融合改进帧差和视觉背景提取算法的前景检测  

Moving Object Detection Algorithm Using Improved ViBe Combined with Frame Difference

作　　者：段锁林[1] 高仁洲 刘福 刘毛毛 王一凡[2] 潘礼正[1] DUAN Suo-lin;GAO Ren-zhou;LIU Fu;LIU Mao-mao;WANG Yi-fan;PAN Li-zheng(Robotics Institute,Changzhou University,Changzhou 213164,China;Mechanical and Electrical Engineering College,Changzhou Vocational Institute of Textile and Garment,Changzhou 213164,China)

机构地区：[1]常州大学机器人研究所,江苏常州213164 [2]常州纺织服装职业技术学院机电工程学院,江苏常州213164

出　　处：《小型微型计算机系统》2019年第9期1903-1908,共6页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61773078)资助;常州市科技支撑计划项目(CE20175040)资助;江苏省高等学校自然科学研究面上项目(18KJB460001)资助

摘　　要：视觉背景提取算法背景建模速度快,受噪声影响小,但是该算法容易产生阴影,且无法快速有效的去除鬼影.针对上述问题,提出了一种融合改进帧差法和改进视觉背景提取的算法.该方法采用帧差的实时特征化方法,将帧间差分图像与视觉背景提取差分图像进行逻辑运算来去除鬼影,同时视觉背景提取算法在判断前景时采用了自适应阈值的方法.通过计算候选前景与背景的亮度、色度和饱和度的变化比率进行阴影检测并将其去除,并对产生的前景图进行开闭操作去除残余的噪点以及填充目标区域的空洞.实验结果表明,改进的算法能够加快鬼影的消除,并且能有效地去除目标阴影.The ViBe( Visual Background Extractor) algorithm has a fast background modeling and is less affected by noise,but it is prone to shadows and can’t quickly and effectively remove ghost area. To solve the above problems,an algorithm that combines improved inter-frame variance and improved ViBe is proposed. The frame difference real-time characterization method is used to perform a logical operation on the inter-frame variance image and the ViBe difference image to remove the ghost area in this method,and an adaptive threshold method in ViBe is used to determine the foreground. Detect shadows and remove them by calculating the ratio of change in brightness,chrominance and saturation of the candidate foreground to the background. At the same time,residual noise is removed and fills is holed in the target area of the generated foreground through a morphology post-processing with opening operation and closing operation. Experiment results show that this method speeds up ghost removal and effectively removes target shadows.

关 键 词：运动目标检测 视觉背景提取算法 帧间差分 鬼影消除 自适应阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]