基于深度学习的数字全息三维路面病害数据识别分类方法

Identification and Classification Method of Digital Holographic 3D Pavement Disease Data based on Deep Learning

为弥补二维图像道路病害识别的不足,本研究采用数字全息三维路面病害数据,设计了适用于三维路面病害数据的识别方法:对三维病害数据进行预处理,获取病害特征高度突出的三维路面数据信息;搭建适用于路面三维病害图像分类的深度学习架构,建立三维路面病害数据库,通过人工标注病害类型和分类,利用不同深度神经网络架构和模型对数据库进行训练,构建了数字全息三维路面病害自动识别算法。三维病害AI智能识别准确率达到95.7%,在路面病害智能检测的应用中取得了较好成效。

In order to make up lacking of road disease identification in 2D images,inthe research,digital holographic 3D road disease datahas been used,and a disease processing method suitable for 3D road data has been designed.The 3D disease data has been pre-processed to obtain 3D road data with highly prominent disease characteristics.A deep learning architecture suitable for the classification of 3D pavement disease imageshas been built,a 3D pavement disease database has been establishedwiththe disease type and classification beingmanually marked,the database with different deep neural network architectures and models has been trained,and a digital holographic 3D pavementdisease automatic identification algorithmhas been constructed.The accuracy rate of AI intelligent recognition of 3D diseases canreached 95.7%,and the results can achieve good results in the application of intelligent detection of road diseases.