基於行車紀錄器影像之號誌燈辨識系統

A Drive Recorder Video Based Traffic Light Detection System

林奇葳(Chi-Wei Lin)

指導教授 ： 郭文嘉

元智大學/資訊學院/資訊管理學系/碩士(2013年)

https://doi.org/10.6838/YZU.2013.00344

摘要

本研究使用行車紀錄器影像為辨識目標，提出了一套號誌燈偵測與辨識系統，並且加入號誌燈對於駕駛人的重要性判斷功能，旨在提供駕駛人正確且有用的號誌燈資訊。本研究使用雙邊模糊化配合Canny邊緣偵測先過濾掉部分號誌燈不會出現的區塊，接著在號誌燈的辨識系統中所採用的Normalized RGB色彩空間可以準確判斷影像中的顏色，並搭配亮度資訊做辨識，以利找出亮起的號誌燈區域，再對候選區域進行號誌燈箱的特徵偵測，以刪除不屬於號誌燈的雜訊，最後則是對得到的號誌燈進行重要性的判斷。本研究未來除了可以結合行車紀錄器提供駕駛人更多資訊以外，也可以應用在車輛自動駕駛系統中或是配合車牌辨識系統整合成闖紅燈偵測系統。

關鍵字

行車紀錄器 ； 號誌燈偵測與辨識 ； Normalized RGB ； 亮度 ； 號誌燈箱偵測

並列摘要

In this study, we proposed a drive recorder video based traffic light detection and recognition system, besides for providing more accurate information to the driver of the traffic lights, the system also decides the importance of each traffic light. The system eliminates a part of input frame that would not contain any traffic light by using bilateral filtering and Canny edge detection. Then we have normalized RGB to find out the colors of the input frame, with the aid of lightness information we can decide the possible traffic light area. Traffic signal box detection then examines the candidates to distinguish real traffic lights from noises. At last, the system would decide the importance of each traffic light to the driver. This system may combine with a drive recorder to provide more useful information to the driver, furthermore, it is possible to be a part of an auto-driving system or combine with a vehicle license plate recognition system to become a red light running detection system.

並列關鍵字

Drive recorder ； Traffic light detection and recognition ； Normalized RGB ； Lightness ； Traffic signal box detection

參考文獻

[11] 王振興， “多標的汽機車車牌辨識系統之研究”，元智大學碩士論文，2003.

[10] 魏銪志， “動態多標的車牌辨識系統之研究” ，元智大學碩士論文，2000.

[2] T.H. Hwang , I.H. Joo, S.I. Cho,“Detection of traffic lights for vision-based car navigation system”, Springer Advances in Image and Video Technology, pp. 682-691, 2006.

[3]J. Levinson, J. Askeland, J. Dolson, S. Thrun, “Traffic light mapping, localization, and state detection for autonomous vehicles”, IEEE Robotics and Automation, pp.5784-5791, 2011.

[4] M. Omachi, S. Omachi, “Traffic light detection with color and edge information”, IEEE Computer Science and Information Technology, pp.284-287, 2009.