基于CE-Net网络的水泥混凝土路面病害检测方法研究

Research on Detection Method of Cement Concrete Pavement Diseases Based on CE-Net Network

水泥混凝土路面病害检测对道路安全与交通效率具有重要意义。为此,引入了一种基于深度学习的分类算法,即上下文编码网络。该网络能够有效捕捉图像的高级特征信息,并保留更多的空间信息。同时,在病害图像分割前,对采集图像进行了灰度化处理和平滑处理。结果显示,提出的上下文编码网络模型的准确率均值高达99.68%,召回率高达98.24%,明显优于其余模型,说明所提网络模型具有显著的病害检测性能,能够应用到实际的水泥混凝土路面病害检测中,为道路路面养护工程提供可靠的技术支持。

The detection of cement concrete pavement diseases is of great significance for road safety and traffic efficiency.To this end,a deep learning based classification algorithm,namely context encoding network,has been introduced.This network can effectively capture advanced feature information of images and retain more spatial information.At the same time,before disease image segmentation,the collected images are subjected to grayscale and smoothing processing.The results show that the proposed context encoding network model has an average accuracy of 99.68%and a recall rate of 98.24%,which is significantly better than other models.This indicates that the proposed network model has significant disease detection performance and can be applied to actual cement concrete pavement disease detection,providing reliable technical support for road surface maintenance projects.