行人检测是物体检测领域的一大难点。为了进一步提高行人检测的精度和速度，将Kobi Levi和Yair Weiss提出的边缘方向直方图特征和传统的Adaboost算法两者有机地结合起来，提出了一种基于改进的EOH特征的行人检测算法。该算法首先对原先只适用于较为简单的人脸检测的EOH(edge orientation histogram)特征进行了改进，弥补了其对于行人的对称性特征的描述能力不足的问题，然后通过改进Adaboost算法中对样本权值进行调整的策略来减少overfitting。实验证明，该方法的检测性能能够接近目前行人检测的领先水平。在误报率为1/10000时，该算法在一个复杂的Inria行人数据集上的检测率可以达到90％。对于640×480大小的图片，该算法的检测速度可以达到2fps。

Pedestrian detection is a difficult problem in the field of object detection. We combine Kobi Levi and Yair Weiss’s edge orientation histogram and Dalal’s hog(histogram of gradients) feature and apply them to pedestrian detection. We improve the algorithm from the following aspects first, we have changed the calculation formula of the original EOH (edge orientation histogram) to gain more descriptive ability. Second, we have changed the policy of updating the weight of the samples of the original Adaboost algorithm in order to reduce overfitting. Experiments show our method is very efficient. When the false positive rate is 1/10000, our detection rate is about 90% on Inria pedestrian dataset. The running speed is about 2 fps with 640×480 images on a 1.8GHz CPU.