基于改进Boosting算法的夜间运动车辆检测

针对夜间交通场景中运动车辆目标提取及实时检测困难的问题,引入改进的Boosting模糊分类算法,提出了一种新的基于车头灯的夜间运动车辆检测方法.通过SIFT算法和融合多特征的方法精确提取夜间运动车辆车头灯特征,利用遗传算法优化Boosting模糊分类算法,并以加权投票方式对提取的车头灯特征进行分类判别,最后结合车头灯空间、运动信息及灯光颜色信息进行同车车头灯配对分组,实现夜间运动车辆的实时检测.实验表明,该方法在夜间交通环境中具有良好的实时检测效果和较高鲁棒性.

Nighttime Motion Vehicle Detection Based on Improved Boosting Fuzzy Classification Algorithm

A method for nighttime motion vehicle detection based on vehicle headlights is proposed to solve the problem of motion vehicle target extraction and real-time detection in nighttime traffic scenes, which introduced an improved Boosting fuzzy classification algorithm. Firstly characteristics of the night moving vehicle headlights are precisely extracted by SIFT algorithm and multi-feature fusion method. And the classifications are determined by a Boosting fuzzy classification algorithm with weighted voting, which is optimized by genetic algorithm. Finally the car headlights are paired grouping with headlight space, motion and lighting color information to achieve real-time detection of nighttime motion vehicles. Experimental results show that the method have better robustness and real-time detection effect in the traffic environment at night.