计算机工程 ›› 2011, Vol. 37 ›› Issue (18): 171-173.doi: 10.3969/j.issn.1000-3428.2011.18.056

• 人工智能及识别技术 • 上一篇    下一篇

融合时空信息的运动目标检测算法

牛武泽,石林锁,金广智,李喜来,白向峰   

  1. (第二炮兵工程学院502教研室,西安 710025)
  • 收稿日期:2011-04-27 出版日期:2011-09-20 发布日期:2011-09-20
  • 作者简介:牛武泽(1987-),男,硕士研究生,主研方向:计算机视觉,智能交通;石林锁,教授;金广智,硕士研究生;李喜来、白向峰,博士研究生

Moving Object Detection Algorithm with Fusion of Time and Spatial Information

NIU Wu-ze, SHI Lin-suo, JIN Guang-zhi, LI Xi-lai, BAI Xiang-feng   

  1. (The 502 Staff Room, The Second Artillery Engineering College, Xi’an 710025, China)
  • Received:2011-04-27 Online:2011-09-20 Published:2011-09-20

摘要: 传统运动目标检测算法在处理诸如树叶晃动、水面波纹等动态场景时效果不理想。为此,针对动态场景下所存在的背景扰动问题,提出一种融合时间和空间信息的运动目标检测算法。该算法通过增量式主成分分析提取空间上图像的背景信息,结合三帧差分法所提取的时域信息进行融合决策以提取运动目标。实验结果表明,该算法能够在动态场景中有效提取运动目标,且检测结果优于混合高斯模型算法。

关键词: 智能视频, 运动目标检测, 时空信息, 增量式主成分分析, 三帧差分法

Abstract: Moving object detection is the basic technology of intelligent video surveillance. The background of scene is modeled on every pixel in traditional algorithms which performs poorly in the scenes with waving leaves and rippling water. Aiming at the problem of background disturbance in dynamic scenes, a kind of time and space information fusion target detection algorithm is put forward. In this algorithm, spatial background information is extracted by incremental Principal Component Analysis(PCA). Decision is made by combination with three frame difference method extracting information of time domain. Experimental results show this algorithm can effectively extract moving targets in dynamic scenes and performs better than Gaussian Mixture Model(GMM) algorithm.

Key words: intelligent video, moving object detection, time and spatial information, incremental Principal Component Analysis(PCA), three frame difference method

中图分类号: