摘要
交通场景图像语义分割是自动驾驶和智能交通等领域的重要研究问题之一。针对交通场景图像中小目标分割精度低的问题,提出一种以全卷积网络和超像素为基础的交通场景小目标改进语义分割算法。首先,基于全卷积网络获得粗糙语义分割结果,提取小目标的位置信息。其次,定位小目标各像素点对应超像素,并提取该超像素内各像素点的类别标签。最后,用超像素内最大可能性的类别为小目标重新标注语义标签。实验结果表明,所提算法能够提高小目标分割的准确率,对道路场景图像语义分割性能的提高是有效的。
Semantic segmentation of traffic scene images is one of the important research problems in fields such as autonomous driving and intelligent transportation.To address the problem of low accuracy of small target segmentation in traffic scene images,an improved semantic segmentation algorithm for small targets in traffic scenes based on full convolutional network and super pixels is proposed.Firstly,the location information of the small target is extracted from the coarse semantic segmentation result based on the full convolutional network.Secondly,the superpixel corresponding to each pixel point of the small target is located,and the category label of each pixel point within this superpixel is extracted.Finally,the semantic labels of the small targets are relabelled with the most probable categories within the superpixels.The experimental results show that the proposed algorithm can improve the accuracy of small target segmentation and can enhance the performance of semantic segmentation of traffic scene images.
作者
罗臣彦
谢新林
刘晓芳
尹东旭
LUO Chen-yan;XIE Xin-lin;LIU Xiao-fang;YIN Dong-xu(School of Electronic and Information Engineering,Taiyuan University of Science and Technology,Taiyuan 030024,China;Shanxi Key Laboratory of Advanced Control and Equipment Intelligence,Taiyuan 030024,China;Shanxi Cloud Era Technology Company Lto,Taiyuan 030006,China)
出处
《太原科技大学学报》
2024年第3期299-305,共7页
Journal of Taiyuan University of Science and Technology
基金
国家自然科学基金青年基金(62006169)
山西省自然科学基金(201901D211304,201903D121130)。
关键词
语义分割
全卷积网络
超像素
交通场景
深度学习
semantic segmentation
fully convolutional network
superpixels
traffic scene
deep learning
