中国机械工程

• 智能制造 • 上一篇    下一篇

基于双目视觉和约束条件的行人目标定位

杨志光1;张永林1;谢经明2;陈幼平2   

  1. 1.武汉轻工大学,武汉,430023
    2.华中科技大学国家数控技术工程研究中心,武汉,430074
  • 出版日期:2016-12-10 发布日期:2016-12-15
  • 基金资助:
    湖北省科技计划资助项目(2014BEC056)

Target Positioning of Pedestrian Based on Binocular Vision and Constraints

Yang Zhiguang1;Zhang Yonglin1;Xie Jingming2;Chen Youping2   

  1. 1.Wuhan Polytechnic University,Wuhan,430023
    2.National NC System Engineering Research Center, Huazhong University of Science and Technology,Wuhan,430074
  • Online:2016-12-10 Published:2016-12-15
  • Supported by:

摘要: 针对行人检测中计算量大、训练分类器耗时和无法满足实时性要求等问题,提出了一种基于双目视觉的行人目标定位方法。该方法利用图像处理技术获取候选轮廓,将轮廓的几何特征作为约束条件来筛选候选轮廓;利用双目视觉获取轮廓的深度信息后,将深度信息作为约束条件对候选轮廓进行进一步筛选。通过深度信息和几何信息的共同约束,识别出行人的头部轮廓,从而实现对行人目标的定位。实验结果表明,该方法减小了计算量,提高了检测精度。

关键词: 双目视觉, 图像处理, 行人检测, 目标定位

Abstract: When confronting the problems in pedestrian detection such as large amount of calculation, time-consuming of classifier training and unfulfilled real-time requirements, a pedestrian detection method was proposed based on binocular vision. In this method, a lot of candidate contours might be obtained by processing image, and the geometrical characteristics of contours were used as a constraint to filter candidate contours. In the meantime, binocular camera might be used to get depth informations of the candidate contours, and the depth informations were used as a constraint to filter the candidate contours. By the constraints of depth informations and geometrical informations, contours of pedestrians' heads might be identified and the pedestrians' localization might get.  The experimental results show that the method reduces the amount of calculation, and enhances the detection accuracy.

Key words:  binocular vision, image processing, pedestrian detection, target positioning

中图分类号: