计算机科学与探索 ›› 2022, Vol. 16 ›› Issue (9): 2132-2142.DOI: 10.3778/j.issn.1673-9418.2012063

• 图形图像 • 上一篇    下一篇

融合转置卷积与深度残差图像语义分割方法

刘腊梅, 王晓娜(), 刘万军, 曲海成   

  1. 辽宁工程技术大学 软件学院,辽宁 葫芦岛 125105
  • 收稿日期:2020-12-17 修回日期:2021-04-08 出版日期:2022-09-01 发布日期:2021-04-19
  • 通讯作者: + E-mail: 1242161005@qq.com
  • 作者简介:刘腊梅(1979—),女,硕士,讲师,CCF会员,主要研究方向为图形图像处理。
    王晓娜(1994—),女,山西朔州人,硕士研究生,主要研究方向为图像与智能信息处理。
    刘万军(1959—),男,辽宁锦州人,硕士,教授,博士生导师,CCF高级会员,主要研究方向为图像与智能信息处理。
    曲海成(1981—),男,博士,副教授,硕士生导师,CCF会员,主要研究方向为遥感影像快速处理、智能大数据处理。
  • 基金资助:
    国家自然科学基金青年项目(41701479);辽宁省自然科学基金(20180550529)

Image Semantic Segmentation Method with Fusion of Transposed Convolution and Deep Residual

LIU Lamei, WANG Xiaona(), LIU Wanjun, QU Haicheng   

  1. College of Software, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Received:2020-12-17 Revised:2021-04-08 Online:2022-09-01 Published:2021-04-19
  • About author:LIU Lamei, born in 1979, M.S., lecturer, member of CCF. Her research interests include graphics and image processing.
    WANG Xiaona, born in 1994, M.S. candidate. Her research interests include image and intelligent information processing.
    LIU Wanjun, born in 1959, M.S., professor, Ph.D. supervisor, senior member of CCF. His research interests include image and intelligent information processing.
    QU Haicheng, born in 1981, Ph.D., associate professor, M.S. supervisor, member of CCF. His research interests include remote sensing image rapid processing and intelligent big data processing.
  • Supported by:
    Young Scientists Fund of National Natural Science Foundation of China(41701479);Natural Science Foundation of Liaoning Province(20180550529)

摘要:

针对深度学习图像语义分割方法中存在分割精度低、损失率高的问题,提出了融合转置卷积与深度残差图像语义分割方法。首先,为了解决神经网络深度增加引起分割精度下降、收敛速度慢的问题,设计一种深度残差学习模块来提升网络的训练效率和收敛速度;然后,为了使上采样过程与特征提取过程中特征图融合精度更高,将深度残差U-net模型中UpSampling2D和转置卷积两种上采样方式进行拼接,形成新的上采样模块;最后,针对网络训练过程中训练集与验证集之间存在的权值过度拟合问题,在网络的跳跃连接层引入Dropout,增强了网络的学习能力。在CamVid数据集上对算法的性能进行了证明,算法语义分割精度达到89.93%,损失率降到0.23,与U-net模型相比,验证集精度提升了13.13个百分点,损失率降低了1.20,优于当前的图像语义分割方法。所提出的图像语义分割新模型,综合了U-net模型的优点,使得图像语义分割精度更高,语义分割的效果更好,有效提升了算法的鲁棒性。

关键词: 图像语义分割, U-net模型, 深度残差网络, 转置卷积

Abstract:

Aiming at the problems of low segmentation accuracy and high loss of deep learning image semantic segmentation methods, image semantic segmentation method with fusion of transposed convolution and deep residual is proposed. Firstly, in order to solve the problems of decreasing segmentation accuracy and slow convergence speed caused by increasing of the depth of neural network, a deep residual learning module is designed to improve the training efficiency and convergence speed of the network. After that, in order to make the feature map fusion more accurate in upsampling and feature extraction process, two upsampling methods of UpSampling2D and transposed convolution in the deep residual U-net model are merged to form a new upsampling module. Finally, to solve the over-fitting of the weights between training set and validation set in the process of network training, Dropout is introduced in the skip connection layer of the improved network, which enhances learning ability of the model. The performance of algorithm is proven on the CamVid datasets. The semantic segmentation accuracy of the algorithm reaches 89.93% and the loss is reduced to 0.23. Compared with U-net model, the verification set accuracy is improved by 13.13 percentage points, and the loss is reduced by 1.20, which is better than the current image semantic segmentation methods. The proposed model of image semantic segmentation combines the advantages of U-net, which makes the image semantic segmentation more accurate, with better effect, and effectively improves the robustness of algorithm.

Key words: image semantic segmentation, U-net model, deep residual network, transposed convolution

中图分类号: