基于混合高斯建模和纹理特征提取的人数统计方法研究  

A Statistical Method for the Number of People Based on Hybrid Gaussian Modeling and Texture Feature Extraction

作　　者：张朋[1] 温宏愿[1] ZHANG Peng;WEN Hong-yuan(Taizhou Institute of Sci.&Tech.,NUST,Taizhou 225300,China)

机构地区：[1]南京理工大学泰州科技学院,泰州225300

出　　处：《价值工程》2018年第10期235-236,共2页Value Engineering

基　　金：江苏省高校"青蓝工程"--江苏省"333工程"科研资助项目(编号:BRA2016177);资助项目:泰州市科技指导性项目(SSF20160033)

摘　　要：针对密集人群环境中行人相互遮挡造成人数难以准确统计问题,提出了一种基于混合高斯模型和图像纹理特征提取的人数密度统计方法。首先通过混合高斯建模提取前景图像,其次通过线性内插权重来进行透视矫正,最后通过构造和提取能量、对比度、熵和相关性四个特征参数进行人数统计。结果表明在不同的测试环境下系统的准确率在90%以上,达到人群密度监控的基本要求。In view of the difficulty of accurately counting the number of pedestrians in dense crowd environment,this paper proposes a population density statistics method based on Gaussian mixture model and image texture feature extraction.Firstly,the foreground image is extracted by means of Gaussian mixture modeling.Secondly,the orthographic correction is performed by linear interpolation weights.Finally,the population statistics are constructed by constructing and extracting the four characteristic parameters of energy,contrast,entropy and correlation.The results show that the accuracy of the system is more than 90%in different test environment,to achieve the basic requirements of population density monitoring.

关 键 词：人群密度估计 特征向量 混合高斯建模 透视矫正 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]