Pedestrian Detection Using Covariance Descriptor and On-line Learning

Pedestrian detection is an important yet challenging problem in object classification due to flexible body pose, loose clothing and ever-changing illumination. In this paper, we employ covariance features and propose an on-line learning classifier which combines naive Bayes classifier and cascade support vector machines (SVM) to improve the precision and recall rate of pedestrian detection in still images. Experimental results show that our strategy can significantly increase both precision and recall rates in some difficult situations. Furthermore, even under the same initial training condition, our method outperforms HOG + AdaBoost in USC Pedestrian Detection Test Set, INRIA Person dataset and Penn-Fudan Database for Pedestrian Detection and Segmentation.