提出一种基于深度学习网络的路面坑洼嵌入式检测系统的设计方案,旨在提升路面养护巡检的效率、降低公路维护费用。该系统首先对大量的样本数据进行网络模型训练,获取最优模型;然后将最优模型部署到英伟达TX2中;最后通过车载摄像头自动检测路面坑洼,并将坑洼信息报送给路面养护部门,实现公路坑洼的自动化巡检。
Abstract
In the paper,the design scheme of a pavement pothole embedded detection system based on deep learning network is proposed,which aims to improve the efficiency of road surface maintenance inspection and reduce highway maintenance costs.The system first trains the network model through a large number of sample data,obtains the optimal model,and then deploys the optimal model to NVIDIA TX2,and finally automatically detects the road potholes through the on-board camera,and submits the road pothole information to the road maintenance department to realize the automatic inspection of the road potholes.
关键词
路面坑洼检测 /
英伟达TX2 /
深度学习 /
YOLOv5S
Key words
pavement potholes detection /
NVIDIA TX2 /
deep learning /
YOLOv5S
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] 高建贞.基于图像分析的道路病害自动检测研究[D].南京:南京理工大学,2003.
[2] 赵珊珊,何宁.基于卷积神经网络的路面裂缝检测[J].传感器与微系统,2017,36(11):135-138.
[3] 郎洪,温添,陆键,等.基于深度学习的三维路面裂缝类病害检测方法[J].东南大学学报(自然科学版),2021,51(1):53-60.
[4] 祝一帆,王海涛,李可,等.一种高精度路面裂缝检测网络结构Crack U-Net[J].计算机科学,2022,49(1):204-211.
[5] Ultralytics.Yolov5[EB/OL].[2022-02].https://github.com/ultralytics/yolov5.
[6] 董红召,方浩杰,张楠.旋转框定位的多尺度再生物品目标检测算法[J].浙江大学学报(工学版),2022,56(1):16-25.
[7] ZHANG H,WANG Y,DAYOUB F,et al.VarifocalNet:An IoU-aware Dense Object Detector,2020.
[8] LIN T-Y,GOYAL P,GIRSHICK R,et al.Focal Loss for Dense Object Detection,2017.
[9] NVIDIA Jetson TX2赋能终端设备实现人工智能[J].单片机与嵌入式系统应用,2017,17(5):62.
[10] 张永合.Jetson TX2平台的最小应用系统硬件设计[J].单片机与嵌入式系统应用,2019,19(10):52-54.
PDF(1420 KB)
