机器学习在植物表型中的应用进展  

Progresses of Machine Learning Application in Plant Phenotype Research

作　　者：李阿蕾 戴志刚[1] 陈基权[1] 邓灿辉 唐蜻[1] 程超华[1] 许英[1] 张小雨 粟建光[1] 杨泽茂[1] LI Alei;DAI Zhigang;CHEN Jiquan;DENG Canhui;TANG Qing;CHENG Chaohua;XU Ying;ZHANG Xiaoyu;SU Jianguang;YANG Zemao(Institute of Bast Fiber Crops,Chinese Academy of Agricultural Sciences,Changsha 410221,Hunan,China)

机构地区：[1]中国农业科学院麻类研究所,湖南长沙410221

出　　处：《中国麻业科学》2023年第5期248-253,260,共7页Plant Fiber Sciences in China

基　　金：国家自然科学基金青年项目(31601351);湖南省自然科学基金面上项目(2022JJ30650);现代农业产业技术体系建设专项(CARS-16-E01);中国农业科学院创新工程(CAAS-ASTIP-2017-IBFC01)。

摘　　要：表型组学涵盖面广、涉及理化性质多,数据有多态性、时效性、数据量大和高维度、高复杂性、高度不确定性的特征。机器学习作为人工智能的重要组成部分,是近年高通量方面的研究热点。植物表型研究与机器学习有效结合不仅可以扩大表型数据量级,还可以挖掘从分子、组织、个体到群体的所有层次植物表型,促进植物研究快速发展。研究以机器学习分类和机器学习在植物表型研究中的应用为出发点,介绍机器学习基本流程,比较众多机器学习算法应用的优缺点,概述国内外最新机器学习在表型组学中的研究热点,探讨植物表型与机器学习未来发展趋势。Phenomics cover a wide range of physical and chemical properties,the data with charac⁃teristics of variety,velocity,volume and high dimension,high complexity,high uncertainty.As an im⁃portant part of artificial intelligence,machine learning is a hot topic in high throughput research in recent years.The effective combination of plant phenotype research and machine learning can not only expand the magnitude of phenotypic data,but also mine plant phenotypic data at all levels,from molecules,tis⁃sues,individuals to groups,promoting the rapid development of plant technology.This article starts from the classification of machine learning and its application in plant phenotype research,describes the basic process of machine learning,compares the advantages and disadvantages of various machine learning al⁃gorithms,summarizes the latest hot research in machine learning and phenotypes at home and abroad,and discusses the future development trend of plant phenotyping and machine learning.

关 键 词：机器学习 表型鉴定 高通量 智慧农业 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]