基于Hadoop的景区人流密度预测模型设计与实现  

Design and implementation of prediction model of tourist flow density in scenic spots based on Hadoop

作　　者：马骞[1] MA Qian(Xi’an Aeronautical Polytechnic Institute,Xi’an 710089,China)

机构地区：[1]西安航空职业技术学院,陕西西安710089

出　　处：《电子设计工程》2021年第18期39-43,共5页Electronic Design Engineering

基　　金：陕西省教育科学规划课题(SGH17V012);西安航空职业技术学院科研计划项目(19XHSK-015)。

摘　　要：针对景区人流监控精细化、高效化的需求,文中对视频图像的背景建模与轮廓提取方法进行研究,构建了面向海量计算需求的人流密度预测模型。该模型需要对输入模型的图像质量进行检测,校核图像的分辨率、像素是否满足算法的输入要求。对图像的背景引入混合高斯模型进行建模,并借助图像的前景边缘检测描绘人体轮廓,基于Canny算子对人体轮廓进行跟踪、提取特征,以这些轮廓为基础实现人流密度的计算。将最终的提取结果输入到神经网络中,根据时间分布完成人流的预测。在算法的实现上考虑到计算需求,文中搭建了7个计算集群的Hadoop分布式计算平台,实现了算法的分布式存储和并行化计算。计算结果表明,相较于传统算法,该算法提取的人体轮廓更精细、轮廓长度更长,漏检率由13.21%降低到5.24%,人流的预测精度达到了92.43%,提升了7.12%。In response to the demand for refined and efficient human flow monitoring in scenic spots,the background modeling and contour extraction methods of video images are studied,and a human flow density prediction model for massive computing needs is built.The model needs to detect the image quality of the input model,and check whether the resolution and pixels of the image meet the input requirements of the algorithm.The Gaussian mixture model is introduced for the background of the image for modeling,and the outline of the human body is drawn with the help of the foreground edge detection of the image,and the outline of the human body is tracked and features are extracted based on the Canny operator,and the crowd density calculation is realized based on these outlines.Input the final extraction results into the neural network,and complete the prediction of the flow of people according to the time distribution.In the realization of the algorithm,considering the computing requirements,a Hadoop distributed computing platform with 7 computing clusters is built in this article to realize the distributed storage and parallel computing of the algorithm.The calculation results show that compared with the traditional algorithm,the human body contour extracted by this algorithm is finer and the contour length is longer.The missed detection rate is reduced from 13.21%to 5.24%,and the prediction accuracy of the flow of people reaches 92.43%,which is improved 7.12%.

关 键 词：边缘监测 背景建模 图像处理 HADOOP 人流预测 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]