摘要
建立了一种适用于复杂交通场景的多层次背景模型,采用随机差影法来获取背景的候选像素,利用空间统计与时间统计的方法确定背景及其可信度。结合目标跟踪结果以区域为单位进行背景更新并排除浅阴影干扰,能在目标出现运动状态变化时快速地恢复背景,并且能够检测出场景中的部分静止目标。实验表明,本算法能及时恢复真实背景,提高了目标检测的准确性。
In this paper, a multilayer background model adapted to complex traffic scene is established. A random image subtraction method is adopted to obtain candidate background pixels. A method integrated with temporal and spatial statistics is used to build the background and the credibility of every pixel. The updating is implemented with fedback of object tracking results and shadows excluding, which contributes in fast background restoring when objects moving status change, and detecting temporary static objects in the scene. Finally, Experiments prove that the algorithm presented in this paper is reliable in restoring real background timely and the veracity of object detection is improved.
出处
《武汉大学学报（信息科学版）》
EI
CSCD
北大核心
2009年第8期906-909,共4页
Geomatics and Information Science of Wuhan University
基金
国家自然科学基金资助项目(40721001)
国家教育部新世纪优秀人才计划资助项目(NCET050625)
关键词
交通信息
差影法
背景提取
目标检测
traffic information
image subtraction
background extraction
object detection
