基于PCA降维的多特征级联的行人检测

Pedestrian Detection Based on PCA Dimension Reduction of Multi-feature Cascade

在行人检测中,针对梯度方向直方图(HOG)冗余信息过多、检测速度慢等不足,提出了运用PCA降维的多特征级联的行人检测。首先利用PCA对HOG特征进行降维,其次将HOG特征和Gabor特征、颜色特征级联作为行人检测的特征,最后使用SVM的径向基(RBF)核函数进行分类。在INRIA行人库上的实验表明,该方法不但提高了分类的速度,而且提高了检测的准确率。

In pedestrian detection, Histogram of oriented gradient（HOG） has the defects of too much redundant information, low detection speed, this paper proposed features cascading pedestrian detection based on PCA dimensional reduction. Firstly,we used PCA to reduce the dimension of HOG features, then took HOG features, Gabor features and color features as the features of pedestrian detection. Finally we used SVM radial basis （RBF） kernel function to classify. Experiments on INRIA pedestrian database show that this method not only increases the speed of classification, but also improve the accuracy of detection.