光学精密工程2025，Vol.33Issue(14)：2262-2277,16.DOI:10.37188/OPE.20253314.2262
全局感知与多尺度特征融合的城市道路语义分割
Urban road semantic segmentation with global awareness and multi-scale feature fusion
摘要
Abstract
Semantic segmentation plays an irreplaceable role in autonomous driving and intelligent trans-portation systems.However,current segmentation networks often suffer from challenges such as blurred object boundaries,mutual occlusions between objects,and significant variations in object scales,which hinder segmentation accuracy.To address these issues,this paper proposed a city road scene semantic seg-mentation network that integrates global context awareness and multi-scale feature fusion.To mitigate the problem of blurred segmentation boundaries,a Global Awareness Module(GAM)was designed to en-hance interaction between spatial and channel-wise information,enabling the network to capture compre-hensive global context.For handling object occlusion and improving recognition of partially obscured re-gions,a Multi-Scale Feature Fusion Module(MSFF)was introduced,which effectively integrated contex-tual cues from different receptive fields to ensure segmentation accuracy for objects of varying sizes.Fur-thermore,a comprehensive Multi-constraint Feature Smoothing Loss was employed to enforce spatial co-herence and semantic consistency,guiding the model toward a more optimal solution by refining feature distributions around object boundaries and within complex scenes.Extensive experiments are conducted on two benchmark datasets.On the Cityscapes dataset,the proposed method achieves mIoU improvements by 0.5％,0.9％,and 1.7％under different input resolutions.On the ADE20K dataset,an mIoU gain of 2.1％is observed.Compared with existing semantic segmentation models,the proposed approach demon-strates superior performance in urban road scene understanding,particularly in terms of boundary delinea-tion,and occlusion robustness.关键词
深度学习/图像处理/语义分割/特征融合/损失函数Key words
deep learning/image processing/semantic segmentation/feature fusion/loss function分类
信息技术与安全科学引用本文复制引用
邬开俊,张治瑞,汪滢,安立伟..全局感知与多尺度特征融合的城市道路语义分割[J].光学精密工程,2025,33(14):2262-2277,16.基金项目
甘肃省自然科学基金项目(No.23JRRA913) （No.23JRRA913）
内蒙古自治区重点研发与成果转化计划项目(No.2023YFDZ0043,No.2023YFDZ0054,No.2023YFSH0043) （No.2023YFDZ0043,No.2023YFDZ0054,No.2023YFSH0043）
兰州交通大学重点研发项目资助(No.ZDYF2304) （No.ZDYF2304）
