Vehicle Detection at Night Based on the Feature of Taillight and License Plate

This paper proposes a algorithm of vehicle feature extraction and detection based on video data for night time. The color characteristics of taillights can be roughly divided into two parts no matter how far or near they are: inner ring-highlights area partial to pink and outer ring-high saturation Red areas. Through a large number of sampling statistics, this method obtains the accurate threshold range of each layer based on HSV color space. Thus, the suspected area of the inner and outer ring of tail lights can be segmented accurately and filtered preliminarily according to the shape characteristics of the tail lamp. In order to improve the detection rate and image recognition quality, the paper carry out AOI region segmentation and median filtering. After getting the suspected area of license plate, the tail light and license plate are combined to determine the rear of the vehicle. Secondly, all the connected regions of the tail lamp suspected area are paired and the confidence level of the pair is established. The confidence level is evaluated according to the characteristics of the tail lamp pair such as the horizontal height, the distance width and the symmetry centered on the license plate. According to the confidence level, whether it is qualified to pair with the license plate is determined Finally, according to the characteristics of taillight pairs, the mismatched relationship pairs are eliminated and the vehicles are identified. The experimental results show that the method can accurately detect the vehicle tail light features to identify the vehicle, and the false detection rate is low.