A robust video based traffic light detection algorithm for intelligent vehicles

Recently, researches on intelligent vehicles which can drive in urban environment autonomously become more popular. Traffic lights are common in cities and are important cues for the path planning of intelligent vehicles. In this paper, a robust and efficient algorithm to detect traffic lights based on video sequences captured by a low cost off-the-shelf video camera is proposed. The algorithm models the hue and saturation according to Gaussian distributions and learns their parameters with training images. From learned models, candidate regions of the traffic lights in the test images can be extracted. Post processing method which takes account of the shape information is applied to the candidate regions. Furthermore, detection results of the previous image frames are aggregated in order to provide a more robust result. Experimental results on several video sequences captured in typical urban environment prove the effectiveness of the proposed algorithm.