Detection of Traffic Lights for Vision-Based Car Navigation System

A recent trend of car navigation system is using actual video captured by camera equipped on a vehicle. The video-based navigation systems displays guidance information overlaid onto video before reaching a crossroad, so it is essential to detect where the crossroads are in the video frame. In this paper, we suggest a detection method for traffic lights that is used for estimating location of crossroads in image. Suggested method can detect traffic lights in a long distance, and estimates pixel location of crossroad that is important information to visually represent guidance information on video. We suggest a new method for traffic light detection that processes color thresholding, finds center of traffic light by Gaussian mask, and verifies the candidate of traffic light using suggested existence-weight map. Experiments show that the detection method for traffic signs works effectively and robustly for outdoor video and can used for video-based navigation system.