基于改进高斯混合模型的体育视频运动目标检测与跟踪  

Detection and tracking of moving object in sports video based on improved Gaussian mixture model

作　　者：高野[1] GAO Ye(Xi’an International Studies University,Xi’an 710128,China)

机构地区：[1]西安外国语大学,陕西西安710128

出　　处：《现代电子技术》2017年第19期91-93,97,共4页Modern Electronics Technique

摘　　要：针对当前体育视频运动目标检测存在的弊端,提出改进高斯混合模型的体育视频运动目标检测与跟踪方法。通过分析高斯混合模型的弊端,保留原有的"背景重建-模型更新-背景更新-目标检测"处理进程,将彩色图像转换为灰度图像,在像素相似度差别小的背景区域进行动态扩张,加入基于灰度直方图的目标跟踪进程,提高高斯混合模型对体育视频运动目标的处理效率与精度。实验结果表明,所提方法对体育视频的干扰处理能力强,检测范围大,检测与跟踪效果好。Aiming at the drawbacks existing in current moving object detection of sports video,a detection and trackingmethod of moving object in sports video based on improved Gaussian mixture model is improved.The shortcomings of Gaussianmixture model are analyzed to retain the original processes of″background reconstruction-model update-background update-targetdetection″,convert the color image into gray image,carry out the dynamic expansion of background area with small difference ofpixel similarity,and add the target tracking process based on gray-level histogram.The improved Gaussian mixture model can improve the processing efficiency and precision of the moving target in sports video.The experimental results show that the method hasstrong ability for interference processing of sports video,wide detection range,and perfect detection effect and tracking effect.

关 键 词：高斯混合模型 体育视频 运动目标 目标检测 目标跟踪 

分 类 号：TN911.73-34[电子电信—通信与信息系统] TP391.41[电子电信—信息与通信工程]