机器学习在无人农场中的应用现状与展望  

Application status and prospect of machine learning in unmanned farm

作　　者：王宝聚 兰玉彬 陈蒙蒙 柳宝虎[2,3] 王国宾 刘海涛[1,2] Wang Baoju;Lan Yubin;Chen Mengmeng;Liu Baohu;Wang Guobin;Liu Haitao(School of Agricultural Engineering and Food Science,Shandong University of Technology,Zibo,255000,China;Shandong University of Technology Sub-center of National Center for International Collaboration Research on Precision Agricultural Aviation Pesticide Spraying Technology,Zibo,255000,China;School of Electrical and Electronics Engineering,Shandong University of Technology,Zibo,255000,China)

机构地区：[1]山东理工大学农业工程与食品科学学院,山东淄博255000 [2]国家精准农业航空施药技术国际联合研究中心山东理工大学分中心,山东淄博255000 [3]山东理工大学电气与电子工程学院,山东淄博255000

出　　处：《中国农机化学报》2021年第10期186-192,217,共8页Journal of Chinese Agricultural Mechanization

基　　金：山东省引进顶尖人才“一事一议”专项经费资助项目(4012118009);中央引导地方科技发展专项资金资助项目(9001118006)。

摘　　要：随着机器学习在生物信息、人脸识别等领域的成功应用,其也为无人农场的发展提供动力。首先阐述无人农场和机器学习的基础概念,同时分别在种植业和畜牧业两个方面对机器学习的应用进行分析,在种植业方面阐述其在田间杂草识别、作物病虫害检测、作物产量预测的应用,在畜牧业方面分析机器学习在鱼类、猪等牲畜的精准识别分类、鱼类的喂食决策系统以及鸡、牛的生产线预测方面的应用现状;提出机器学习存在训练样本难获取、难标记的问题,嵌入式芯片的性能缺陷,以及专业人才缺乏等劣势;应建立通用的无人农场数据库,研究可以预测动物的健康状况以及对动物的生长环境状况进行实时监测的专家系统,还应加强机器学习的嵌入式研究,以及结合5G、大数据、传感器等技术的机器学习将成为无人农场未来的研究方向。本文对机器学习在无人农场的应用现状、问题及展望进行总结叙述,期望为以后的进一步研究提供参考。The successful application of machine learning in biological information and face recognition has provided an impetus for the development of unmanned farms. This article first explained the basic concepts of unmanned farms and machine learning and then analyzed the application of machine learning in plantation and animal husbandry. In terms of the planting industry, this paper explained the application in field weed identification, crop disease, and insect pest detection, and crop yield prediction. In the aspect of animal husbandry, it analyzed the application status of machine learning in the accurate identification and classification of fish, pigs, and other livestock, fish feeding decision-making systems, and chicken and cattle production line prediction. It was pointed out that machine learning has the disadvantages of difficulty obtaining training samples and labeling, performance defects of embedded chips, and lack of professional talents. Future researches should focus on establishing a generic unmanned farm database and studying expert systems that can predict animal health and real-time monitoring of animal growth environment conditions. Future researches should also strengthen the embedded research of machine learning to combine machine learning with 5 G, big data, sensors, and other technologies.

关 键 词：机器学习 无人农场 作物管理 畜牧养殖 精准识别 生产预测 

分 类 号：S237[农业科学—农业机械化工程]