草地学报

草地学报 ›› 2024, Vol. 32 ›› Issue (12): 3688-3697.DOI: 10.11733/j.issn.1007-0435.2024.12.004

• 研究论文 • 上一篇    

不同燕麦品种种子自然老化表型和多光谱变化特征

刘俊泽1, 贺英彩3, 马秀琴2, 倪浩然1, 王雪萌1, 王青1, 贺晓帆1, 聂嘉欣1, 胡昊1, 毛培胜1, 贾善刚1   

  1. 1. 中国农业大学草业科学与技术学院, 北京 100193;
    2. 青海省祁连县畜牧兽医站, 青海 海北藏族自治州 810499;
    3. 青海省祁连县草原站, 青海 海北藏族自治州 810499
  • 收稿日期:2024-06-20 修回日期:2024-09-19 发布日期:2024-12-14
  • 通讯作者: 贾善刚,E-mail:shangang.jia@cau.edu.cn
  • 作者简介:刘俊泽(1999-),男,汉族,辽宁葫芦岛人,硕士研究生,主要从事育种与种子科学研究,E-mail:junze.liu@cau.edu.cn;
  • 基金资助:
    “现代农业产业技术体系”(CARS-34)；“国家重点研发计划”(2022YFD1300804)；“四川省省院省校合作重点研发项目”(2023YFSY0012)资助

Characteristics of Natural Ageing Phenotypes and Multispectral Changes in Seeds of Different Oat Varieties

LIU Jun-ze1, HE Ying-cai3, MA Xiu-qin2, NI Hao-ran1, WANG Xue-meng1, WANG Qing1, HE Xiao-fan1, NIE Jia-xin1, HU Hao1, MAO Pei-sheng1, JIA Shan-gang1   

  1. 1. College of Grassland Science and Technology, China Agricultural University, Beijing 100193, China;
    2. Qilian County Animal Husbandry and Veterinary Station, Haibei Tibetan Autonomous Prefecture, Qinghai Province 810499, China;
    3. Qilian County Grassland Station, Haibei Tibetan Autonomous Prefecture, Qinghai Province 810499, China
  • Received:2024-06-20 Revised:2024-09-19 Published:2024-12-14

PDF

摘要： 本研究利用多光谱成像技术对12个燕麦(Avena sativa)品种三年自然老化前后种子进行检测，并对其形态和光谱两种特征参数进行分析。结果显示，不同品种及其自然老化种子形态和光谱特征参数存在差异。基于形态和光谱特征参数的k-means聚类分析将不同燕麦品种种子大致归为两大类群：自然老化前和自然老化后。此外，运用课题组此前开发的nCDA-CNN模型对种子老化和发芽率预测，发现多光谱图像的预测准确度达到了100%。进一步分析发现，自然老化前后种子发芽率与颜色参数L、630 nm和690 nm波长光谱反射率具有显著正相关性(P<0.05)；种子老化与四个形态参数(表面积、长度、形态参数B和饱和度)以及365 nm波长光谱反射率具有显著正相关性(P<0.05)。以上研究结果表明，不同品种的燕麦种子在自然老化前后的外部形态及光谱特征上存在显著差异(P<0.05)，运用基于图像的机器学习模型能够准确鉴别种子老化和预测发芽率，对进一步研究老化种子的生理生化特征有一定的意义。

关键词: 多光谱成像, 燕麦种子, 无损检测, 自然老化种子, 机器学习

Abstract: In this study, seeds of 12 oat (Avena sativa) varieties before and after three years of natural aging were examined using multispectral imaging and analyzed for both morphological and spectral characterization parameters. The results showed that different varieties and their naturally aged seeds showed differences in both morphological and spectral characterization parameters. Based on the k-means clustering analysis of morphological and spectral parameters, the different oat varieties were roughly categorized into two major groups：one of which was the pre-naturally aged seeds and the other was the naturally aged seeds. In addition, the previously developed nCDA-CNN model was utilized for seed aging and germination probability prediction based on the training and validation sets, and it was found that the prediction accuracies of the multispectral images all reached 100%. Further analysis showed that the seed germination rate before and after natural aging was significantly correlated with different morphological parameters and spectral reflectance at 630 nm and 690 nm wavelengths;seed aging was significantly correlated with the four morphological parameters (surface area, length, morphological parameter B and saturation) as well as with spectral reflectance at 365 nm. The above findings indicated that different oat varieties seeds and their aging before and after natural aging had external morphological and spectral variability, and that seed aging and germination probability could be accurately identified and predicted by using an image-based machine learning model.

Key words: Multispectral imaging, Oat seeds, Non-destructive testing, Natural aging seeds, Machine learning

中图分类号: