基于卷积神经网络的桥梁病害识别与裂缝特征测量方法

打开文本图片集

关键词：深度学习；病害识别；裂缝特征；卷积神经网络；贝叶斯优化；语义分割中图分类号：TP391 文献标志码：A 文章编号：1005-8249（2025）06-0140-10DOI：10.19860/j.cnki.issn1005-8249.2025.06.024

Abstract：Toimprovetheefciencyandaccuracyofbridgedefectidentification，adeeplearmingalgorithmbasedonbayesian optimizationandconvolutionalneuralnetworksisproposedforidentificationoffourtypesofbridgedefects：pockmarkedsurface， cracks，exposed rebar，and spalling. For cracks，DeepLabv3 + with Mobilenet - v2 as the feature extraction network is establishedas thesemantic segmentation modelforcrack images.Theresultsshowthatthe proposed methodmaintains nearlythe sameaccuracyandrobustnessin identifying thefourtypesof bridgedefects，whilereducingtraining timebyaproximately （204号 80% .Through the semantic segmentation model and image procesing techniques，precise segmentation of cracks and automatic extractionofgeometricinformationareachieved，withanMIoUofO.95forcracksegmentation.Theeffcientandaccurate identification of bridge diseasesprovides more precise data references for bridgeperformance prediction and analysis.

Key words：deep learning；damage identification；crack features；convolutional neural network；Bayesianoptimization; semantic segmentation

0 引言

基于深度学习的病害识别方法已成为桥梁检测的新兴手段之一，特别是基于卷积神经网络（CNN）的图像处理技术，具有效率高、可达性广等优点。（剩余9995字）