融合时域信息的自适应ViBe算法  

Self-adaptive ViBe algorithm with time domain information

作　　者：瞿中[1] 刘帅 刘妍 QU Zhong;LIU Shuai;LIU Yan(College of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065

出　　处：《计算机工程与设计》2019年第3期782-787,共6页Computer Engineering and Design

基　　金：国家级大学生创新创业训练基金项目(201710617003);重庆市基础科学与前沿技术研究基金项目(cstc 2015jcyjBX0090;cstc2015jcyjA40034;cstc 2014jcyjA10051)

摘　　要：针对ViBe算法采用第一帧初始化背景模型没有考虑像素值在时域上变化的情况,提出一种融合时域信息的自适应ViBe算法。提出一个历史像素值队列反映像素值在时域上的变化情况;提出一种结合自适应思想的阈值分割方法对前景像素和背景像素进行分割,使ViBe算法适应场景中不同动态区域;提出一种自适应的更新因子,通过自适应更新机制使背景模型能够在不同背景下更加可靠。实验结果表明,改进后的ViBe算法能够快速适应动态背景,使检测结果更为准确。The initialization of background model in the first frame neglects the pixel values changes in the time domain, to solve the problem, a self-adaptive ViBe algorithm with time domain information was proposed. A historical pixel value queue to reflect the pixel changes in the time domain was proposed. A segmentation method based on the self-adaptive idea was proposed, which ensured ViBe adapting to different dynamic regions in scene. A self-adaptive updating factor was proposed, which made the background model reliable in different backgrounds. Experimental results show that the improved algorithm can quickly adapt to the dynamic background and make the detection results more accurate.

关 键 词：ViBe 目标检测 自适应阈值 背景模型 动态更新 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]