基于深度卷积网络的路面图像裂纹检测与识别  

Pavement Image Crack Detection and Recognition Based on Deep Convolutional Network

作　　者：汪荣娟 王子宁 陈嘉艺 李世新 WANG Rong-juan;WANG Zi-ning;CHEN Jia-yi;LI Shi-xin(School of Mechanical Engineering,Shenyang University of Technology,Shenyang 110000,China)

机构地区：[1]沈阳工业大学机械工程学院,辽宁沈阳110000

出　　处：《机械工程与自动化》2022年第6期22-24,共3页Mechanical Engineering & Automation

基　　金：辽宁省教育厅项目(青年项目)(LQGD2020016)。

摘　　要：针对道路裂纹人工检测精度低、效率低、误差高的问题,提出了一种基于深度卷积网络的路面图像裂纹检测与识别方法。对采集到的混凝土裂纹和砂岩裂纹图像的数据集,利用ResNet-34模型对其进行数据处理,并调整了学习策略的参数设置。将该模型与其他三种神经网络模型进行比较,通过试验得出:基于深度卷积网络的模型在路面图像裂纹检测与识别方面具有很好的准确性并且有效降低了检测误差。Aiming at the problems of low precision,low efficiency and high error in manual detection of road cracks,a method of road crack detection and recognition based on depth convolution network is proposed.Through the collected data set of concrete crack and sandstone crack images,ResNet-34 model is used to process the data and adjust the parameter settings of the learning strategy,and the model is compared with other three neural networks.The experiment shows that the optimized model has good accuracy in pavement image crack detection and recognition and effectively reduces the detection error.

关 键 词：深度卷积神经网络 路面图像 裂纹检测 识别 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]