摘要
鉴于目前视频序列的运动目标检测常使用背景差分法,但其受光照变化以及阴影影响,不能准确区分出运动物体的特点,因此提出一种新的背景建模算法.该算法融合纹理信息和颜色信息建立背景模型,其中的纹理信息采用LBP算子进行描述.为更好地描述纹理信息,进一步改进了基本的LBP算子,并同时引入抗噪因子增强抗噪声影响.实验证明,提出的算法在大多数情况下都取得良好的效果.
Currently, the background subtraction is often used in moving target detection of video sequences. However, with the varying illumination and shadow, this method can not accurately distinguish motioning objects. To tackle this shortfall, a new background modeling algorithm which fuses texture and color information to create a background model is proposed in this paper. The texture information is often described with LBP (local binary pattern, local binary mode) operator. The basic LBP operator is improved in this work to characterize the texture information, while the anti-noise factor is also introduced to enhance the anti-noise impact. Experimental results show that the proposed algorithm has achieved satisfactory results in most cases.
出处
《宁波大学学报(理工版)》
CAS
2013年第1期43-47,共5页
Journal of Ningbo University:Natural Science and Engineering Edition
基金
国家科技重大专项(2011ZX03002-004-02)
浙江省移动网络应用技术联合重点实验室项目(2010E10005)
关键词
运动目标检测
背景差分
背景建模
纹理信息
局部二值模式
抗噪因子
moving objects detection
background subtraction
background model
texture information
localbinary pattern
anti-noise factor