Abstract: Haar and Gabor wavelet transforms are two commonly used methods for feature extraction.The former is widely used in object detection and the latter is commonly used in face recognition.A Gabor wavelet based feature extraction method for human detection is proposed，and verified using three main pedestrian datasets.Experiments show that：Gabor wavelet representation for ima-
ges has the ability to describe the local intensity variation at different orientations of different scales.As a result，it achieves better performance than Haar wavelet representation，which can only encode image regions in vertical，horizontal and diagonal directions.

Key words: human detection, Gabor wavelet, Haar wavelet, pedestrian dataset

摘要： Haar小波和Gabor小波变换是常用的特征提取方法，前者广泛用于目标检测，后者则常用于人脸识别。针对人体目标检测，提出采用Gabor小波变换进行特征提取，并采用三个主要的行人库与Haar小波方法进行对比实验，实验显示：由于二维Gabor小波变换响应能够在多个尺度的多个方向上对目标的局部区域像素值变化进行描述，所以相比只能在水平、垂直和对角线三个方向上描述目标的Haar小波，其优势明显。

关键词: 人体目标检测, Gabor小波, Haar小波, 行人库