中文摘要：

为快速定位车辆前方的行人,提出一种基于腿部区域多尺度的梯度方向直方图（HOG）特征与hu矩特征融合的行人检测方法,融合腿部区域的轮廓和纹理特征,有效地降低特征向量维数。实验结果表明,该算法能在保持较高的检测准确率的同时提高检测速度。

英文摘要：

Abstract： To gain a quick assessment of pedestrian position moment is proposed. It makes use of the features and textures dimensions. The results indicate that this method can maintain , a method combining of legs and it effectively high detection accuracy multi-scale HOG and hu lowers the feature vector and speed.