摘要： 提出一种结合常规几何参数及利用亮度、颜色四重阈值进行汽车尾灯检测的算法，以应用于夜间条件下智能汽车追尾预警系统。该算法利用亮度低阈值确定汽车尾灯的疑似区域，利用常规几何参数对上述疑似尾灯区域进行排除，利用红色比例低阈值和亮点比例阈值检测出亮度较高尾灯光斑，利用红色比例高阈值检测出亮度较低的尾灯光斑。实验结果表明，该算法可以快速、准确地检测出夜间的汽车尾灯。

关键词: 追尾预警系统, 夜间尾灯检测, 亮度阈值, 红色阈值

Abstract: This paper proposes a new vehicle taillights detection algorithm for the night view rear-end collision alarm system of smart vehicle. The algorithm combines with routine geometry parameters, makes use of four thresholds of brightness and color to detect vehicle taillights. The TB_L is used to locate the suspected taillights facular areas, the routine geometry parameters are employed to check and eliminate the suspected taillights facular areas, the TRP_L and TBP are adopted to determine the facular areas of bright taillights, the TRP_H is used to determine the facular areas of dim taillights. Experimental results show that the algorithm can detect taillights accurately and quickly.

Key words: rear-end collision alarm system, night view taillight detection, brightness threshold, red threshold

中图分类号: 

• TP391.41