Abstract: Pedestrian detection has a wide application in video surveillance and intelligent vehicle system.This paper applies boosted cascade algorithm，which originates from face detection，to the pedestrian detection.Some modifications to this algorithm are also introduced.Firstly，it uses Gentle AdaBoost for classifier training to improve training efficiency.Secondly，it introduces a feature filter procedure before training to decrease training time and memory occupation.The experiment result indicates that the algorithm is efficient in both training and detection，it also has a relatively high detection rate.

Key words: Gentle AdaBoost, pedestrian detection, rectangle feature, cascade classifier

摘要： 行人检测在视频监控以及智能车系统中有着广泛的应用前景，为了能够更有效地检测行人，将人脸检测中级联检测框架引入行人检测中，并对其进行改进，采用Gentle AdaBoost算法进行分类器训练，以提高训练效率，同时在训练前引入了特征预筛选，以减少训练时间和系统开销。实验表明，改进后的方法训练时间短，检测精度高，同时具有较快的检测速度。

关键词: Gentle AdaBoost, 行人检测, 矩形特征, 级联分类器