| 基于深度学习的柑橘病虫害动态识别系统研发 |
| |
| 引用本文： | 李昊,刘海隆,刘生龙.基于深度学习的柑橘病虫害动态识别系统研发[J].中国农机化学报,2021,42(9):195. |
| |
| 作者姓名： | 李昊 刘海隆 刘生龙 |
| |
| 作者单位： | 电子科技大学资源与环境学院;四川煤矿安全监察局安全技术中心; |
| |
| 基金项目： | 四川省国际科技合作项目(2020YFH0067) 中科院先导项目(XDA20060303) |
| |
| 摘 要： | 柑橘是我国重要的经济林果之一,因种植区多在山区坡地,病虫害防治给管理带来了很大困难,在线监测与专家决策成为现代农业发展的方向。本文采用物联网技术和深度学习方法,基于尺度可变视频流信息,设计并构建了一套基于柑橘叶片的病虫害动态识别系统。该系统实现了全方位智能控制,解决了实时叶片图像变形和尺度缩放等问题,实现了柑橘图像的动态采集和智能识别。叶片检测的MAP达到87.72%,病害识别准确率达到95.46%,系统运行结果表明,该系统可有效实现柑橘智能监控的管理,为病虫害物联网监控提供参考。
|
| 关 键 词： | 物联网 视频监控 深度学习 叶片检测 病害识别 |
Research on dynamic identification system of citrus diseases and pests based on deep learning ［ |
| |
| Abstract: | Citrus is one of the important economic trees in our country. Because planting areas are mostly on mountainous slopes, the prevention and control of diseases and insect pests have brought great difficulties to management. Online monitoring and expert decision making have become the direction of modern agricultural development. In this paper, using the Internet of Things technology and deep learning methods, a set of dynamic identification systems of diseases and insect pests based on citrus leaves is designed and constructed based on the variable scale video stream information. The system realizes all round intelligent control, solves real time leaf image deformation and scale scaling, and realizes the dynamic collection and intelligent recognition of citrus images. The map of leaf detection is 87.72%, and the accuracy of disease identification is 95.46%. The system operation results show that the system can effectively realize the management of intelligent citrus monitoring and provide a reference for the monitoring of diseases and insect pests. |
| |
| Keywords: | Internet of Things video monitoring deep learning leaf detection disease recognition |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《中国农机化学报》浏览原始摘要信息 |
| 点击此处可从《中国农机化学报》下载免费的PDF全文 |
|
