Automatic road extraction framework based on codec network  

基于编解码网络的自动道路提取框架

作　　者：WANG Lin SHEN Yu ZHANG Hongguo LIANG Dong NIU Dongxing 王霖;沈瑜;张泓国;梁栋;牛东兴(兰州交通大学电子与信息工程学院,甘肃兰州730070;中铁科学研究院有限公司,四川成都610036)

机构地区：[1]School of Electronic and Information Engineering,Lanzhou Jiaotong University,Lanzhou 730070,China [2]China Railway Scientific Research Institute Co.,Ltd.,Chengdu 610036,China

出　　处：《Journal of Measurement Science and Instrumentation》2024年第3期318-327,共10页测试科学与仪器（英文版）

基　　金：supported by National Natural Science Foundation of China(No.61864025);2021 Longyuan Youth Innovation and Entrepreneurship Talent(Team),Young Doctoral Fund of Higher Education Institutions of Gansu Province(No.2021QB-49);Employment and Entrepreneurship Improvement Project of University Students of Gansu Province(No.2021-C-123);Intelligent Tunnel Supervision Robot Research Project(China Railway Scientific Research Institute(Scientific Research)(No.2020-KJ016-Z016-A2);Lanzhou Jiaotong University Youth Foundation(No.2015005);Gansu Higher Education Research Project(No.2016A-018);Gansu Dunhuang Cultural Relics Protection Research Center Open Project(No.GDW2021YB15).

摘　　要：Road extraction based on deep learning is one of hot spots of semantic segmentation in the past decade.In this work,we proposed a framework based on codec network for automatic road extraction from remote sensing images.Firstly,a pre-trained ResNet34 was migrated to U-Net and its encoding structure was replaced to deepen the number of network layers,which reduces the error rate of road segmentation and the loss of details.Secondly,dilated convolution was used to connect the encoder and the decoder of network to expand the receptive field and retain more low-dimensional information of the image.Afterwards,the channel attention mechanism was used to select the information of the feature image obtained by up-sampling of the encoder,the weights of target features were optimized to enhance the features of target region and suppress the features of background and noise regions,and thus the feature extraction effect of the remote sensing image with complex background was optimized.Finally,an adaptive sigmoid loss function was proposed,which optimizes the imbalance between the road and the background,and makes the model reach the optimal solution.Experimental results show that compared with several semantic segmentation networks,the proposed method can greatly reduce the error rate of road segmentation and effectively improve the accuracy of road extraction from remote sensing images.基于深度学习的道路提取是近十年来语义分割的热点之一。本研究提出一种基于编解码网络的遥感图像道路自动提取框架。首先,将预训练好的ResNet34网络迁移到U-Net网络并替换其编码结构,加深网络层数,降低道路分割的错误率和细节丢失。其次,利用通道注意力机制,对编码器上采样得到的特征图进行信息选择,将目标特征进行权重优化,在增强图像目标区域特征的同时抑制背景及噪声区域,优化背景复杂度高的遥感图像特征提取效果。最后,提出一种自适应sigmoid损失函数,解决道路和背景占比不平衡的问题,使模型在客观程度上达到最优解。实验结果表明,与其他几种语义分割网络方法相比。本研究所提的方法有效降低了道路提取的错误率,提高了道路图像的分割精度。

关 键 词：remote sensing image road extraction ResNet34 U-Net channel attention mechanism sigmoid loss function 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] U491[自动化与计算机技术—控制科学与工程]