摘要
格拉斯曼尼(Grassmannian)算法是一种可以由高度不完整信息追踪子空间的在线学习算法,它在视频运动目标跟踪时具有鲁棒性和低复杂度等优点,可以应用在视频前景与背景的实时分离的情况。针对格拉斯曼尼算法在前景分离中,面对室内全局光线突变会产生大量噪声的问题,提出了一种优化的预处理方法。通过HSV色彩空间变换对视频进行阴影检测,根据阈值判断光线变化情况并自适应调整前景内容,最终实现在光照变化情况下的运动目标检测,并有效去除了原格拉斯曼尼算法在光线突变会产生的大量噪声,提高了对光照变化的鲁棒性。
Grassmannian robust adaptive subspace tracking algorithm is a low-complexity and robust online algorithm for tracking subspaces from highly incomplete information. It can solve the problems of real-time separation of background from foreground in videos. In this paper, an improved pre-processing method is proposed, which can deal with the noise when indoor illumination has great changes. The new method includes a shadow detection based on HSV color space and observation of illumination changes. The foreground can be changed adaptively according to a threshold. Finally, moving objects can be tracked and noise caused by sudden illumination changes can be eliminated. The new algorithm becomes more robust to illumination changes.
出处
《电视技术》
北大核心
2015年第10期1-4,30,共5页
Video Engineering
基金
国家自然科学基金项目(61373151)
上海市自然科学基金项目(13ZR1415000)
关键词
前景提取
光照变化
HSV颜色空间
GRASTA
separation of background from foreground
illumination change
HSV color space
GRASTA
