作者：李光东，程进，周子肖，刘宝宇，王骏琦，史立根，姚宇军，常创创，李文月，邹小平

单位：北京信息科技大学北京市传感器重点实验室，北京 100101

中图分类号：TP391.41                 

文献标识码：A                 

文章编号：1006-883X(2020)02-0007-05

收稿日期：2019-12-04

 

摘要：随着公交车和出租车的数量不断增加，司机在开车时的不规范驾驶动作造成很多交通事故。为了有效降低这种交通事故发生的可能性，设计了一个基于深度学习的不规范驾驶行为智能识别系统。本系统将摄像头安装在车内并实时采集司机驾驶状态的图像，控制器与后台电脑服务器远程连接，服务器通过网络通信接收司机图像并通过深度学习对采集的司机图像进行识别并处理，再把处理的结果传给控制器。这样司机在开车时发生喝水、打电话、玩手机等7种危险动作行为时，就可以通过控制器及时对司机进行智能安全语音提示，后台也可以实时监督和警示司机。

关键词：不规范驾驶动作；网络通信；深度学习；图像处理；智能安全语音提示

 

Irregular Driving Behavior Intelligent Recognition System Based on Deep Learning

LI Guang-dong, CHENG Jin, ZHOU Zi-xiao, LIU Bao-yu, WANG Jun-qi, SHI Li-gen, YAO Yu-jun, CHANG Chuang-chuang, LI Wen-yue, ZOU Xiao-ping 

Beijing Key Laboratory of Sensors, Beijing Information Science and Technology University, Beijing 100101, China

Abstract: With the increasing number of buses and taxis, drivers irregular driving behaviors while driving cause many traffic accidents. In order to effectively reduce the possibility of such traffic accidents, an irregular driving behavior intelligent recognition system is designed based on deep learning. A camera with a controller is installed in the car to collect images of the drivers driving status in real time, and the controller is remotely connected to the background computer server. The server receives the collected images through network communication, identifies and processes these images based on deep learning, and then transmits the processed results to the controller. In this way, 7 kinds of dangerous actions while driving, such as drinking water, talking or playing on the phone and so on, can be promptly and intelligently given voice prompts to the driver under the control of the controller in the car, and the background server can also monitor and alert the driver in real time.

Key words: irregular driving actions; network communication; deep learning; image processing; intelligent safety voice prompts

 

 

备注：2020年 第26卷 第02期