《计算机应用》唯一官方网站 ›› 2020, Vol. 40 ›› Issue (2): 566-570.DOI: 10.11772/j.issn.1001-9081.2019091625

• 第七届CCF大数据学术会议 • 上一篇    下一篇

基于循环一致性对抗网络的数码迷彩伪装生成方法

滕旭1, 张晖2(), 杨春明1, 赵旭剑1, 李波1   

  1. 1.西南科技大学 计算机科学与技术学院,四川 绵阳 621010
    2.西南科技大学 理学院,四川 绵阳 621010
  • 收稿日期:2019-08-20 修回日期:2019-09-24 接受日期:2019-10-09 发布日期:2019-10-14 出版日期:2020-02-10
  • 通讯作者: 张晖
  • 作者简介:滕旭(1996—),男,江苏徐州人,硕士研究生,主要研究方向:深度学习、生成对抗式网络
    杨春明(1980-),男,云南华坪人,副教授,硕士,主要研究方向:数据挖掘、自然语言处理
    赵旭剑(1984—),男,山西运城人,副教授,博士,主要研究方向:机器学习、自然语言处理
    李波(1977—),男,四川绵阳人,讲师,博士研究生,主要研究方向:信息安全、信息过滤。
  • 基金资助:
    教育部人文社科基金资助项目(17YJCZH260);赛尔网络下一代互联网技术创新项目(NGII20180403)

Digital camouflage generation method based on cycle-consistent adversarial network

Xu TENG1, Hui ZHANG2(), Chunming YANG1, Xujian ZHAO1, Bo LI1   

  1. 1.School of Computer Science and Technology,Southwest University of Science and Technology,Mianyang Sichuan 621010,China
    2.School of Science,Southwest University of Science and Technology,Mianyang Sichuan 621010,China
  • Received:2019-08-20 Revised:2019-09-24 Accepted:2019-10-09 Online:2019-10-14 Published:2020-02-10
  • Contact: Hui ZHANG
  • About author:TENG Xu, born in 1996, M. S. candidate. His research interests include deep Learning, generative adversarial network.
    YANG Chunming, born in 1980, M. S., associate professor. His research interests include data mining, natural language processing.
    ZHAO Xujian, born in 1984, Ph. D., associate professor. His research interests include machine learning, natural language processing.
    LI Bo, born in 1977, Ph. D. candidate, lecturer. His research interests include information safety, information filtering.
  • Supported by:
    the Foundation of Humanities and Social Sciences of Ministry of Education(17YJCZH260);the CERNET Innovation Project for Next Generation Internet Technology(NGII20180403)

摘要:

针对传统的数码迷彩生成方法无法根据背景实时生成数码迷彩的问题,提出一种基于循环一致性对抗网络的数码迷彩生成方法。首先,使用密集连接卷积网络提取图像特征,将学习到的数码迷彩特征映射到背景图像中;其次,加入颜色保持损失来提高数码迷彩的生成质量,保证生成的数码迷彩与周围的背景颜色相一致;最后,在判别器中加入自归一化神经网络以提高模型对噪声的鲁棒性。由于缺乏数码迷彩伪装效果的客观评价标准,采用边缘检测算法与结构相似性(SSIM)算法对生成的数码迷彩的伪装效果进行评估。实验结果表明,该方法在自制数据集上生成的数码迷彩伪装的SSIM得分比已有算法的得分降低了30%以上,验证了它在数码迷彩生成任务上的有效性。

关键词: 深度学习, 生成对抗式网络, 数码迷彩, 边缘检测, 密集连接卷积网络

Abstract:

Traditional methods of generating digital camouflages cannot generate digital camouflages based on the background information in real time. In order to cope with this problem, a digital camouflage generation method based on cycle-consistent adversarial network was proposed. Firstly, the image features were extracted by using densely connected convolutional network, and the learned digital camouflage features were mapped into the background image. Secondly, the color retention loss was added to improve the quality of generated digital camouflages, ensuring that the generated digital camouflages were consistent with the surrounding background colors. Finally, a self-normalized neural network was added to the discriminator to improve the robustness of the model against noise. For the lack of objective evaluation criteria for digital camouflages, the edge detection algorithm and the Structural SIMilarity (SSIM) algorithm were used to evaluate the camouflage effects of the generated digital camouflages. Experimental results show that the SSIM score of the digital camouflage generated by the proposed method on the self-made datasets is reduced by more than 30% compared with the existing algorithms, verifying the effectiveness of the proposed method in the digital camouflage generation task.

Key words: deep learning, generated adversarial network, digital camouflage, edge detection, densely connected convolutional network

中图分类号: