基于卷积神经网络的实时人群密度估计  

Real-Time Crowd Density Estimation Based on Convolutional Neural Networks

作　　者：李白萍[1] 韩新怡 吴冬梅[1] LI Baiping;HAN Xinyi;WU Dongmei(College of Communication and Information Engineering,Xi’an University of Seience and Technology,Xi’an Shaanxi 710054,China)

机构地区：[1]西安科技大学通信与信息工程学院,陕西西安710054

出　　处：《图学学报》2018年第4期728-734,共7页Journal of Graphics

基　　金：陕西省重点研发计划项目(2017GY-095)

摘　　要：针对传统实时人群密度估计方法存在误差大、分类效果不佳等缺陷,提出了基于卷积神经网络的实时人群密度估计方法。通过对比4种常见网络结构:Alex Net、VGGNet、Goog Le Net和Res Net的准确度与实时性,选择综合性较好的Goog Le Net作为人群密度估计的模型,利用关键帧截取技术实现人群密度的实时估计并简要分析人群密度特征图。最后用实例验证了该方法的实时性与准确度,证明了其可行性。In response to the deficiencies such as big error and poor performance in the traditional method of real-time crowd density estimation,a new one based on CNN is proposed.By comparing the accuracy and real-time of four common network structures-AlexNet,VGGNet,GoogLeNet,and ResNet,the GoogLeNet which has relatively better comprehensive performance is chosen as the model for crowd density estimation.We used the key-frame extraction technology to realize real-time crowd density estimation and briefly analyze the crowd density feature map.Finally,examples are analyzed to verify the real time,accuracy,and feasibility of this new method of real-time crowd density estimation.

关 键 词：人群密度 卷积神经网络 视频处理 实时估计 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]