摘要
给出了一种对光照变化等实际环境中干扰不敏感的基于水平集和背景减的运动目标检测与分割方法.首先建立了当前帧图像与背景参考帧图像间的纹理差,并将其与灰度差相结合构成差图像;然后,分析了纹理特征,并采用适当措施减小由场景中杂乱运动产生的不利影响,从而实现运动目标的鲁棒检测;最后采用了基于区域的水平集方法,将多个信息有机地结合在一起,实现对多个刚体或非刚体运动目标的分割.实际采集序列图的仿真实验验证了该方法的有效性.
Based on level set method and background subtraction, this paper proposed an approach to realize robust detection and segmentation of multiple motion objects. Firstly, a novel texture difference was built and combined with intensity difference to develop difference image, which contributes to the elimination of the influence from illumination changes in real environment. Then, certain measures deriving from texture density similarity comparison were taken against clutter motion in the scene. Finally, region-based level set method was adopted to simultaneously segment moving objects, on consideration of no limitation on rigid objects and non-rigid. The experimental results of two real image sequences show that the approach is favorable.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
2004年第5期713-717,共5页
Journal of Shanghai Jiaotong University
基金
国家自然科学基金资助项目(30170264)
