首页 | 本学科首页   官方微博 | 高级检索  
     检索      

牛结节性皮肤病病毒基因组学研究进展
引用本文:翟 颀,黄敏霞,吕殿红,贾春玲,周秀蓉,温肖会,罗胜军.牛结节性皮肤病病毒基因组学研究进展[J].广东农业科学,2022,49(6):97-104.
作者姓名:翟 颀  黄敏霞  吕殿红  贾春玲  周秀蓉  温肖会  罗胜军
基金项目:深圳市科技计划项目(CJGJZD20210408092401004)
摘    要:【目的】通过调节训练集内实验室场景图片与田间场景图片的分布,提高深度学习模型的准确度, 以减少植物病害识别深度学习模型对田间场景数据的依赖。【方法】通过调节训练集内实验室场景图片和田间 场景图片的分布,使用 ResNeSt-50、VGG-16、ResNet-50 等 3 种神经网络结构分别对训练得到的深度学习模型 进行测试和比较,从而优化植物病害识别模型。【结果】在由一定数量的植物病害图像组成的训练集内,调节 其中不同场景图片的分布会对模型的准确率产生影响。当训练集内的田间场景图片分布达 30% 时,模型准确率 提升 18% 以上。在 100% 实验室场景图片的训练集内添加 30% 田间场景图片,可提升模型准确率 17% 以上;在 100% 田间场景图片的训练集内添加实验室场景图片,模型准确率随图片数量增加而提升,提升幅度为 2%~4%。 【结论】该方法适用于农业复杂环境下高准确度病害识别模型的快速建立,可减少深度学习模型对田间场景数 据的依赖,缩短模型建立初期的田间数据采集周期,降低田间数据采集成本,促进人工智能技术在无人农场及 智慧农业中更有效地运用。

关 键 词:植物病害  深度学习  卷积神经网络  数据分布  实验室场景  田间场景

Research Progress in Genomics of Lumpy Skin Disease Virus
ZHAI Qi,HUANG Minxi,LYU Dianhong,JIA Chunling,ZHOU Xiurong,WEN Xiaohui,LUO Shengjun.Research Progress in Genomics of Lumpy Skin Disease Virus[J].Guangdong Agricultural Sciences,2022,49(6):97-104.
Authors:ZHAI Qi  HUANG Minxi  LYU Dianhong  JIA Chunling  ZHOU Xiurong  WEN Xiaohui  LUO Shengjun
Abstract:Lumpy skin disease (LSD) is an animal infectious disease caused by Lumpy Skin Disease Virus (LSDV) that produces nodular lesions in cattle, and is widespread in East, South and Southeast Asian countries, causing serious economic losses to the local cattle industry. Currently, LSD is managed as a second-class animal disease in China. Since 2019, more than ten provinces in China have reported LSD epidemics, whose situation is very serious. At present, the superior cell for LSDV isolation and proliferation is Madin-Darby bovine kidney (MDBK). As a member of the Goat Pox Virus genus of the Poxviridae family, LSDV has a virus particle diameter about 290 nm × 270 nm, with two existence forms. The viral genome is complex and huge, about 151 kb, and contains more than 150 genes. Currently, only approximately 40 LSDV strains containing the whole genome sequence information are stored in the GenBank database. In addition to field wild strains, recombinant strains of field wild strains and vaccine strains have also emerged in some parts of Asia in recent years, and the prevalent recombinant strains are pathogenic, which poses new challenges for development of vaccine and preparation of prevention and control strategies for LSD in the future. The study mainly summarizes the etiology of LSD and the progress in the genome-wide study of LSDV strains in different regions of Africa, Europe and Asia, with a view to providing references for the prevention and control of LSD and further researches.
Keywords:plant disease  deep learning  convolutional neural network  data distribution  lab-condition  field-condition
点击此处可从《广东农业科学》浏览原始摘要信息
点击此处可从《广东农业科学》下载免费的PDF全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号