是指利用第一次训练的分类器在负样本原图上进行行人检测时所有检测到的矩形框，这些矩形框区域很明显都是误报，把这些误报的矩形框保存为图片，加入到初始的负样本集合中，重新进行SVM的训练，可显著减少误报。这种方法叫做自举法-The first is the use of trained classifier all detected rectangle detect pedestrians in the negative when the sample picture, the rectangular frame are obviously false, to save these false rectangle for the picture, adding to negative in the initial sample collection, re-SVM training, can significantly reduce false positives. This method is called bootstrap method