基于循环一致对抗网络的玉米灰斑病图像迁移方法研究

doi:10.13733/j.jcam.issn.2095-5553.2023.02.023

摘要/Abstract

摘要： 针对玉米病害图像采集困难，特别是灰斑病表现差异性较多问题，提出一种基于循环一致对抗网络（CycleGAN）的玉米灰斑病图像生成算法，通过病害图像迁移，使得健康的作物图像可以生成患病作物图像。此方法首先通过特征提取分别提取出健康玉米图像特征和灰斑病图像特征；然后把两种特征图像输入到CycleGAN的生成器Gs中，结合生成器中的残差网络提高图像传输时的准确性，利用两个判断器判断生成的图像是否一致；最后通过对健康玉米图像进行病害迁移得到所需的玉米灰斑病图像。试验结果表明：与VAE、GAN的图像进行迁移比较，结构相似SSIM值整体分别提升50.434%、18.762%，均方误差MSE值整体减少12.891%、9.558%；直观效果上CycleGAN迁移后的不同病害程度的玉米灰斑病效果更好，因此使用CycleGAN网络生成的玉米灰斑病图像更准确。

关键词: 玉米灰斑病, 病害迁移, 残差网络, 循环一致对抗网络（CycleGAN）, 数据扩充

Abstract: Aiming at the difficulty of image acquisition of maize disease, especially the problem of many different manifestations of gray spot disease, an image generation algorithm of corn gray spot disease was proposed based on cyclic uniform countermeasure network (CycleGAN), which could make healthy crop images through the migration of disease images. This method extracted the healthy corn and the gray spot disease image features through feature extraction. Then, the two feature images were input into CycleGANs generator Gs. The residual network in the generator was combined to improve the accuracy of image transmission, and the two judges were used to judge whether the generated images were consistent. Finally, the desired corn gray leaf spot image was obtained by disease migration on the healthy corn image. The results showed that compared with VAE and GAN images, SSIM value increased 50.434% and 18.762% respectively, MSE value decreased 12.891% and 9.558%. In terms of visual effect, the effect of gray spot of corn with different disease degrees after CycleGAN migration was better, so the image of gray spot of corn generated by CycleGAN network was more accurate.

Key words: gray leaf spot, disease migration, residual network, CycleGAN, data augmentation

中图分类号:

S435.13

韩烨, 侯睿峥, 陈霄, . 基于循环一致对抗网络的玉米灰斑病图像迁移方法研究[J]. 中国农机化学报, 2023, 44(2): 163-171.

Han Ye, Hou Ruizheng, Chen Xiao.. Research on images migration method of maize gray disease based on cyclic consistent adversarial network[J]. Journal of Chinese Agricultural Mechanization, 2023, 44(2): 163-171.

参考文献

［1］　
王铎, 温长吉, 王希龙, 等. 基于深度卷积条件生成对抗网络的虫害分类算法研究［J］. 中国农机化学报, 2020, 41(6): 179-187.

Wang Duo, Wen Changji, Wang Xilong, et al. Generating pest identification classification against network based on deep convolution conditions ［J］. Journal of Chinese Agricultural Mechanization, 2020, 41(6): 179-187.

［2］　
王聃, 柴秀娟. 机器学习在植物病害识别研究中的应用［J］. 中国农机化学报, 2019, 40(9): 171-180.

Wang Dan, Chai Xiujuan. Application of machine learning in plant diseases recognition ［J］. Journal of Chinese Agricultural Mechanization, 2019, 40(9): 171-180.

［3］　
张开兴, 吕高龙, 贾浩, 等. 基于图像处理和BP神经网络的玉米叶部病害识别［J］. 中国农机化学报, 2019, 40(8): 122-126.

Zhang Kaixing, Lü Gaolong, Jia Hao, et al. Identification of corn leaf disease based on image processing and BP neural network ［J］. Journal of Chinese Agricultural Mechanization, 2019, 40(8): 122-126.

［4］　
Gatys L A, Ecker A S, Bethge M. Image style transfer using convolutional neural networks ［C］. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2016: 2414-2423.

［5］　
黄琼男, 朱卫纲, 李永刚. 基于GAN的SAR数据扩充研究综述［J］. 兵器装备工程学报, 2021, 42(11): 31-38.

Huang Qiongnan, Zhu Weigang, Li Yonggang. Review on SAR data expansion based on GAN ［J］. Journal of Ordnance Equipment Engineering, 2021, 42(11): 31-38.

［6］　
刘可杰, 董怀玉, 王丽娟. 我国玉米灰斑病菌的种类及其分布调查［J］. 植物保护, 2021, 47(5): 266-270.

Liu Kejie, Dong Huaiyu, Wang Lijuan. Investigation on the species and distribution of pathogens causing gray leaf spot of maize in China ［J］. Plant Protection, 2021, 47(5): 266-270.

［7］　
陈佳, 董学良, 梁金星, 等. 基于注意力机制的CycleGAN服装局部风格迁移研究［J］. 计算机工程, 2021, 47(11): 305-312.

Chen Jia, Dong Xueliang, Liang Jinxing, et al. Research on the local style transfer of clothing by CycleGAN based on attention mechanism ［J］. Computer Engineering, 2021, 47(11): 305-312.

［8］　
Efros A A, Freeman W T. Image quilting for texture synthesis and transfer ［C］. Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques, 2001: 341-346.

［9］　
Hertzmann A, Jacobs C E, Oliver N, et al. Image analogies ［C］. Proceedings of the 28th Annual Conference on Computer Graphics and Interactive Techniques, 2001: 327-340.

［10］　
张海嵩, 尹小勤, 于金辉. 实时绘制3D中国画效果［J］. 计算机辅助设计与图形学学报, 2004, 16(11): 1485-1489.

Zhang Haisong, Yin Xiaoqin, Yu Jinhui. Realtime rendering of 3D chinese painting effects ［J］. Journal of ComputerAided Design & Computer Graphics, 2004, 16(11): 1485-1489.

［11］　
钱小燕, 肖亮, 吴慧中. 快速风格迁移［J］. 计算机工程, 2006, 32(21): 15-17, 46.

Qian Xiaoyan, Xiao Liang, Wu Huizhong. Fast style transfer ［J］. Computer Engineering, 2006, 32(21): 15-17, 46.

［12］　
张惊雷, 厚雅伟. 基于改进循环生成式对抗网络的图像风格迁移［J］. 电子与信息学报, 2020, 42(5): 1216-1222.

Zhang Jinglei, Hou Yawei. Imagetoimage Translation based on improved cycleconsistent generative adversarial network ［J］. Journal of Electronics & Information Technology, 2020, 42(5): 1216-1222.

［13］　
Huang X, Belongie S. Arbitrary style transfer in realtime with adaptive instance normalization ［C］. Proceedings of the IEEE International Conference on Computer Vision, 2017: 1501-1510.

［14］　
LeCun Y, Bengio Y, Hinton G. Deep learning ［J］. Nature, 2015, 521(7553): 436-444.

［15］　
Gatys L A, Ecker A S, Bethge M. A neural algorithm of artistic style ［J］. arXiv preprint arXiv:1508.06576, 2015.

［16］　
Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks ［J］. arXiv preprint arXiv:1511.06434, 2015.

［17］　
Zhu J Y, Park T, Isola P, et al. Unpaired imagetoimage translation using cycleconsistent adversarial networks ［C］. Proceedings of the IEEE International Conference on Computer Vision, 2017: 2223-2232.

［18］　
Kingma D P, Welling M. Autoencoding variational bayes ［J］. arXiv preprint arXiv: 1312.6114, 2013.

［19］　
来杰, 王晓丹, 向前, 等. 自编码器及其应用综述［J］. 通信学报, 2021, 42(9): 218-230.

Lai Jie, Wang Xiaodan, Xiang Qian,et al. Review on autoencoder and its application ［J］. Journal on Communications, 2021, 42(9): 218-230.

［20］　
Goodfellow I, PougetAbadie J, Mirza M, et al. Generative adversarial nets ［J］. Advances in Neural Information Processing Systems, 2014, 27.

［21］　
胡铭菲, 左信, 刘建伟. 深度生成模型综述［J］. 自动化学报, 2022, 48(1): 40-74.

Hu Mingfei, Zuo Xin, Liu Jianwei. Survey on deep generative model ［J］. Acta Automatica Sinica, 2022, 48(1): 40-74.

［22］　
林泓, 任硕, 杨益, 等. 融合自注意力机制和相对鉴别的无监督图像翻译［J］. 自动化学报, 2021, 47(9): 2226-2237.

Lin Hong, Ren Shuo, Yang Yi, et al. Unsupervised imagetoimage translation with selfattention and relativistic discriminator adversarial networks ［J］. Acta Automatica Sinica, 2021, 47(9): 2226-2237.

[1]	于雪莹, 高继勇, 王首程, 李庆盛, 王志强. 基于生成对抗网络和混合注意力机制残差网络的苹果病害识别[J]. 中国农机化学报, 2022, 43(6): 166-174.
[2]	张继成, 李德顺. 基于深度残差学习的成熟草莓识别方法[J]. 中国农机化学报, 2022, 43(2): 136-142.
[3]	陆雅诺;陈炳才;. 基于注意力机制的小样本啤酒花病虫害识别[J]. 中国农机化学报, 2021, 42(3): 189-196.
[4]	张星;高巧明;潘栋;张伟伟;. 基于改进YOLOv3的田间复杂环境下菠萝拾捡识别研究[J]. 中国农机化学报, 2021, 42(1): 201-206.
[5]	闫建伟;赵源;张乐伟;张富贵;. 基于残差网络的自然环境中刺梨果实的识别[J]. 中国农机化学报, 2020, 41(10): 191-196.