摘要
道路在长期使用情况下易产生路面损伤,进而降低交通效率,严重时甚至会危及人身和财产安全。为此设计了一种基于YOLO目标检测算法的边缘到终端的道路损伤检测系统。系统包括路侧信息采集平台、边缘计算设备、云传输系统、客户使用终端。首先设计了一个轨道式巡检机器人作为持续的信息采集平台完成道路损坏检测任务;该机器人沿路侧轨道移动进行巡检,配备光学摄像头、Jetson NX、4G无线路由器用于信息采集、边缘计算和通信;云服务器用于临时存储和传输机器人平台收集的信息;最终将信息发送给客户端,用于道路损坏检测的可视化显示;路面损伤检测算法基于YOLOv5框架训练得到,在RDD2020数据集上对比了YOLOv5、YOLOv4和YOLOv4-tiny的效果,并基于NVIDIA Jetson NX平台实现了模型部署和模型优化。实验结果表明,该系统可以实现道路损伤检测的实时显示,具有一定的工程应用价值。
Road are prone to damage under long-term use,which reduces traffic efficiency,and even endangers personal and property safety in serious cases.Therefore,an edge-to-client road damage detection system based on YOLO object detection algorithm is designed.The system includes roadside information collection platform,edge computing equipment,cloud transmission system,and client terminal.Firstly,a track-mounted inspection robot is designed as a continuous information collection platform to complete the road damage detection task.The robot moves along roadside tracks for inspection,and is equipped with optical cameras,Jetson NX,and 4G wireless routers for information collection,edge computing and communications.Cloud server is used to temporarily store and transmit the information collected by the robot platform.The information is finally sent to the client terminal for visual display of road damage detection.Road damage detection algorithm trained based on YOLOv5 framework.The effects of YOLOv5,YOLOv4,and YOLOv4-tiny on the RDD2020 data set are compared,and model deployment and model optimization are implemented based on the NVIDIA Jetson NX platform.Experimental results show that the system can realize real-time display of road damage detection and has certain engineering application value.
作者
李明珏
张高兴
Li Mingjue;Zhang Gaoxing(Kunming Shipborne Equipment Research&Test Center,Kunming 650051,China)
出处
《机电工程技术》
2023年第9期291-295,共5页
Mechanical & Electrical Engineering Technology
关键词
目标识别
破损检测
YOLO
边缘计算
机器人
object detection
road damage detection
YOLO network
edge computing
robot
