摘要
针对多尺度下的目标分割问题,提出了一种多尺度特征融合的图像语义分割方法.该方法改进了空洞空间金字塔池化(ASPP)模块,采用6种不同采样率空洞卷积并行架构增强了对输入图像信息的提取,该模块可以获取多尺度的卷积特征,从而进一步增强图像分割能力.通过在PASCAL VOC 2012数据集上进行训练和验证,达到了82.0%的平均像素准确率和71.9%的平均交并比.实验结果表明:通过改进空间维度上的ASPP模块,该方法可以达到良好的图像语义分割效果.
To handle the problem of segmenting objects at multi-scale,this paper proposed a method for image semantic segmentation based on multi-scale feature fusion,which improves the atrous spatial pyramid pooling(ASPP)module.we designed modules which employ atrous convolution in six parallel to capture multi-scale context by adopting multiple atrous rates.This module can obtain convolution features on multiple scales and further improve the performance.The training and validation on PASCAL VOC 2012 data sets showed that the experimental results achieve 82.0%of the mean pixel accuracy and 71.9%of the mean intersection over union.The results show that the model can achieve good image semantics segmentation effect by improving the ASPP method.
作者
马冬梅
杨彩锋
李鹏辉
MA Dong-mei;YANG Cai-feng;LI Peng-hui(College of Physics and Electronic Engineering,Northwest Normal University,Lanzhou 730070,Gansu,China)
出处
《西北师范大学学报（自然科学版）》
CAS
北大核心
2020年第1期44-50,共7页
Journal of Northwest Normal University(Natural Science)
基金
国家自然科学基金资助项目(61961037)
