基于改进HOG描述符的行人检测算法

Pedestrian Detection Algorithm Based on Improved HOG Descriptor

针对现有面向梯度的直方图（HOG）描述符及其改进描述方式应用于行人检测时存在的运算速度较慢问题，提出了一种新的改进HOG描述方法。该方法对HOG描述符进行降维处理，选取特征中最具特征表达能力的部分特征向量，引入支持向量机（SVM）线性分类器对其进行训练，最后，利用所得结果进行行人检测。试验结果表明，基于该算法的行人检测算法在对检测正确率影响很小的情况下，可以提高约30％的运算速度。

Histograms of oriented gradients （HOG） descriptor and the improved description meth- ods for pedestrian detection exists a problem with low operation speeds. To solve this problem, an improved HOG description method is proposed. The method can reduce the dimension of the HOG descriptor and choose the most important features. The support vector machine （SVM） lin- ear classifier is used to train these features. Finally, results are used for pedestrian detection. Experimental results show that the pedestrian detection algorithm based on the algorithm can in- crease the detection speed by 30% with little effect on the accuracy rate.