MFR-MAP相结合的运动目标检测方法研究  

The Research of Moving Object Detection Based on MRF-MAP Theory

作　　者：赵彤洲[1,2] 王海晖[1,2] 

机构地区：[1]武汉工程大学智能机器人湖北省重点实验室,湖北武汉430073 [2]武汉工程大学计算机科学与工程学院,湖北武汉430073

出　　处：《电脑与电信》2010年第10期42-44,共3页Computer & Telecommunication

基　　金：湖北省自然科学基金"多模态信息融合图像分析的理论及方法研究"资助;项目编号:2008CDB333

摘　　要：针对运动目标检测的准确性和完整性要求,将静态图像分割算法的空域信息与运动目标检测的时域信息相结合,提出了采用基于MRF-MAP框架的时-空联合的目标检测方法。利用Mean-Shift算法进行空域检测,利用在显著性水平约束下的帧间差分法进行时域检测,构造了时-空联合的MFR模型和相应的能量函数,通过求解能量函数最优解,得到较为理想的运动目标检测结果。For the demand of accuracy and integrity of moving object detection,the spatial space information which is used in static image segmentation algorithm and the temporal space information which is used in moving object detection are combined,and a spatial-temporal joint detection method based on MRF-MAP frame is proposed.In spatial detection,it uses the Mean-Shift algorithm.In temporal detection,it uses frame difference method which is confined by significance level.Then the spatial-temporal joint MFR model and the corresponding energy function are made,the ideal results on moving object detection can be obtained by resolving the optimal solution of the energy function.

关 键 词：运动目标检测 时-空联合 MFR-MAP框架 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]