Abstract: To solve the auto-recognition problem of multi-type traffic lights in complex background, a novel method based on brightness value division,K-means clustering and histogram analysis of foreground objects was proposed. Firstly, self-localization was realized by brightness value division, geometric features analysis and classification statistics. Along with signal color judgment by K-means clustering algorithm, the type and direction information of traffic lights would be acquired by analyzing the histograms of foreground objects; accordingly auto-detection was realized. Experimental results showed that the proposed method had high detection accuracy for multi-type traffic lights in different scenarios. It was proved that this method was highly reliable and widely applicable.

Key words: self-localization, traffic light recognition, brightness value division, K-means clustering, foreground object histogram