Acta Agrestia Sinica ›› 2023, Vol. 31 ›› Issue (10): 2985-2991.DOI: 10.11733/j.issn.1007-0435.2023.10.009

Previous Articles     Next Articles

Seed Identification of Alfalfa Varieties Based on Multispectral Imaging

NI Hao-ran, WANG Run, HU Hao, LIU Jun-ze, TIAN Pei-xin, JIA Shan-gang   

  1. College of Grassland Science and Technology, China Agricultural University, Beijing 100193
  • Received:2023-05-24 Revised:2023-07-05 Online:2023-10-15 Published:2023-11-02

基于多光谱成像技术鉴别不同紫花苜蓿品种的种子

倪浩然, 王润, 胡昊, 刘俊泽, 田沛鑫, 贾善刚   

  1. 中国农业大学草业科学与技术学院, 北京 100193
  • 通讯作者: 贾善刚,E-mail:shangang.jia@cau.edu.cn
  • 作者简介:倪浩然(2000-),男,汉族,四川自贡人,硕士研究生,主要从事育种与种子科学研究,E-mail:haoran.ni@cau.edu.cn
  • 基金资助:
    “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);“现代农业产业技术体系”(CARS-34)资助

Abstract: Non-destructive identification of seeds for different alfalfa (Medicago sativa) varieties is a highly promising field of research. In this study,multispectral imaging (MSI) was used to screen seeds of 25 varieties of alfalfa,based on morphological and spectral features. The results of both morphological and spectral data showed significant differences among the varieties. Using k-means clustering of morphological data,alfalfa varieties were classified into four categories,among which,three ones enriched in foreign varieties and the other one enriched in domestic varieties. By combining k-means clustering of morphological and spectral data,it was sufficient to generally differentiate the seeds of 25 varieties. Additionally,MSI output was used to correlate the substance content inside the seed,and correlate the ascorbic acid (AsA) content as an example. The results showed a correlation between AsA content inside the seed and the morphological parameters such as “BetaShape_a”“BetaShape_b”“RatioWidthLength” and “Compactness Circle”,and their correlation coefficients were -0.59,-0.58,-0.67 and -0.68,respectively. These findings demonstrate the potential of MSI for identifying alfalfa varieties,and suggest a certain correlation between the stored substances inside the seed and seed images.

Key words: Multispectral imaging, Alfalfa seeds, Varieties identification, Non-destructive identification

摘要: 不同紫花苜蓿(Medicago sativa)品种种子的无损检测是一个极具潜力的研究方向,本研究利用多光谱成像(Multispectral imaging,MSI)技术对25个紫花苜蓿品种的种子进行检测,并对形态和光谱两种特征参数进行分析。结果显示,不同品种种子的形态和光谱特征参数均有区别。基于形态特征参数的k-means聚类分析,将不同苜蓿品种大致归为四类,其中三类偏向于国外品种,一类偏向于国内品种。再结合基于光谱特征参数的k-means聚类分析,根据聚类结果,可以大致区分25个品种的种子。此外,本研究尝试用多光谱数据来关联种子内物质含量,对种子抗坏血酸(Ascorbic acid,AsA)含量与形态、光谱特征参数的相关性进行分析,结果表明,种子AsA含量与形状参数a、形状参数b、长宽比以及紧实度圆等形态参数具有一定的相关性,相关系数分别为-0.59,-0.58,-0.67,-0.68。以上研究结果表明,利用多光谱成像技术检测不同紫花苜蓿品种具有可行性,种子内储存物质含量也与种子图像存在一定相关性。

关键词: 多光谱成像, 紫花苜蓿种子, 品种鉴别, 无损检测

CLC Number: