A fast and efficient pedestrian detection method

The invention discloses a fast and efficient pedestrian detection method. The invention improves on the traditional pedestrian detection framework, and before extracting features, an improved algorithm HWEBING based on BING features is proposed for pre-detecting images, which can reduce a large number of windows of non-objects and screen out possible candidate windows of objects. The detection rate of the MLBP feature of the invention is increased by 3.5% and 2.1% respectively compared with the detection rate of the uniform mode LBP and the basic mode LBP. Compared with the traditional pedestrian detection method, the speed of extracting HOG and MLBP features after pre-detection by HWEBING algorithm is 5.5 times faster than that of extracting HOG and MLBP features directly. The combinationof the HWEBING algorithm and HOG-MLBP features achieves better pedestrian detection results. 本发明公开了种快速高效的行人检测方法。本发明在传统的行人检测框架上做出了改进，在提取特征之前提出了种基于BING特征的改进算法HWEBING对图像进行预检测，预检测能够减少大量的非物体的窗口，筛选出可能是物体的候选窗口。本发明的MLBP特征的检测率比均匀模式LB