即時性的行人偵測是電腦視覺領域其中一個最具有挑戰性的項目。此項目中最為重要而且最為困難的部份是如何在偵測的同時兼顧準確率和執行速度，這也是本論文所設計的偵測框架欲達到的目標。本論文提出了一種複合性、以輪廓特徵為偵測基礎的行人偵測器，偵測器基於兩種輪廓性質的特徵: "Census Transform Histogram" (CENTRIST) 特徵，和 "dominant orientation templates" (DOT) 特徵，本論文的偵測框架會在數位影像畫面中提取影像的這兩種特徵和事先訓練好的行人特徵分類器進行比較，以辨識出行人的存在。本論文提出的偵測框架基於上述兩種低計算時間與資源耗時的特徵，和低計算複雜度的行人特徵分類器訓練流程，因此具備了即時性的優點。此外，本論文提出的偵測框架不需要影像來源的額外三維深度資訊，事先定義的感興趣領域 (region of interest)，或是連續影像的動態資訊，採用了基礎的單張畫面處理 (frame-by-frame) 的流程。此偵測框架目前已經在 INRIA, ETH, TUD-Brussels 公開的影片資料集中進行實測，方法偵測結果達到了可接受程度的準確度並且可以即時性運行。

Real-time pedestrian detection is one of the most challenging topics in the field of computer vision. The most important and yet difficult issue is achieving accuracy and detection speed simultaneously. Accordingly, the pedestrian detection framework proposed in this paper is aimed at this goal. The proposed compound contour detector is based on two features: Census Transform Histogram (CENTRIST) feature, and dominant orientation templates (DOT) feature. They are used as the main extracted pedestrian features in a real-time detection framework. A low computational complexity training process speed up the feature extraction algorithm, and a combination of the above mentioned two features is applied to a real-time pedestrian detection system architecture. The proposed method does not require additional three-dimensional depth information, region of interest, or momentum information for background subtraction. It employs a conventional frame-by-frame processing method. It was applied to the TUD-Brussels, INRIA, and ETH public data sets on a standard PC and achieved acceptable accuracy. The results show that the proposed method is feasible and can be applied to real-time pedestrian detection.

Abstract i
Acknowledgements ii
Contents iii
1 Introduction 1
1.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Objective . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature Survey 5
3 The Proposed Compound Contour Detector Framework 8
3.1 Framework Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Pedestrian Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.1 Edge Extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.2.2 Census Transform Histogram Feature . . . . . . . . . . . . . . . . . . 13
3.2.3 Dominant Orientation Templates Feature . . . . . . . . . . . . . . . . 17
iii
3.3 Pedestrian Feature Classier and Non-maximum Suppression Algorithm . . . 19
3.3.1 Support Vector Machine Classier . . . . . . . . . . . . . . . . . . . . 19
3.3.2 Histogram Intersection Kernel Support Vector Machine . . . . . . . . 22
3.3.3 Nonmaximum Suppression . . . . . . . . . . . . . . . . . . . . . . . . 24
4 Experimental Data Sets and Experimental Results 27
4.1 Experimental Data Sets Overview . . . . . . . . . . . . . . . . . . . . . . . . 27
4.2 Training Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
4.3 Receiver Operating Characteristic Curve . . . . . . . . . . . . . . . . . . . . 30
4.4 Experimental Results of Pedestrian Detection . . . . . . . . . . . . . . . . . 33
4.5 Methods Comparision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
5 Conclusion and Future Work 41
5.1 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .44

Bibliography
[1] N. Dalal and B. Triggs, \"Histogram of oriented gradient for human detection"," IEEE
Computer Society Conference on Computer Vision and Pattern Recognition, vol. 1,
pp. 886{893, June 2005.
[2] T. Watanabe, S. Ito, and K. Yokoi, \"Co-occurrence Histograms of Oriented Gradients
for Pedestrian Detection"," Advances in Image and Video Technology Lecture Notes in
Computer Science, vol. 5414, pp. 37{47, 2009.
[3] Y. Bai, L. Guo, and L. Jin, \"A novel feature extraction method using Pyramid His-
togram of Orientation Gradients for smile recognition"," 16th IEEE International Con-
ference on Image Processing, pp. 3305{3308, November 2009.
[4] Y. Goto, Y. Yamauchi, and H. Fujiyoshi, \"CS-HOG: Color similarity-based HOG","
19th Korea-Japan Joint Workshop on Frontiers of Computer Vision, pp. 266{271,
February 2013.
[5] X. Wang, T. Han, and S. Yan, \"An HOG-LBP Human Detector with Partial Occlu-
sion Handling, Computer Vision"," IEEE 12th International Conference on Computer
Vision, pp. 32{39, September 2009.
[6] N. D. Kumar and K. Suresh, \"HOG-PCA descriptor with optical
ow based human
detection and tracking"," International Conference on Communications and Signal Pro-
cessing, pp. 900{904, April 2014.
[7] D. Sun and J. Watada, \"Detecting pedestrians and vehicles in trac scene based on
boosted HOG features and SVM"," 9th IEEE International Symposium on Intelligent
Signal Processing, pp. 1{4, May 2015.
[8] S. Madhogaria, P. Baggenstoss, M. Schikora, W. Koch, and D. Cremers, \"Car detection
by fusion of HOG and causal MRF"," IEEE Transactions on Aerospace and Electronic
Systems, vol. 51, pp. 575{590, January 2015.
[9] P. Dollar, Z. Tu, P. Perona, and S. Belongie, \"Integral Channel Features"," in Pro-
ceedings of the British Machine Vision Conference, pp. 91.1{91.11, BMVA Press, 2009.
doi:10.5244/C.23.91.
[10] P. Felzenszwalb, R. Girshick, D. McAllester, and D. Ramanan, \"Object Detection with
Discriminatively Trained Part-Based Models"," IEEE Transactions on Pattern Analysis
and Machine Intelligence, vol. 32, pp. 1627{1645, September 2010.
[11] J. Gall, A. Yao, N. Razavi, L. V. Gool, and V. Lempitsky, \"Hough Forests for Object
Detection, Tracking, and Action Recognition"," IEEE Transactions on Pattern Analysis
and Machine Intelligence, vol. 33, no. 11, pp. 2188{2202, 2011.
[12] A. Lehmann, B. Leibe, and L. V. Gool, \"Fast PRISM: Branch and Bound Hough
Transform for Object Class Detection"," International Journal of Computer Vision,
vol. 94, pp. 175{197, June 2011.
[13] X. Zeng, W. Ouyang, and X. Wang, \"Multi-stage Contextual Deep Learning for Pedes-
trian Detection"," IEEE International Conference on Computer Vision, pp. 121{128,
December 2013.
[14] H. Yoshida, D. Suzuo, D. Deguchi, I. Ide, H. Murase, T. Machida, and Y. Kojima,
\"Pedestrian detection by scene dependent classiers with generative learning"," IEEE
on Intelligent Vehicles Symposium, pp. 654{659, June 2013.
[15] H. Zheng, C. Tian, and W. Wei, \"Visual tracking based on the color attention pre-
served sparse generative object model"," IEEE International Conference on Orange
Technologies, pp. 1{4, September 2014.
[16] L. Dong, K. Weng, C. Wang, G. Liang, H. Chen, and X. Wu, \"A 3D object recognition
and pose estimation system using deep learning method"," 4th IEEE International
Conference on Information Science and Technology, pp. 401{404, April 2014.
[17] Z. Zhu, X. Wang, S. Bai, C. Yao, and X. Bai, \"Deep learning representation using
autoencoder for 3D shape retrieval"," International Conference on Security, Pattern
Analysis, and Cybernetics, pp. 279{284, October 2014.
[18] J. Wu and J. Rehg, \"CENTRIST:A Visual Descriptor for Scene Categorization","
IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, pp. 1489{
1501, August 2010.
[19] J.Wu, C. Geyer, and J. Rehg, \"Real-time human detection using contour cues"," IEEE
International Conference on Robotics and Automation, pp. 860{867, 2011.
[20] J. Wu and J. Rehg, \"C4 : A Real-Time Objectdetection Framework"," IEEE Trans-
actions on Image Processing, vol. 22, pp. 4096{4107, October 2013.
[21] S. Hinterstoisser, V. Lepetit, S. Ilic, P. Fua, and N. Navab, \"Dominant orientation tem-
plates for real-time detection of texture-less objects"," IEEE Conference on Computer
Vision and Pattern Recognition, pp. 2257{2264, June 2010.
[22] C. Cortes and V. Vapnik, \"Support-vector networks"," Machine Learning, vol. 20,
pp. 273{297, September 1995.
[23] S. Maji, A. Berg, and J. Malik, \"Classication using intersection kernel support vector
machines is ecient"," IEEE Conference on Computer Vision and Pattern Recognition,
pp. 1{8, June 2008.
[24] J. Wu, \"Ecient HIK SVM learning for image classication"," IEEE Transactions on
Image Process, vol. 21, pp. 1489{1501, October 2012.
[25] A. Neubeck and L. Gool, \"Ecient Non-Maximum Suppression"," 18th International
Conference on Pattern Recognition, vol. 3, pp. 850{855, 2006.
[26] \INRIA Person Dataset." http://pascal.inrialpes.fr/data/human/.
[27] \Robust Multi-Person Tracking from Mobile Platforms." https://data.vision.ee.
ethz.ch/cvl/aess/dataset/.
[28] \Multi-Cue Onboard Pedestrian Detection." https://www.mpi-inf.
mpg.de/departments/computer-vision-and-multimodal-computing/
research/people-detection-pose-estimation-and-tracking/
multi-cue-onboard-pedestrian-detection/.
[29] P. Dollar, C.Wojek, B. Schiele, and P. Perona, \"Pedestrian detection: A benchmark","
IEEE Conference on Computer Vision and Pattern Recognition, pp. 304{311, June 2009.
[30] \Caltech Pedestrian Detection Benchmark." http://www.vision.caltech.edu/
Image_Datasets/CaltechPedestrians/.
[31] P. Viola and M. Jones, \"Robust Real-Time Face Detection"," International Journal
of Computer Vision, vol. 57, pp. 137{154, May 2004.
[32] P. Felzenszwalb, D. McAllester, and D. Ramanan, \"A discriminatively trained, mul-
tiscale, deformable part model"," IEEE Conference on Computer Vision and Pattern
Recognition, pp. 1{8, June 2008.
[33] S. Paisitkriangkrai, C. Shen, and A. van den Hengel, \"Ecient pedestrian detection
by directly optimize the partial area under the ROC curve"," IEEE International Con-
ference on Computer Vision, pp. 1057{1064, December 2013.
[34] R. Benenson, M. Mathias, R. Timofte, and L. V. Gool, \"Pedestrian detection at 100
frames per second"," IEEE Conference on Computer Vision and Pattern Recognition,
pp. 2903{2910, June 2012.
[35] R. Benenson, M. Mathias, T. Tuytelaars, and L. V. Gool, \"Seeking the Strongest Rigid
Detector"," IEEE Conference on Computer Vision and Pattern Recognition, pp. 3666{
3673, June 2013.
[36] S. Paisitkriangkrai, C. Shen, and A. van den Hengel, \"Strengthening the Eectiveness
of Pedestrian Detection with Spatially Pooled Features"," European Conference on
Computer Vision, vol. 8692, pp. 546{561, 2014.
[37] B. S. C. Wojek, \"A Performance Evaluation of Single and Multi-feature People Detec-
tion"," Pattern Recognition, vol. 5096, pp. 83{91, June 2008.
[38] \PASCAL2007." http://host.robots.ox.ac.uk/pascal/VOC/voc2007/index.html.