摘要
针对视觉背景提取(Vi Be)算法在初始化背景模型时易产生鬼影以及对户外动态场景适应能力差、检测精度低等问题,在算法初始化阶段,通过采用前景点计数的方法,统计静止像素点占前景块的百分比,有效抑制了鬼影的产生,提高了算法检测精度;在前景检测阶段,根据场景的动态程度自适应调整阈值,提高了算法对动态背景的适应能力。实验表明:与原算法相比,在动态场景中算法检测精度提高了20%以上。
Aiming at problem that visual background extraction（ Vi Be） algorithm is prone to produce ghost when initializing the background model and its adaptability is poor in outdoor dynamic scenes and the detection precision is low,an improved visual background extraction algorithm is proposed. In initial stage of the algorithm,to improve precision of the algorithm,the ghost is effectively suppressed by counting the percentage of still pixels in foreground blocks. In the foreground detection stage,improve the adaptability in dynamic scenes,dynamically adjust threshold according to the scene dynamic degree. Experimental results demonstrate that precision of this algorithm is improved by 20 % compared with the original Vi Be algorithm.
作者
徐赟
彭勇
胡旻涛
XU Yun, PENG Yong, HU Min-tao(School of Internet of Things Engineering, Jiangnan University, Wuxi 214122, Chin)
出处
《传感器与微系统》
CSCD
2018年第3期38-40,43,共4页
Transducer and Microsystem Technologies
基金
江苏省产学研联合创新资金资助项目(BY2013015-35)
关键词
背景建模
视觉背景提取
前景检测
鬼影
自适应阈值
background modeling
visual background extraction
foreground detection
ghost
adaptive threshold
