基于自编码器的深度对抗哈希方法在覆冰电网图像检索中的应用  

Deep Adversarial Hashing Method Based on Auto-encoder for Frozen Power Line Image Retrieval

作　　者：强彦 何龙 张丽敏 王继宗 吕军[3] 史国华 陈琪 QIANG Yan;HE Long;ZHANG Limin;WANG Jizong;LYU Jun;SHI Guohua;CHEN Qi(College of Information and Computer,Taiyuan University of Technology,Taiyuan 030024,China;State Grid Jincheng Power Supply Company,Jincheng Shanxi 048000,China;Department of Computer Science and Technology,Lyuliang University,Lyuliang Shanxi 033001,China;College of Engineering,Oregon State University,Corvallis,Oregon 97331,USA)

机构地区：[1]太原理工大学信息与计算机学院,太原030024 [2]国网山西省电力公司晋城供电公司,山西晋城048000 [3]吕梁学院计算机科学与技术系,山西吕梁033001 [4]俄勒冈州立大学工程学院,美国俄勒冈州科瓦利斯97331

出　　处：《太原理工大学学报》2020年第4期485-494,共10页Journal of Taiyuan University of Technology

基　　金：山西省电力公司科技项目(5205E0160009)。

摘　　要：为了提高覆冰电网图像的检索性能,提出一种基于自编码器的深度对抗哈希方法。首先,通过在现有的生成器和鉴别器之上添加新的编码鉴别器以鼓励生成的图像样本更好地表示真实数据分布。其次,构建哈希编码网络以学习生成紧凑的二进制哈希码。我们在WGAN-GP损失的基础上引入了新的基于长尾柯西分布的交叉熵损失和量化损失函数以优化汉明空间检索性能。实验结果表明,该深度对抗哈希方法能够通过编码鉴别器和柯西损失函数解决模式崩溃和图像模糊的问题,图像检索性能相比于其他方法有明显提高。In order to improve the retrieval performance of frozen power line images,a deep adversarial hashing method based on auto-encoder was proposed.First,a new code discriminator was added on top of the existing generator and discriminator to encourage the generated image samples to better represent the real data distribution.Second,a hash coding network was built to learn to generate a compact binary hash code.On the basis of the WGAN-GP loss,we introduced new cross-entropy loss and quantization loss functions based on the long-tail Cauchy distribution to optimize the Hamming space retrieval performance.The experimental results show that the deep adversarial hashing method can solve the problems of mode collapse and image blurring through the code discriminator and Cauchy loss functions,and the image retrieval performance is significantly improved with respect to other methods.

关 键 词：覆冰图像 图像检索 哈希编码 生成对抗网络 长尾柯西分布 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]