基于嵌入式GPU的车流量检测系统  

Vehicle Flow Detection System Based on Embedded GPU

作　　者：储泽楠[1] 韩毅[1] 宋倍倍 CHU Zenan;HAN Yi;SONG Beibei(Anyang Institute of Technology,Anyang Henan 455000;Anyang Quality and Technical Supervision,Inspection and Testing Center,Anyang Henan 455000)

机构地区：[1]安阳工学院,河南安阳455000 [2]安阳市质量技术监督检验测试中心,河南安阳455000

出　　处：《河南科技》2020年第5期29-31,共3页Henan Science and Technology

基　　金：河南省科技攻关项目(182102210197)。

摘　　要：交通拥堵是当今世界交通领域面临的主要问题之一,如何通过现有的交通设备获取更加精准的交通信息是亟待解决的问题。图像识别技术在智能交通系统中有着广泛的应用,基于深度学习的车流量检测技术是智能交通的重要组成部分。本项目设计了一个基于嵌入式GPU的智能车流量检测系统,该系统架设在NVIDIA JetsonTX2平台上,采用基于深度学习YOLO v3的车辆检测模型,检测道路上的车辆目标,设置兴趣区域,对检测到的目标进行识别计数,实现对交通视频的实时车流量检测。试验验证分析表明,该系统具有较高的检测精度。Traffic congestion is one of the main problems in the world's transportation field,how to obtain more accu rate traffic information through existing transportation equipment is an urgent problem.Image recognition technology has been widely used in intelligent transportation systems,vehicle flow detection technology based on deep learning is an important part of intelligent transportation.This project designed an intelligent vehicle flow detection system based on embedded GPU,which was built on the NVIDIA JetsonTX2 platform,used a YOLO v3 vehicle detection model based on deep learning to detect vehicle targets on the road,set areas of interest,identify and count the detect ed targets,and realize real-time traffic flow detection on traffic videos.Test verification analysis shows that the sys tem has high detection accuracy.

关 键 词：深度学习 车流量检测 YOLO Jetson TX2 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]