东北大学学报:自然科学版 ›› 2016, Vol. 37 ›› Issue (4): 481-485.DOI: 10.12068/j.issn.1005-3026.2016.04.006

• 信息与控制 • 上一篇    下一篇

基于改进ViBe的多行人检测方法

朴春赫1,2, 潘怡霖1, 赵海1, 朱宏博1   

  1. (1. 东北大学 计算机科学与工程学院, 辽宁 沈阳110819; 2. 金策工业综合大学 电子工程系, 朝鲜 平壤999093)
  • 收稿日期:2015-04-13 修回日期:2015-04-13 出版日期:2016-04-15 发布日期:2016-04-05
  • 通讯作者: 朴春赫
  • 作者简介:朴春赫(1976-),男, 朝鲜平壤人,东北大学博士研究生; 赵海(1959-),男,辽宁沈阳人,东北大学教授,博士生导师.
  • 基金资助:
    中央高校基本科研业务费优秀科技人才培育资金资助项目(N140405004); 辽宁省科技厅重点实验室建设基金资助项目(LZ201015).

Multi-pedestrian Detection Approach Based on Improved ViBe Algorithm

PAK Chun-hyok1,2, PAN Yi-lin1, ZHAO Hai1, ZHU Hong-bo1   

  1. 1.School of computer Science & Engineering, Northeastern University, Shenyang 110819, China; 2. School of Electronic Engineering, Kim Chaek University of Technology, Pyongyang 999093, DPRK.
  • Received:2015-04-13 Revised:2015-04-13 Online:2016-04-15 Published:2016-04-05
  • Contact: PAK Chun-hyok
  • About author:-
  • Supported by:
    -

摘要: 利用改进的ViBe算法及模板匹配方法对多行人场景进行了目标检测.为了提高运动目标检测性能,提出将原ViBe算法与帧间差分相结合的方法,该方法使二值图像中的鬼影比原ViBe算法消除更为迅速.根据待处理像素点周围已完成的前背景分离结果,利用动态计算模型参数估计前背景,以提高前景目标提取的准确度.基于HOG算法识别运动目标中的行人目标.实验结果表明,该方法在没有降低运动物体检测实时性的同时,明显提高了算法检测的准确度.

关键词: 运动目标检测, 鬼影消除, ViBe, 支持向量机, 行人检测

Abstract: The moving object detection of multi-pedestrian scenario was conducted by using the improved ViBe algorithm and the template matching method. The original ViBe algorithm was combined with the frame difference method to improve the performance of moving object detection, which can more rapidly eliminate the ghosts in the result binary image than the original ViBe algorithm. According to the treated foreground result around the treating pixel, the dynamic model parameters were used to estimate the foreground for improving the accuracy of the foreground object detection. The HOG algorithm was used to detect the pedestrians in the moving objects. The results showed that the method does not reduce the real-time performance of moving object detection, and significantly improve the accuracy of the detection algorithm.

Key words: moving object detection, ghost elimination, ViBe, SVM, pedestrian detection

中图分类号: