摘要
为了精确、快速地识别道路表面的质量情况,及时进行维保,文中提出了一种基于红外光谱分析与改进CNN的智能道路质量检测技术。该方案根据路面图像的红外光谱特征,通过红外光谱识别技术来判断路面是否存在裂缝。为了精准地检测出路面上的裂缝,同时还对CNN模型的ResNet34进行改进,提出了一种编码器-解码器体系结构的检测网络。在仿真实验中,所提出的道路质量检测方法得到的准确率为98.70%,召回率为99.00%,F1值为98.34%,mIoU为76.24%,其检测所需的时间消耗仅为0.51 s,优于SegNet、U-Net和DFN等深度学习方法。
In order to accurately and quickly identify the quality of road surfaces and carry out maintenance in a timely manner,this paper proposes an intelligent road quality detection technology based on infrared spectroscopy analysis and improved CNN.This scheme uses infrared spectral recognition technology to determine whether there are cracks on the road surface based on the infrared spectral features of the road surface image.In order to accurately detect cracks on the outlet surface and improve ResNet34 of the CNN model,an encoder decoder architecture detection network is proposed.In the simulation experiment,the accuracy of the road quality detection method proposed in the article is 98.70%,the recall rate is 99.00%,the F1 value is 98.34%,and the mIoU is 76.24%.Its detection time consumption is only 0.51 seconds,which is superior to deep learning methods such as SegNet,U-Net,and DFN.
作者
熊涛
席恩伟
闫文佳
XIONG Tao;XI Enwei;YAN Wenjia(Yunnan Communications Investment&Construction Group CO.,LTD.;Yunnan Key Laboratory of Digital Communications,Yunnan Kunming 650228,China;Yunnan Yunling Highway Engineering Consulting CO.,LTD.,Yunnan Kunming 650220,China)
出处
《工业仪表与自动化装置》
2023年第5期98-102,共5页
Industrial Instrumentation & Automation
基金
云南省数字交通重点实验室(202205AG070008)
云南交投科技创新计划项目(YCIC-YF-2021-11)。
