基于Adaboost和SVM的人头实时检测

Realtime head detection based on Adaboost and SVM

针对复杂场景图像中的人头检测问题,提出一种Adaboost与支持向量机(SVM)相结合的检测算法。该算法重点对Adaboost特征进行了改进,用Adaboost对人头进行快速检测,并引入级联的SVM分类器对Adaboost检测结果进行逐级筛选,从而实现对人头的精确检测。实验表明,该方法降低了Adaboost运算复杂度,提高了特征分类能力,引入级联SVM分类器在保证高检测率的同时,降低了误检率,对复杂场景具有较强的鲁棒性。

A combined classification algorithm based on Adaboost and support vector machine is proposed in order to deal with the head detection problems in complex condition.Firstly,Adaboost with improved features are used to detect the heads with little time consuming.Then,SVM-cascade detectors are followed to detect the results of the Adaboost detector.Experiment results show that the method proposed simplifies the extraction of Adaboost features,decreases the false negative rate while maintaining a high detection rate.Moreover,this algorithm has perfect flexibility in complex background.