基于深度学习的园林智能浇灌系统  

Intelligent Garden Irrigation System Based on Deep Learning

作　　者：刘鹏[1] 曹晓辉 胡文鹏 殷伟铭 左辰 罗亚波[1] Liu Peng;Cao Xiaohui;Hu Wenpeng;Yin Weiming;Zuo Chen;Luo Yabo(School of Mechanical and Electronic Engineering,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学机电工程学院,湖北武汉430070

出　　处：《湖北汽车工业学院学报》2018年第2期61-64,71,共5页Journal of Hubei University Of Automotive Technology

基　　金：武汉理工大学节能减排社会实践与科技竞赛项目(WHUT-2018-11)

摘　　要：园林浇灌有自动浇灌和人工浇灌2种模式,但都存在水资源浪费、人力资源浪费和不合理浇灌等问题。针对这些问题,文中将深度学习和机器视觉技术应用于土壤湿度分类,以土壤图像与对应的湿度为样本,建立了卷积神经网络框架,并基于大量实验对卷积神经网络进行了训练和验证,实现了基于土壤图像信息的浇灌需求智能决策,决策正确率达到85%以上,并与控制系统相结合,实现了园林智能浇灌系统,从而达到合理浇灌、节约水资源与人力资源的目标。The garden watering including both automatic irrigation and artificial irrigation, has problems such as waste of water resources, human resources and unreasonable irrigation. Deep learning and machine vision were applied to the classification of soil moisture, soil images and corresponding humiditywere taken as samples, and a frame of convolution neural network was established. According to a largenumber of experiments, the convolution neural network was trained and verified to realize the intelligentdecision of irrigation demand based on soil image information. The correct decision rate is above 85%,and the intelligent irrigation system is realized by combining with the control system, so as to achieve the goal of rational irrigation and saving water resources and human resources.

关 键 词：深度学习 卷积神经网络 机器视觉 智能浇灌 

分 类 号：TP271[自动化与计算机技术—检测技术与自动化装置]