GMFNet:全局多尺度和多级别的特征融合语义分割网络  

GMFNET:GLOBAL MULTI-SCALE AND MULTI-LEVEL FEATURE FUSION NETWORK FOR SEMANTIC SEGMENTATION

作　　者：陈金令 赵成明 李洁 Chen Jinling;Zhao Chengming;Li Jie(School of Electrical Information,Southwest Petroleum University,Chengdu 610500,Sichuan,China)

机构地区：[1]西南石油大学电气信息学院,四川成都610500

出　　处：《计算机应用与软件》2025年第4期311-318,334,共9页Computer Applications and Software

基　　金：成都市科技局创新创业资助项目(2018YF0500893GX)。

摘　　要：语义分割网络在编码器-解码器中融合高低水平特征存在以下问题:(1)在空间和通道中特征提取无法同步,导致特征组合无法获取全局上下文信息;(2)特征融合无法充分利用高低水平特征图像,导致语义边界模糊。设计全局空洞空间金字塔池化,该结构不仅在空间上提取多尺度信息和通道上对图像信息充分利用,还增强编码器阶段的特征重用。设计特征融合注意力模块,在编码器中连接不同阶段的高低水平特征和新特征。实验表明,该算法在Cityscapes数据集上达到了77.92%mIoU。For the semantic segmentation network,the following problems exist in the fusion of low-level and high-level feature in the encoder-decoder:(1)feature extraction in space and channel cannot be synchronized,resulting in feature combinations that cannot obtain global context information;(2)feature fusion cannot be fully utilized low-level and high-level feature images,resulting in blurred semantic boundaries.The global atrous spatial pyramid pooling was designed.This structure not only extracted multi-scale information in space and utilized image information in channels,but also enhanced feature reuse in the encoder stage.A feature fusion attention module was designed to connect low-level and high-level features and new features at different stages in the encoder.Experiments show that the algorithm achieves 77.92%mIoU on the Cityscapes dataset.

关 键 词：语义分割 卷积神经网络 全局上下文信息 特征融合 编码器-解码器 

分 类 号：TP394.41[自动化与计算机技术—计算机应用技术]