基于双卷积神经网络的行人精细化识别  

Fine pedestrians recognition based on double convolutional neural network

作　　者：王枫[1] 刘青山[1] 

机构地区：[1]南京信息工程大学信息与控制学院,南京210044

出　　处：《中国科技论文》2017年第14期1578-1582,共5页China Sciencepaper

基　　金：江苏省高校自然科学研究重大项目(15KJA520001)

摘　　要：针对监控场景的行人属性识别问题,提出了基于双卷积神经网络的属性识别方法,给定1幅行人图像,首先对行人的头、上身、下身、鞋、包、帽子等部件采用基于单次检测器(single shot multibox detector,SSD)的深度卷积网络框架进行检测,对于检测出的部件采用VGG-weights(visual geometry group)的深度网络进行精细化的属性识别,如行人的性别、头发长度、衣着款式、背包款式等;最后设计了行人属性识别系统,实现了行人部件的自动检测与属性识别。实验表明,所提出的识别方法性能好、识别度高、系统检测时间短、应用价值高。Focused on the pedestrian attributes recognition problem in visual surveillance,apedestrian attributes recognition system based on double convolutional neural network is proposed.Given a pedestrian image,we first detect head,upper body,lower body,shoes,bag and hat etc.using SSD（single shot multiBox detector）network architecture.Then we recognize the attributes for the detected parts with VGG-weights（visual geometry group）network,including gender,hair length,clothing style,bag style and etc.A system is designed for pedestrian parts detection and attributes recognition.Experiments show that the system can recognizes pedestrian attributes fast and accurately,which presents high application value.

关 键 词：目标检测 行人属性识别 卷积神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程]