一种新的混合模型运动检测算法  

Novel hybrid motion detection algorithm based on auto threshold segmentation

作　　者：曹鹏[1] 张鹏[2] 

机构地区：[1]西北工业大学电子信息学院,西安710072 [2]解放军理工大学通信工程学院研究生四队,南京210007

出　　处：《计算机工程与应用》2012年第31期171-174,190,共5页Computer Engineering and Applications

摘　　要：设计了一种新的基于自动阈值分割的混合模型运动检测算法。将单高斯背景模型中求解门限阈值的思想引入,实现了混合模型门限的自适应调整和运动目标的分割。为提高运动检测的准确性,使用相邻三帧差法生成的时域运动前景掩模对检测结果进行修正。实验结果表明,混合模型算法不仅能够清晰准确地获得运动目标的完整信息,而且较好地消除了噪声的干扰。This paper designs a novel hybrid moving object detecting algorithm based on auto threshold segmentation. Through introducing the idea of uni-Gaussian model thresholding method, the proposed hybrid algorithm succeeds in realizing the auto threshold adjustment and foreground object segmentation. To ensure the accuracy of moving object detection, a temporal foreground mask is generated by three-frame differencing, and then applied to correcting the detected result. Experimental results show that this algorithm can obtain clear and complete information of the moving object, and eliminate noise fundamentally for the scene at the same time.

关 键 词：运动目标检测 自动阈值分割 时域前景掩模 混合模型算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]