Support Vector Machine Approach for Detection of Microcalcifications,” IEEE Transactions on Medical Imaging, vol. 21, no. 12, pp. 1552-1563, 2002.
[33]D. Comaniciu, V. Ramesh, and P. Meer, “Real-Time Tracking of Non-Rigid Objects using Mean Shift,” Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Hilton Head Island, South Carolina, U.S.A., vol. 2, pp. 142-149, 2000.
[34]M. S. Khalid, M. U. Ilyas, M. S. Sarfaraz, and M. A. Ajaz, “Bhattacharyya Coefficient in Correlation of Gray-Scale Objects,” Proceedings of the 4th International Symposium on Image and Signal Processing and Analysis, pp. 209-214, 2005.
[35]G. Welch and G. Bishop, “An Introduction to the Kalman Filter,” Technical Report, Department of Computer Science, University of North Carolina at Chapel Hill, Chapel Hill, North Carolina, U.S.A., 2006.
[36]“Template matching,“ [Online] Available: http://en.wikipedia.org/wiki/Template_matching (accessed on October 15, 2014).
[37]D. A. Klein, D. Schulz, S. Frintrop, and A. B. Cremers, “Adaptive Real-Time Video-Tracking for Arbitrary Objects,” IEEE/RSJ International Conference on Intelligent Robots and Systems, Taipei, Taiwan, pp. 772-777, 2010.
[38]H. Wu, X. Chen, Y. Gao, H. Zhou, and X. Zhang, “An Effective Algorithm of Tracking Multiple Objects in Occlusion Scenes,” in Proceedings of 2010 2nd International Conference on Industrial Mechatronics and Automation, Wuhan, China, vol. 2, pp. 409-413, 2010.
[39]S. Gidel, P. Checchin, C. Blanc, T. Chateau, and L. Trassoudaine, “Pedestrian Detection and Tracking in an Urban Environment Using a Multilayer Laser Scanner,” IEEE Transactions on Intelligent Transportation Systems, vol. 11, no. 3, pp. 579-588, 2010.
[40]A. Prioletti, A. Mogelmose, P. Grisleri, M. M. Trivedi, A. Broggi, and T. B. Moeslund, “Part-Based Pedestrian Detection and Feature-Based Tracking for Driver Assistance Real-Time, Robust Algorithms, and Evaluation,” IEEE Transactions on Intelligent Transportation Systems, vol. 14, no. 3, pp. 1346-1359, 2013.