【摘要】 基于传统AdaBoost算法的识别率和误报率同时低的原因。提出一种改进AdaBoost算法的行人检测方法,采用类Haar特征作为行人特征,引入误报率来更新样本权重,使得识别率降低得更慢,实现一个级联的行人检测分类器。相比于传统AdaBoost算法,改进算法不仅取得较优的识别率,并且可以降低分类器训练的层数。实验证明了改进算法的有效性。更多还原

【Abstract】 The recognition rate and misinformation rate of traditional AdaBoost algorithm have been reduced simultaneously.The article proposes a pedestrian detection method based on the improved AdaBoost algorithm.Haar-like feature is selected as pedestrian feature and the introduction of misinformation rate updates the sample weight which makes recognition rate reduced slowly,the pedestrian detection cascaded classifier is designed.Compared with traditional AdaBoost algorithm,it can get higher recognition rate,and reduce the numbers of classifier training layers.Experiment shows the effective of improved algorithm.更多还原