期刊文献+

基于Micro-CT分析大豆种子结构表型及构建种子重量预测模型

Analysis of Soybean Seed Structure Phenotype and Construction of Seed Weight Prediction Model Based on Micro-CT
在线阅读 下载PDF
导出
摘要 大豆是重要的粮油兼用作物,是植物蛋白质的重要来源,为明确大豆种子结构特征并构建种子重量预测模型,以42个不同大豆品种为材料,利用Micro-CT技术扫描测试样本,通过CT图像的处理解析,获取大豆种子的长度、宽度、厚度、体积、表面积、种胚体积、表面积及空腔体积特征,人工称重测定单粒重量指标。系统分析种子形态特征及其与重量的相关关系,并对不同品种进行聚类分析,通过对多项形态表型和重量指标进行主成分分析,确定主要贡献指标,基于机器学习算法构建重量预测模型。结果表明:种子形态特征与种子重量显著相关,但种子形状特征对重量无显著影响;42个大豆品种可以分为4类,其中第一类品种蒙豆375和蒙豆60的种子大小和重量指标均显著高于其他3类品种。利用随机森林模型和偏最小二乘回归方法构建的重量预测模型效果优于单一指标的简单线性回归效果。其中,随机森林回归模型训练集和测试集的R^(2)分别为0.80和0.66,RMSE分别为0.017和0.021 g,偏最小二乘回归模型训练集和测试集的R^(2)分别为0.75和0.72,RMSE分别为0.019和0.020 g。研究结果为大豆种子外部形态和内部结构的研究提供了新的技术和方法,为大豆品种分类、产量和品质性状的评价提供理论和技术参考。 Soybean is an important dual-purpose crop used for both food and oil production.It is a significant source of plant protein.In order to clarify the structural characteristics of soybean seeds and construct a seed weight prediction model,42 different soybean varieties were used as materials and scanned using Micro-CT technology to obtain the seed length,seed width,seed height,seed volume,seed surface area,embryo volume,embryo surface area,and cavity volume features.The single seed weight was measured manually.The morphological characteristics of the seeds and their correlation with weight were systematically analyzed.Cluster analysis was conducted on different varieties,and principal component analysis was applied to multiple morphological and weight indicators to determine the main contributing factors.A weight prediction model was then built based on machine learning algorithms.The results showed that seed morphological characteristics were significantly correlated with seed weight,but seed shape had no significant impact on weight.The 42 soybean varieties could be divided into four categories,with the first category consisting of the varieties Mengdou 375 and Mengdou 60,which had significantly larger seed size and higher weight indicators compared to the other three categories.The weight prediction model constructed using random forest regression and partial least squares regression methods outperformed simple linear regression using a single indicator.The random forest regression model achieved R^(2) values of 0.80 and 0.66 for the training and test sets,respectively,with RMSE values of 0.017 and 0.021 g.The partial least squares regression model achieved R^(2) values of 0.75 and 0.72 for the training and test sets,respectively,with RMSE values of 0.019 and 0.020 g.The research findings provide new techniques and methods for the study of the external morphology and internal structure of soybean seeds.They also offer theoretical and technical references for soybean variety classification,yield,and quality evaluation.
作者 刘长斌 李远鲲 郭民坤 樊江川 郭新宇 卢宪菊 LIU Changbin;LI Yuankun;GUO Minkun;FAN Jiangchuan;GUO Xinyu;LU Xianju(NongXin Science&Technology(Beijing)Co.,Ltd.,Beijing 100097,China;Information Technology Research Center,Beijing Academy of Agriculture and Forestry Science/Beijing Key Lab of Digital Plant,Beijing 100097,China)
出处 《大豆科学》 北大核心 2025年第1期11-21,共11页 Soybean Science
基金 国家重点研发计划(2023YFD2301803) 北京市农林科学院作物表型组学协同创新中心(KJCX20240406)。
关键词 MICRO-CT 大豆 结构表型 重量预测 模型构建 机器学习 Micro-CT soybean structure phenotype seed weight prediction modeling machine learning
  • 相关文献

参考文献10

二级参考文献187

共引文献790

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部