Traffic flow detection based on the rear-lamp and virtual coil for nighttime conditions

Traffic flow detection is an important topic in intelligent transportation systems (ITS). At present, most of the traffic flow detection method for nighttime are based on the headlights. However, headlights are not available in bad weather while the vehicle taillights have more stable features such as colors. so that in this paper, a novel method based on vehicle rear lamp and virtual coil is presented. Firstly, the Hough transform is used to extract the lane to set up the virtual coil in which the rear lamp objects can be extracted. Secondly, a color threshold segmentation algorithm is used to extract the suspected taillights areas in RGB color space. Then, the threshold of brightness is determined automatically by the maximum variance method to separate the rear lamp and label it by minimum bounding rectangle(MBR). Finally, the detection is accomplished by matching and tracking the rear lamps in the virtual coil. Experimental results show that the algorithm can statistic traffic flow accurately and quickly in various weather conditions.