基于自适应ViBe算法的动态背景运动目标检测  

Moving object detection in dynamic background based on adaptive ViBe algorithm

作　　者：汤旻安[1] 王晨雨 罗引航 TANG Min’an;WANG Chenyu;LUO Yinhang(School of Automation and Electrical Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China)

机构地区：[1]兰州交通大学自动化与电气工程学院,甘肃兰州730070

出　　处：《传感器与微系统》2023年第11期119-122,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61663021,61763025,61861025)。

摘　　要：针对传统ViBe算法存在鬼影、背景适应性差以及前景提取效果不佳的问题,提出了一种改进的自适应ViBe算法。首先,使用哈希(Hash)算法提取关键帧进行差分运算,将差分图像与频率调谐(FT)显著性检测图像结合,提取当前帧的目标区域,填充背景像素,滤除真实目标之外的鬼影区域;然后,利用像素“最小距离”和局部标准差计算出背景的复杂度,自适应地调整像素点的半径阈值和更新速率,减少背景像素影响;最后,由比例图像矩阵信息与HSV空间中的颜色特征粗略判断出阴影区域,由拉普拉斯算子计算阴影区域像素梯度并将梯度值较小的阴影像素滤除。实验结果表明:改进后的算法能够相对有效地抑制鬼影和阴影现象,前景检测时F-Measure值保持在较高水平,对动态背景具有一定的鲁棒性。Aiming at the problem that the traditional ViBe algorithm has ghost,low background adaptability and poor foreground extraction effect,an improved adaptive ViBe algorithm is proposed.Firstly,Hash algorithm is used to extract the key frame for differential operation,frequency tuning(FT)saliency detection image and the difference image are combined to extract the target area of the current frame,fill with background pixels and filter out the ghost area outside the real target.Secondly,the pixel"minimum distance"and local standard deviation are used to compute the complexity of the background which adjusts the radius threshold pixel points and updates rate adaptively to reduce the influence of background pixels.Finally,the shadow area is roughly determined by the ratio image matrix information and the color features in HSV space.The pixel gradient of the shadow area is calculated by Laplace operator,and the shadow pixels with small gradient value are filtered out.Experimental results show that the improved algorithm can effectively suppress ghost and shadow phenomena,the F-Measure value is kept at a high level during foreground detection and has a certain degree of robustness to dynamic backgrounds.

关 键 词：ViBe算法 动态背景 帧间差分 图像处理 运动目标检测 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]