改进特征融合的实时语义分割方法  

Real-time Semantic Segmentation Method Based on Improved Feature Fusion

作　　者：王小玉[1] 李志斌 WANG Xiaoyu;LI Zhibin(School of Computer Science and Technology,Harbin University of Science and Technology,Harbin 150080,China)

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《哈尔滨理工大学学报》2023年第6期87-94,共8页Journal of Harbin University of Science and Technology

基　　金：国家自然科学基金(61772160);黑龙江省教育厅科学技术研究项目(12541177).

摘　　要：针对实时语义分割任务中需要同时兼顾位置信息和语义信息的问题,提出一种改进特征融合的实时语义分割方法。该方法由卷积神经网络、轻量级注意力模块(light attention module,LAM)和双通道特征融合模块(bilateral feature fusion module,BFFM)组成。首先,使用卷积神经网络结合轻量级注意力模块快速提取图像的位置信息和语义信息。然后,使用双通道特征融合模块指导位置信息和语义信息的特征图融合。所提方法在CamVid上,平均交并比达到67.8%,分割速度可达到52.6帧/s。在Cityscapes上,平均交并比达到73.5%,分割速度可达到31.8帧/s。实验结果表明,提出的分割方法满足分割的准确性和实时性要求,能够适用于实时语义分割任务中。Aiming at the problem that both location information and semantic information need to be considered in real-time semantic segmentation tasks,we proposed a real-time semantic segmentation method based on improved feature fusion.The method consists of a convolution neural network,a light attention module(light attention module,LAM),and a bilateral feature fusion module(bilateral feature fusion module,BFFM).Firstly,quickly extract the location information and semantic information of the image by combining the convolutional neural network with the lightweight attention module.Then,the bilateral feature fusion module is used to guide the feature fusion of location information and semantic information.The results of the method on the data set of CamVid,mIoU reached 67.8%,and the running speed reached 52.6 fps.On the data set of Cityscapes,mIoU reached 73.5%,and the running speed reached 31.8 fps.The results show that the proposed segmentation method meets the accuracy and real-time requirements of segmentation,and can be applied to real-time semantic segmentation tasks.

关 键 词：图像处理 卷积神经网络 实时语义分割 特征融合 注意力机制 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]