计算机工程 ›› 2012, Vol. 38 ›› Issue (20): 132-135.doi: 10.3969/j.issn.1000-3428.2012.20.034

• 人工智能及识别技术 • 上一篇    下一篇

运动摄像机下多目标检测与跟踪

陈李迪超,郭继昌   

  1. (天津大学电子信息工程学院,天津 300072)
  • 收稿日期:2011-11-30 修回日期:2012-01-31 出版日期:2012-10-20 发布日期:2012-10-17
  • 作者简介:陈李迪超(1988-),女,硕士研究生,主研方向:机器视觉,数字图像处理;郭继昌,教授、博士生导师
  • 基金项目:
    天津市科技支撑计划基金资助项目(10ZCKFGX00700)

Detection and Tracking of Multi-object Under Moving Camera

CHEN Li-di-chao, GUO Ji-chang   

  1. (School of Electronic Information Engineering, Tianjin University, Tianjin 300072, China)
  • Received:2011-11-30 Revised:2012-01-31 Online:2012-10-20 Published:2012-10-17

摘要: 针对摄像机运动情况下多目标的检测与跟踪问题,提出一种将Global K均值与模板匹配相结合的方法。利用六参数仿射模型得到摄像机运动参数,对图像进行全局运动补偿,用Global K均值算法对前景点进行循环聚类,判断目标数目并进行跟踪,通过对目标区域进行模板匹配使跟踪结果更准确。实验结果表明,该方法能够在运动摄像机下稳定、实时地跟踪多个目标,对发生形变的目标基本也能稳定跟踪。

关键词: 运动摄像机, 多目标, 跟踪, 仿射变换模型, K均值

Abstract: In order to solve the problem of multi-object detection and tracking in real-time with the moving camera, this paper uses Global K-means based on affine transform model along with template matching. It adopts the affine transform model to get the global motion and compensates it between images. The Global K-means method is used to iteratively cluster the foreground points and decide the numbers of the objects to track. The template matching of objects regions makes the tracking more accurate. Experimental results demonstrate that the method can track multiple-object steadily in real-time with moving camera.

Key words: moving camera, multi-object, tracking, affine transforming model, K-means

中图分类号: