一种基于ASPPUnet的道路裂缝检测模型  

A Pavement Crack Detection Model Based on ASPPUnet

作　　者：曹一冰[1] 张江水[1] 张政[1] 赵鑫科 CAO Yibing;ZHANG Jiangshui;ZHANG Zheng;ZHAO Xinke(Information Engineering University,Zhengzhou 450001,China)

机构地区：[1]信息工程大学,河南郑州450001

出　　处：《测绘科学技术学报》2025年第1期49-56,共8页Journal of Geomatics Science and Technology

基　　金：国家重点研发计划项目(2021YFB3900900)。

摘　　要：为了更加精确高效地对道路裂缝进行分割提取,提出一种基于多尺度特征与上下文信息融合的ASPPUnet道路裂缝检测模型。ASPPUnet通过U形编码解码器进行多尺度特征的提取,通过引入ASPP模块进行不同范围上下文信息的融合;同时模型还引入了深度可分离卷积模块,用以实现模型的轻量化;采用融合Dice和交叉熵的损失函数,均衡模型的查全率和查准率;采用动态数据集增广方法,使得模型在小数据集上也能实现良好的检测效果。通过与Unet等模型的实验对比可以看出,ASPPUnet拥有更好的检测效果和可塑性,具有较好的应用价值。In order to segment and extract pavement cracks more accurately and efficiently,an ASPPUnet pavement crack detection model is proposed based on multi-scale features and context information fusion in this paper.In ASPPUnet,multi-scale features are extracted through U-shaped Encoder-Decoder,and context information of different ranges is fused by introducing ASPP module.And depthwise separable convolution is introduced into the model to realize lightweight of the model.The loss function of fusion Dice and cross entropy is adopted to balance the recall and precision of the model.Meanwhile,dynamic dataset augmentation method is adopted to achieve good detection effect on small data sets.Compared with Unet and other models,ASPPUnet has better detection effect and plasticity,which leads to better application value.

关 键 词：裂缝检测 图像分割 深度可分离卷积 损失函数 ASPP模块 Unet模型 

分 类 号：P258[天文地球—测绘科学与技术]