行人检测是物体检测领域的一大难点。为了更加快速地检测行人，将人脸检测中BoostedCascade算法应用到行人检测中，并对其进行了改进，即先利用加权平均矢量投影的方法将高维的梯度直方图特征化为1维，再利用lookuptable进行概率密度估计，从而将GentleAdaboost成功地应用于行人检测。实验表明，该方法不仅训练时间短、检测速度快，而且检测精度接近目前的最佳水平。

Pedestrian detection is a difficult problem in the field of object detection. This paper applies the boosted cascade algorithm which originates from face detection to pedestrian detection. We improve the algorithm by first project the high dimensional hog(histogram of gradient) feature to 1 dimension, then estimate the posterior probability with lookup table so as to apply the gentle adaboost algorithm. Experiments indicate our algorithm is both efficient in training and detection. The detection rate is close to the state of the arts.