点击次数：

论文类型：期刊论文

发表时间：2012-12-01

发表刊物：Journal of Beijing Institute of Technology (English Edition)

收录刊物：EI、Scopus

卷号：21

期号：SUPPL.2

页面范围：1-5

ISSN号：10040579

摘要：Overtaking vehicle detection is an important component of advanced driver assistance system because its speed is usually higher than that of the host vehicle. Failure to detect the overtaking vehicle will result in serious collision accident. The variations of the third order cumulant in the monitoring windows between two consequent images were analyzed to judge the emerging of overtaking vehicles. Then the two-dimension maximum entropy segmentation method and multi-threshold technique were utilized to segment the shadow or the bottom dark area of the overtaking vehicle, which was helpful to locate its position. Experiments were conducted on highway scenarios with various weather conditions. Results show that the proposed methodology has good reliability and robustness under different circumstances and the correct detection rate is close to 93%. ? right.