PDF全文下载：
201904015.pdf
文章编号： 16726987（2019）04011306； DOI： 10.16351/j.16726987.2019.04.015
朱善良a， 王浩宇a， 高鑫a, 赵玉b， 谢秋玲c, 周伟峰a, 杨树国a*(青岛科技大学a.数理学院；b.经济与管理学院；c.财务处，山东 青岛 266061)
摘要： 针对运动目标检测中ViBe算法的鬼影、阴影和噪声干扰问题，本研究提出一种融入改进混合高斯模型（GMM）的ViBe算法。该算法改进混合高斯模型的自适应性，使混合高斯模型的K值与学习率对背景进行自适应调节；对视频帧进行训练，构造“虚拟”背景代替第一帧图像进行背景建模，算法能够有效地提取背景建模初始化的视频运动目标，从而消除鬼影现象。该算法用像素分类法提取前景目标，经形态学处理得到完整的运动目标。实验结果表明：与几种运动目标检测算法相比，本研究提出的算法不仅能够有效地抑制鬼影、阴影和噪声干扰，而且该算法自适应性强、检测速度快、检测结果可靠。
关键词： 运动目标检测； ViBe算法； 混合高斯模型； 形态学方法
中图分类号： TP 391文献标志码： A
引用格式： 朱善良， 王浩宇， 高鑫， 等. 一种基于改进混合高斯模型的运动目标检测算法\[J\]. 青岛科技大学学报（自然科学版）， 2019， 40（4）： 113118.
ZHU Shanliang, WANG Haoyu, GAO Xin, et al. A moving object detection algorithm based on improved gaussian mixture model\[J\]. Journal of Qingdao University of Science and Technology（Natural Science Edition), 2019, 40(4)： 113118.
A Moving Object Detection Algorithm Based on Improved Gaussian Mixture Model
ZHU Shanlianga, WANG Haoyua, GAO Xina, ZHAO Yub, XIE Qiulingc, ZHOU Weifenga, YANG Shuguoa
(a.College of Mathematics and Physics; b. College of Economics and Management; c. Finance Office,
Qingdao University of Science and Technology, Qingdao 266061, China)
Abstract: Aiming at the problems of ghost，shadow and noise interference of the Vibe algorithm in moving object detection, this paper presents a novel algorithm for moving object detection based on improved Gaussian mixture model (GMM) and the Vibe algorithm. The proposed algorithm improves the adaptability of the GMM, which makes the K value and learning rate of the GMM to adjust the background adaptively. The algorithm trains the video frames to construct a "virtual" background instead of the first frame image for background modeling. And the algorithm effectively extracts video moving objects of background initialization and the ghosting phenomenon is eliminated. Then the foreground object is extracted by pixel classification method, which obtains the complete moving object with morphological method. The experimental results show that, in comparison with some moving object detection algorithms, Our proposed algorithm not only can effectively eliminate the ghost image, shadow and noise，but also effectively works on a wide range of complex scenarios, faster detection speed, and more reliable detection results.
Key words: moving object detection； ViBe algorithm； Gaussian mixture model； morphological method
收稿日期： 2018 12 10
基金项目： 山东省高校科研计划项目（J18KA314）；青岛市源头创新计划项目（182264jch）；青岛科技大学大学生创新与创业训练计划项目（20180426202）.
作者简介： 朱善良（1977—），男，副教授，博士研究生.*通信联系人.
