期刊文献+

基于可变形部件模型的粒子滤波快速行人检测与跟踪 预览

RAPID PARTICLE FILTER PEDESTRIAN DETECTION AND TRACKING BASED ON DEFORMABLE PART MODELS
在线阅读 下载PDF
分享 导出
摘要 针对视频中的行人检测和跟踪问题,提出一种基于可变形部件模型的快速行人检测、改进粒子滤波的行人跟踪算法。在行人检测阶段,为了改善非刚体行人的检测精度,采用了混合多尺度可变形部件模型;同时为了加速行人底层特征的计算,采用了基于预测算法的快速特征金字塔计算行人特征,代替传统的计算图像特征金字塔的每一个尺度特征。在行人跟踪阶段,采用时变的状态空间模型和基于颜色梯度直方图的观测模型对检测到的行人进行跟踪。实验证明,改进的行人检测算法可以在性能损失忽略不计的条件下,大大提高检测速度,并且相对于传统的行人跟踪,改进的粒子滤波算法对行人这一非刚性目标能实现较好的跟踪。 Aiming at the problem of pedestrian detection and tracking,a new fast pedestrian detection based on deformable part models and an improved tracking algorithm based on particle filter are proposed. In the pedestrian detection stage,the mixtures of multiscale deformable part models is adopted to improve the detection accuracy of non-rigid pedestrian; meanwhile,the fast feature pyramids based on the prediction algorithm is adopted to reduce the computing time of multi-scale pedestrian features instead of traditional calculation of each scale characteristics of the pyramid. In the pedestrian tracking stage,the state space model of time-varying and the observation model based on color gradient histogram are used to track the pedestrian. Experiments show that the modified pedestrian detection algorithm yields considerable speedups with negligible loss in detection accuracy,and the modified particle filter algorithm can achieve a better tracking for the non-rigid pedestrian compared with the traditional pedestrian tracking.
作者 王传旭 郝艳婷 Wang Chuanxu Hao Yanting(College of Information Science and Technology, Qingdao University of Science and Technology, Qingdao 266061, Shandong, China)
出处 《计算机应用与软件》 2017年第1期160-164,179共6页 Computer Applications and Software
基金 国家自然科学基金项目(61472196) 山东省自然科学基金项目(ZR2015FM012)
关键词 行人检测 行人跟踪 可变形部件模型 快速特征金字塔 粒子滤波 Pedestrian detection Pedestrian tracking Deformable part models Fast feature pyramids Particle filter
作者简介 王传旭,教授,主研领域:图像处理,计算机视觉. 郝艳婷,硕士生.
  • 相关文献

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部 意见反馈