摘要
交通的快速发展给人们生活带来很多的便利,但由于司机违规行为而造成的交通事故也频频发生。针对司机在道路行驶过程中会经常出现抽烟和玩手机两种违规行为,通过对深度学习中的经典网络卷积神经网络进行改进。提出神经网络融合的模型,作用于真实的、标注的司机违规行为数据集。与经典的LetNet和AlexNet进行对比,实验结果表明,提出的方法能够有效地识别司机违规行为。
The rapid development of traffic has brought a lot of convenience to people s lives, but traffic accidents caused by drivers illegal behaviors also occur frequently. In view of two drivers illegal behaviors, smoking and playing with mobile phones, this paper proposed a neural network fusion model by improving the classical convolution neural network in depth learning. This method was applied to the real and tagged data sets of drivers irregularities and compared with the LetNet and AlexNet. The comparison experiment results show that the method can effectively identify drivers illegal behaviors.
作者
李俊俊
杨华民
张澍裕
李松江
Li Junjun;Yang Huamin;Zhang Shuyu;Li Songjiang(Internet of Things and Intelligent Structure Room,R ＆ D Center,Beijing Aerospace Institute of Control Devices,Beijing 100854,China;School of Computer Science and Technology,Changchun University of Science and Technology,Changchun 130022,Jilin,China)
出处
《计算机应用与软件》
北大核心
2018年第12期222-227,319,共7页
Computer Applications and Software
基金
吉林省产业技术研究与开发专项项目(2016C090)
大数据与社会治理研究国家社科基金项目(17BSH135)
关键词
深度学习
卷积神经网络
激活函数
行为识别
Deep learning
Convolution neural network
Activation function
Behavior recognition
