基于DPM模型的行人检测技术的研究

在行人检测领域,当场景很复杂时,一般行人检测算法往往得不到很好的检测效果。比如在行人很多且靠的很近时,用基于梯度直方图的检测算法时,检测效果不是很好。由P.Felzenszwa提出一种以可变形部件模型为基础的检测算法,能够检测多样变化的目标类型并且在挑战Pascal目标检测中达到较高水平。该算法使用隐变量支持向量机,是一种在支持向量机基础上添加潜在变量而重新构建的支持向量机。本文提出了一种基于可变形部件模型的行人检测算法,通过建立多人体模板,在行人相互靠近有重叠的场景下有着很好的检测效果。

Research on pedestrian detection based on DPM

In the field of pedestrian detection, when the scene is very complicated, generally the pedestrian detection algorithm tends to be not very good detection effect.Forexample,when pedestrians are many and very close, the detection effect is not very good ,that detection algorithm based on gradient histogram .Part basedobjectdetectionsystemwhichfromP.Felzenszwais based ondeformablepart models.This systemis able to dealwithhighlyvariableobjectclasses and achieves state-of-the-art results in the PASCALobjectdetectionchallenges.A latent SVM is a reformulation of SVM interms of latent variables.This paper proposes a pedestrian detection algorithm baseddeformable parts model, in the scenario of pedestrian near each other has a very good detection effect.