Traffic Light Recognition

• Main Authors: Kuo-Hao Lu, 呂國豪
• Other Authors: Shu-Yuan Chen
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/67328113508308738839

碩士 === 元智大學 === 資訊工程學系 === 93 === Advanced technology improves the capabilities of modern vehicles. The innovations of senor-based systems support surrounding survey of the vehicle and display relevant information to the driver. Thus, construction of a safety, convenient and efficiency driving environment can be achieved. This paper is to propose an automatic traffic light recognition system so that car drivers have sufficient information to make a correct decision. This in turn facilitates the construction of ITS (Intelligent Transportation System). The proposed method can be applied not only to fixed camera but also to movable camera. Our method consists of three phases: traffic light detection, extraction and classification. The three phases are based on color information, region information and geometric and appearance constraints. In the detection stage, the RGB color space is first converted into HSI color space to detect those regions with specific colors of traffic lights. The morphology technology is then employed to remove hole and noise. In the extraction stage, region labeling is involved to detect candidate regions of traffic lights. Border detection is then employed to obtain region border. In the classification stage, geometric and appearance constrains are derived respectively from traffic light shape and color and used for classification. In this study, traffic lights of circle and arrow shape can both be coped with. Various experiments have been conducted to demonstrate the effectiveness and practicability of proposed method.