摘要
混合高斯背景模型算法被广泛地运用于运动检测中,但是该算法在一些复杂的室外场景下未能有效地反映背景,容易出现误检测。为此提出一种改进的算法,该算法在更新背景模型时对不同的区域采用不同的更新速率,并在进行前景检测时加入一种阈值判断,最后对检测结果进行去噪处理。实验结果表明,改进后的算法能够更好地处理多模态区域,减少前景检测中出现的空洞,避免由于方差过度收敛引起的误检测,从而更精确地实现运动目标分割。
The Gaussian mixture background algorithm has been used in motion detection widely,but it can’t effectively reflect the background in some complex outdoor situations and may bring the false detection problem easily. This paper proposed an improved algorithm. This algorithm used different update rates for the different regions when updated the background model,and added a threshhold is added in the process of foreground detection. Finally,done a denosing process for the detection results. The experimental result shows that the proposed algorithm can deal with the multi-modal regions more effectively,reduce the detection hole and avoid false detection which is caused by the excessive convergence of the variance,thereby achieving more accurate moving object segmentation.
出处
《计算机应用研究》
CSCD
北大核心
2010年第9期3561-3563,3566,共4页
Application Research of Computers
基金
国家“863”计划资助项目(2007AA01Z104)
国家自然科学基金资助项目(60973030)
关键词
智能视频监控
运动检测
混合高斯模型
更新速率
多模态
intelligent video surveillance
motion detection
mixture gaussian model
update rates
multi-modal
