本論文提出一個強健的紅綠燈偵測系統，並將此系統中基本紅綠燈偵測功能實現於Altera的FPGA板上。此系統不僅可成功偵測一般圓形的交通號誌，亦能正確辨識箭頭符號號誌燈之指示方向。首先，根據CIELab色彩空間的L亮度和a色度過濾出可能是紅綠燈的部分，再利用L亮度之方向梯度直方圖截取影像可能屬於紅綠燈的特徵點，接著採用支持向量機分類器進行分類。若判定為交通號誌燈時，系統將再根據特徵點顏色判定燈號，如顏色為紅色則顯示停止字樣；如為綠色則需再辨別是否為箭頭符號。若為箭頭符號，則根據所提出的多模識別演算法進行箭號方向辨識。多模箭號辨識演算同時考慮兩種辨識模式，根據箭號斜率及模板進行匹配，經兩者辨識分數經由決定模組投票以判定最終辨識結果。實驗結果顯示，本論文提出之紅綠燈偵測系統可達到86%之偵測率及92%之成功率。
最後，本論文將基本紅綠燈偵測核心模組實現於FPGA板上，為了節省方向梯度直方圖所耗費的硬體成本，本論文提出一個強度量化的方向梯度直方圖，在不影響偵測率及成功率的前提下有效地節省了12%的硬體花費。實驗結果顯示，本硬體架構之操作速度可達161 MHz，適合使用於需要即時運算的車用紅綠燈偵測之應用。

In this thesis, a robust traffic light detection (TLD) system is proposed and the classification kernel of red/green light detection is also realized in Altera FPGA board. The system not only successfully detects the common green and red traffic lights but also correctly recognizes the arrow green sign traffic lights. The proposed TLD system filters the valid red and green regions from CIELab color domain followed by traffic light object detection classified by support vector machine (SVM) classifier with the histogram of oriented gradient (HOG) features. If the HOG and SVM determine as a traffic light object, the system will detect if the color is green or red. If it is green color, the system should further detect whether the green light is an arrow sign or not by the proposed multimodal arrow direction algorithm, which computes both slope counting and pattern score. According to the experimental result, the proposed TLD system achieves the 86.55% detection rate and 92.41% successful rate.
Finally, the VLSI architecture of the classification kernel of the proposed TLD system is also provided. We suggest a magnitude quantization method for simple realization of HOG features. According to simulation results, the classification kernel of the proposed TLD system achieves 161 MHz operating frequency to process a full HD video in real time application.

[1] J3016_201609, “Taxonomy and definitions for terms related to driving automation systems for on-road motor vehicles,” Sep. 2016.
[2] Global Status Report on Road Safety 2015, [Online]. Available: http://www.who.int/violence_injury_prevention/road_safety_status/2015/en/.
[3] Road Traffic Accident – by Causes, [Online]. Available: https://www.npa.gov.tw/NPAGip/wSite/lp?ctNode=11500&xq_xCat=10&mp=4.
[4] Y. Zhou, Z. Chen, and X. Huang, “A system-on-chip FPGA design for real-time traffic signal recognition system,” in Proc. IEEE Int. Symp. Circuits Syst. (ISCAS), Montreal, QC, Canada, May 2016, pp. 1778–1781.
[5] M. Omachi and S. Omachi, “Detection of traffic light using structural information,” in Proc. IEEE Conf. Signal Process. (ICSP), Beijing, China, Oct. 2010, pp. 809–812.
[6] S. Sooksatra and T. Kondo, “Red traffic light detection using fast radial symmetry transform,” in Proc. 11th Int. Conf. Elect./Electron, Comput., Telecommun. Inform. Technol. (ECTI-CON), Nakhon Ratchasima, Thailand, May 2014, pp. 1–6.
[7] Y. Ji, M. Yang, Z. Lu and C. Wang, “Integrating visual selective attention model with HOG features for traffic light detection and recognition,” in Proc. Intell. Vehicles Symp. (IV), Seoul, South Korea, July 2015, pp. 280–285.
[8] J. L. Binangkit and D. H. Widyantoro, “Increasing accuracy of traffic light color detection and recognition using machine learning,” in Proc. 10th Int. Conf. Telecommun. Syst. Services Applicat. (TSSA), Denpasar, Indonesia, Oct. 2016, pp. 1–5.
[9] Trichromatic Theory and Color Spaces, [Online]. Available: http://www.drmoron.org/isblack-a-color/ .
[10] Wikipedia: HSL and HSV, [Online]. Available: https://en.wikipedia.org/wiki/HSL_and_HSV .
[11] Defining a color location using CIE L*a*b* coordinates, [Online]. Avilabe: http://the-print-guide.blogspot.tw/2010/04/tolerancing-color-in-presswork-cie-lab.html .
[12] J. Levinson, J. Askeland, and J. Dolson, “Traffic light mapping, localization, and state detection for autonomous vehicles,” in Proc. 2011 IEEE Int. Conf. Robotics Automation (ICRA), Shanghai, China, May 2011, pp. 5784–5791.
[13] N. Dalal and B. Triggs, “Histograms of oriented gradients for human detection,” in Proc. IEEE Comput. Vision Pattern Recognition (CVPR), San Diego, CA, USA, Jun. 2005, pp. 1–8.
[14] T. H. P. Tran, C. C. Pham, and T. P. Nguyen, “Real-time traffic light detection using color density,” in Proc. IEEE Int. Conf. Consumer Electron.-Asia (ICCE-Asia), Seoul, South Korea, Oct. 26–28, 2016. DOI: 10.1109/ICCE-Asia.2016.7804791
[15] Z. Shi, Z. Zou, and C. Zhang, “Real-time traffic light detection with adaptive background suppression filter,” IEEE Trans. Intell. Transportation Syst., vol. 17, pp. 690–700, Oct. 2015.
[16] F. Oniga, S. Prodan, and S. Nedevschi, “Traffic light detection on mobile devices,” in Proc. IEEE Int. Conf. Intell. Comput. Commun. Process. (ICCP), Cluj-Napoca, Romania, Sep. 2015, pp. 287–292.
[17] K. Mizuno, Y. Terachi, K. Takagi, S. Izumi, H. Kawaguchi and M. Yoshimoto, “An FPGA implementation of a HOG-based object detection processor,” IPSJ Trans. Syst. LSI Design Methodology, vol. 6, pp 42–51, Feb. 2013.
[18] P. Y. Chen, C. C. Huang, and C. Y. Lien, “An efficient hardware implementation of HOG feature extraction for human detection,” IEEE Trans. Intell. Transportation Syst., vol. 15, pp 656–662, Apr. 2014.
[19] M. P. Philipsen, M. B. Jensen, and M. M. Trivedi, “Ongoing work on traffic lights: detection and evaluation,” in Proc. IEEE Int. Conf. Advanced Video Signal Based Surveillance (AVSS), Karlsruhe, Germany, Aug. 25–28, 2015. DOI: 10.1109/AVSS.2015.7301730
[20] A. Geiger, P. Lenz, and R. Urtasun, “Are we ready for autonomous driving? the KITTI vision benchmark suite,” in Proc. IEEE Comput. Vision, Pattern Recogn. (CVPR), Providence, RI, USA, June 2012, pp 3354–3361.
[21] M. P. Philipsen, M. B. Jensen, and A. Mogelmose, “Traffic light detection: a learning algorithm and evaluations on challenging dataset,” in Proc. IEEE Int. Conf. Intell. Transportation Syst. (ITSC), Las Palmas, Spain, Sep. 2015, pp 2341–2345.
[22] ITU-T SC16/Q6, “Calculation of average PSNR differences between RD-curves,” Apr. 2001.
[23] Y. Zhou, Z. Chen, and X. Huang, “A pipeline architecture for traffic sign classification on an FPGA,” in Proc. IEEE Int. Symt. Circuits Syst. (ISCAS), Lisbon, Portugal, May 2015, pp 950–953.
[24] G. Siogkas, E. Skodras, and E. Dermatas, “Traffic lights detection in adverse conditions using color, symmetry and spatiotemporal information,” in Proc. Int. Conf. Comput. Vision Theory Applicat. (VISAPP), Rome, Italy, Feb. 2012, pp 620–627.
[25] M. Omachi and S. Omachi, “Traffic light detection with color and edge information,” in Proc. IEEE Int. Conf Comput. Science, Inform. Tech. (ICCSIT), Beijing, China, Aug. 2009, pp 284–287.
[26] R. Charatte and F. Nashashibi, “Real time visual traffic lights recognition based on spot light detection and adaptive traffic lights templates,” in Proc. IEEE Intell. Vehicles Symp., Xian, China, June 2009, pp 358–363.
[27] V. John, L. Zheming, and S. Mita, “Robust traffic light and arrow detection using optimal camera parameters and GPS-based priors,” in Proc. Asia-Pacific Conf. Intell. Robot Syst. (ACIRS), Tokyo, Japan, July 2016, pp 204–208.
[28] W. Zong and Q. Chen, “Traffic light detection based on multi-feature segmentation and online selecting scheme,” in Proc. IEEE Int. Conf. Syst., Man Cybernetics (SMC), San Diego, CA, USA, Oct. 2014, pp 204–209.
[29] G. G. Lee and B. K. Park, “Traffic light recognition using deep neural networks,” in Proc. 2017 IEEE Int. Conf. Consumer Electron., Las Vegas, USA, Jan. 8–10, 2017. DOI: 10.1109/ICCE.2017.7889317
[30] W. Liu, S. Li, J. Lv, B. Yu, T. Zhou, H. Yuan, and H. Zhao, “Real-time traffic light recognition based on smartphone platforms,” IEEE Trans. Circuits Syst. Video Tech., vol. 27, pp 1118–1131, Jan. 2016.