摘要
假定图像序列的背景图像已经获得,提出一种基于时空背景差的运动目标检测算法.该算法融合背景差分、基于时间信息的帧间差分及基于空间信息的背景差分信息,得到真实运动物体的运动种子点,认为背景差分图像中包含运动种子点的连通区域为真实的前景目标,从而可以检测出正确而完整的前景目标.仿真实验表明,该算法可以避免背景模型对场景的表征不足及背景更新阶段造成的错误检测,即使在场景中存在微小运动的复杂环境下,仍能实现准确的运动分割.
Assuming that background had been extracted from input images, a new method is proposed for the effective detection of moving objects from an image sequence in this paper. The background difference, frame difference based on time information and background difference based on spatial information were fused to get the moving seeds of real moving objects, and then only those connected moving components, which contain the seed pixels, in the background difference results, were selected as the final moving foreground objects. Simulation results show that false detection from wrong modeling or inadequate background updating is avoided completely. And the proposed motion detection algorithm is also shown to perform correctly in case that the scene contains small motions.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2006年第7期1044-1048,共5页
Journal of Computer-Aided Design & Computer Graphics
基金
国家重点基础研究发展规划项目(2001CB309403)
关键词
背景差
时空信息
运动检测
视频分析
background difference
space-time information
motion detection
video sequence analysis
