基于DeepLabv3+算法的城市街景语义分割算法研究  

Research on urban street view semantic segmentation algorithm based on DeepLabv3+algorithm

作　　者：陈文艺 苗宗成[1] Chen Wenyi;Miao Zongcheng(School of Electronic Information,Xijing University,Xi’an 710123,China)

机构地区：[1]西京学院电子信息学院,陕西西安710123

出　　处：《无线互联科技》2024年第9期103-105,共3页Wireless Internet Technology

摘　　要：在自动驾驶领域,城市街道场景的语义分割对于提升系统的安全性和效率至关重要。针对传统语义分割模型参数过多、泛化性能较差以及分割效果有限等问题,文章提出了一种基于改进DeepLabv3+的解决方案。此改进模型融合了轻量级MobileNetv2主干网络和SE注意力机制,优化了空洞金字塔池化(Atrous Spatial Pyramid Pooling,ASPP)模块,将其从并行结构改进成串行结构,采用深度可分离卷积结构。在Cityscapes数据集上,文章的方法取得了75.90%的平均交并比(Mean Intersection over Union,MIoU),显著提升了分割精度与计算效率。In the field of autonomous driving,semantic segmentation of urban street scenes is crucial for improving the safety and efficiency of the system.Aiming at the problems of too many parameters,insufficient portability and limited segmentation effect of traditional semantic segmentation models,this paper proposes a solution based on improved DeepLabv3+.This improved model incorporates a lightweight MobileNetv2 backbone network and the SE attention mechanism,and optimizes the Atrous spatial pyramid pooling(ASPP)module from a parallel to a serial structure with a depth-separable convolutional structure.On the Cityscapes dataset,the method in this paper achieves 75.90%MIoU,which significantly improves the segmentation accuracy and computational efficiency.

关 键 词：深度学习 语义分割 ASPP 

分 类 号：TP389.1[自动化与计算机技术—计算机系统结构]