摘要
将深度残差神经网络应用到农作物疾病识别中,对14类番茄细粒度病症进行识别,使用数据增强以及改变优化器等方式对深度残差网络进行优化.实验表明:调优后的深度残差网络在粗粒度番茄病症的识别中达到了97.18%的准确率,在细粒度番茄病症的识别中达到了81.68%的准确率,与传统的深度学习模型相比,有更好地识别效果.细粒度病症识别的研究,能够准确定位到疾病的严重程度,对番茄病症的治疗提供了准确的决策支持,有利于农业生产和环境保护.
The deep residual neural network was applied to the recognition of crop diseases and 14 kinds of tomato fine-grained diseases were identified.The deep residual neural network was optimized by using data enhancement and changing optimizer.The experiment shows that the optimized depth residual network achieved an accuracy of 97.18%in the recognition of coarse-grained tomato disease and an accuracy of 81.68%in the recognition of fine-grained tomato disease.Compared with the traditional deep learning model,it has better recognition effect.The study of the fine-grained disease identification can accurately judge the severity of the disease and provide appropriate decision support for the treatment of tomato disease,which is conducive to agricultural production and environmental protection.
作者
胡伟健
樊杰
杜永兴
李宝山
李灵芳
杨颜博
HU Weijian;FAN Jie;DU Yongxing;LI Baoshan;LI Lingfang;YANG Yanbo(Information Engineering School, Inner Mongolia University of Science and Technology, Baotou 014010, China)
出处
《内蒙古科技大学学报》
CAS
2020年第3期251-256,共6页
Journal of Inner Mongolia University of Science and Technology
基金
国家自然科学基金资助项目(61661044,61961033)
内蒙古自治区高等学校青年科技英才计划资助项目(NJYT-19-A15)
内蒙古科技大学创新基金项目-优秀青年科学基金资助项目(2017YQL10)。
关键词
深度学习
番茄病症
细粒度识别
深度残差网络
deep learning
tomato illness
fine-grained identification
deep residual network
