摘要
提出一种能在动态摄像机场景下检测前景物的算法(AGMM).本方法采用角点特征对前后两帧图像进行匹配,估算两帧图像的移动向量,并以此校正高斯混合模型(GMM),并在此基础上进行背景的重建以及前景物的分割.以不同场景的视频序列对本算法和GMM算法进行比较.实验结果表明,提出的算法能够适应动态摄像机场景,以牺牲一点复杂度为代价,大大提高检测精度,并且在摄像机移动比较大的位移时仍然可以得到正确的结果.
A new algorithm （AGMM） to detect foreground objects in an image sequence obtained with dynamic cameras is proposed in this paper. By matching corner features between the background image and the input image, the motion vector between the background image and the input image can be esti- mated. After a process corresponds to an alignment between the Gaussian mixture model and the input image, the background can be rebuilt and the final foreground separation result can be obtained. Com- pared between our method and GMM by two different scenic serials, experiment results show that the proposed algorithm can adapt dynamic cameras and improve accuracy greatly with increasing a little more complexity. Even when the displacement between two consecutive images is very large, it can still get the right result.
出处
《福州大学学报（自然科学版）》
CAS
CSCD
北大核心
2012年第2期188-192,共5页
Journal of Fuzhou University(Natural Science Edition)
基金
福建省自然科学基金资助项目(2009J05130)
福建省教育厅科研资助项目(JB11127)
关键词
背景建模
运动检测
前景背景分割
特征匹配
background modeling
motion detection
background foreground segmentation
features match
