在人機互動的應用上最主要的角色就是人，所以賦予機器能夠偵測行人的能力即為重要，相關應用包含了互動型機器人、進階駕駛輔助系統、防盜系統與長照弱勢關懷系統。行人偵測系統是希望透過影像處理的方式，從攝影機所攝取的影像進行行人偵測，將結果資訊提供給有需要的應用上。

本行人偵測系統架構分為兩大部分，分別為積分通道影像轉換以及快速多尺度搜尋。通道影像使用YCbCr三通道色彩空間、梯度影像、以及 6 個方向梯度影像，共使用 10 個通道影像來提取行人特徵，為了達到快速提取特徵的目的，將這 10 個通道影像轉換為積分通道影像。快速多尺度搜尋有別於一般縮放影像的多尺度搜尋，改為在原始影像上使用不同大小的掃瞄視窗，節省重複計算縮放後的積分通道影像所需的處理時間，行人分類器採用AdaBoost 所訓練出的串聯式分類器，並利用經由實驗統計所產生的特徵修正公式調整行人分類器的特徵阀值，以對應在不同大小掃瞄視窗中的行人特徵。在快速多尺度搜尋後，本行人偵測系統會利用結果分群的方式，去除重複的偵測結果，並顯示整張影像中的多尺度行人偵測結果。

People are among the most important role of the Human–Machine Interaction (HCI) applications, and endowing machines with ability to detect human is very significant. The relevant applications conclude interactive robots, advanced driver assist systems (ADAS), burglarproof systems, and care for the elderly and disabled. Pedestrian detection system can detect people from the input image by camera and provide the result for the relevant applications.

The system architecture is composed of integral channel images transformation and fast multiscale detection. Channel images consist of YCbCr color channels, gradient magnitude channel and 6 quantized orientations channels. Aim at fast features extraction, we compute the integral channel images of 10 channel images. Different from rescaling the image several times in multiscale detection, fast multiscale detection rescales the detection windows to save the process time for computing the rescaled integral channel images. Our pedestrian detector is based on a cascade classifier applying AdaBoost algorithm, and can adjust the integral channel features’ thresholds with the empirical approximation function for different scale pedestrian. After fast multiscale detection, our system applies result clustering to reduce multiple detection results and demonstrates the multiscale pedestrian detection results.

[1] N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” Computer Vision and Pattern Recognition, vol. 1, pp. 886-893, 2005.
[2] X. Wang, T.X. Han, and S. Yan, “An HOG-LBP human detector with partial occlusion handling,” International Conference on Computer Vision, pp. 32-39, 2009.
[3] Y. Pang, Y. Yuan, X. Li, and J. Pan, “Efficient HOG human detection,” Signal Processing, vol. 91, no. 4, pp. 773–781, 2011.
[4] C. Hou, H. Ai, and S. Lao, “Multiview pedestrian detection based on vector boosting,” Asian Conference on Computer Vision, vol. 1, pp. 210-219, 2007.
[5] C. Keller, M. Enzweiler, M. Rohrbach, D. F. Llorca, C. Schnorr, and D.M. Gavrila, “The benefits of dense stereo for pedestrian recognition,” IEEE Transactions on Intelligent Transportation Systems, vol. 12, no. 4, pp. 1096-1106, 2011.
[6] D. Park, C. L. Zitnick, D. Ramanan, and P. Dollar, “Exploring weak stabilization for motion feature extraction,” Computer Vision and Pattern Recognition, pp. 2882-2889, 2013.
[7] C. Premebida, J. Carreira, J. Batista, and U. Nunes, “Pedestrian detection combining RGB and dense LIDAR data,” International Conference on Intelligent Robots and Systems, pp. 4112-4117, 2014.
[8] C. Wojek and B. Schiele, “A performance evaluation of single and multi-feature people detection,” 30th DAGM Symposium Pattern Recognition, pp. 82-91, 2008.
[9] F. Khan, R. Anwer, J. van de Weijer, A. Bagdanov, M. Vanrell, and A. Lopez, “Color attributes for object detection,” Computer Vision and Pattern Recognition, pp. 3306-3313, 2012.
[10] X. Wang, M. Yang, S. Zhu, and Y. Lin, “Regionlets for generic object detection,” International Conference on Computer Vision, vol. 37, pp. 2071-2084, 2013.
[11] P. Luo, Y. Tian, X. Wang, and X. Tang, “Switchable deep network for pedestrian detection,” Computer Vision and Pattern Recognition, pp. 899-906, 2014.
[12] P. Viola and M. Jones, “Robust real-time face detection,” International Journal of Computer Vision, vol. 57, no. 2, pp. 137-154, 2004.
[13] P. Viola and M. Jones, “Rapid object detection using a boosted cascade of simple features,” Computer Vision and Pattern Recognition, vol. 1, pp. 511-518, 2001.
[14] P. Dollár, Z. Tu, P. Perona and S. Belongie, “Integral channel features,” British Machine Vision Conference, pp. 91.1-91.11, 2009.
[15] R.-E. Fan, P.-H. Chen and C.-J. Lin, “Working set selection using second order information for training support vector machines,” Journal of Machine Learning Research, vol. 6, pp. 1889-1918, 2005.
[16] P.-H. Chen, R.-E. Fan, and C.-J. Lin, “A study on SMO-type decomposition methods for support vector machines,” IEEE Transactions on Neural Networks, vol. 17, pp. 893-908, July 2006.
[17] B. Leo, J. Friedman, R. Olshen, C. Stone, “Classification and regression trees,” Belmont, California: Wadsworth, 1984.
[18] C. Zhang and P. Viola, “Multiple-instance pruning for learning efficient cascade detectors,” Conference on Neural Information Processing Systems, pp. 1681-1688 2007.
[19] P. Dollár, S. Belongie and P. Perona, “The fastest pedestrian detector in the west,” British Machine Vision Conference, pp. 68.1-68.11, 2010.
[20] NICTA, NICTA pedestrian dataset, Available:
https://www.nicta.com.au/category/research/computer-vision/tools/automap-datasets-and-code/
[21] Elektra, CVC-02: Onboard sequence pedestrian dataset 2, Available:
http://adas.cvc.uab.es/site/elektra/
[22] EPFL, Multi-camera pedestrians video, Available: http://cvlab.epfl.ch/data/pom/