草地学报

草地学报 ›› 2026, Vol. 34 ›› Issue (9): 3185-3196.DOI: 10.11733/j.issn.1007-0435.2026.09.005

• 研究论文 • 上一篇    

水肥管理对紫花苜蓿种子生产的影响及多光谱机器学习模型构建

鞠克升1, 周金露2, 王青2, 贺晓帆2, 陈盈燕2, 聂嘉欣2, 刘芳茜2, 李昱宏2, 赵琨2, 贾善刚2   

  1. 1. 酒泉市农产品质量安全监督管理站酒泉市种业发展中心, 甘肃 酒泉 735000;
    2. 中国农业大学草业科学与技术学院, 北京 100193
  • 收稿日期:2026-05-06 修回日期:2026-07-01 发布日期:2026-09-11
  • 通讯作者: 贾善刚, E-mail: shangang.jia@cau.edu.cn
  • 作者简介:鞠克升(1971-),男,汉族,天津人,高级畜牧师,主要从事农产品质量安全和牧草种子推广示范研究,E-mail:349176563@qq.com
  • 基金资助:
    “国家重点研发计划”（2022YFD1300804）； 山东省自然科学基金（ZR2024MC066）资助

Effects of Water and Fertilizer Management on Seed Production of Alfalfa （Medicago sativa L.） and Construction of Multispectral Machine Learning Models

JU Ke-sheng1, ZHOU Jin-lu2, WANG Qing2, HE Xiao-fan2, CHEN Ying-yan2, NIE Jia-xin2, LIU Fang-qian2, LI Yu-hong2, ZHAO Kun2, JIA Shan-gang2   

  1. 1. Jiuquan City Agricultural Products Quality Safety Supervision and Management Station, Jiuquan City Seed Industry Development Center,Jiuquan, Gansu Province 735000, China;
    2. College of Grassland Science and Technology, China Agricultural University, Beijing 100193, China
  • Received:2026-05-06 Revised:2026-07-01 Published:2026-09-11

PDF

摘要： 豆科牧草紫花苜蓿（Medicago sativa L.）种子产量与品质受水肥管理影响显著。本研究旨在探讨水肥梯度处理对紫花苜蓿种子产量及内容物含量的影响，试验设计了不同灌溉（600 m3·hm-2， 900 m3·hm-2， 1200 m3·hm-2）和磷肥（100 kg·hm-2， 200 kg·hm-2， 300 kg·hm-2）组合处理方案，测定了种子产量、产量组分、可溶性蛋白质和淀粉含量等指标，获取种子19个波段的多光谱信息，构建了nCDA-CNN模型，评估多光谱图像技术在水肥处理种子检测中的应用潜力。结果表明，适量水肥施用显著提高了种子产量、蛋白和淀粉含量（P<0.05）。其中灌水量600 m3·hm-2、施磷量200 kg·hm-2处理的种子产量最高（885.44 kg·hm-2），较对照增产287.96 kg·hm-2；灌水量600 m3·hm-2、施磷量100 kg·hm-2处理的种子可溶性蛋白和淀粉含量较高。通径分析显示，每荚果种子数对种子产量影响较大。种子多光谱图像分析发现，nCDA-CNN模型能够准确预测水肥处理的紫花苜蓿种子，识别准确率达到95%，这为紫花苜蓿种子生产高通量无损检测提供了可能。优化水肥管理可显著提升紫花苜蓿种子生产效益，多光谱图像技术结合深度学习模型在草种业质量检测与可持续发展中具有重要应用前景。

关键词: 水肥管理, 紫花苜蓿, 种子产量, 多光谱检测

Abstract: The seed yield and quality of the leguminous forage alfalfa （Medicago sativa L.） are significantly influenced by water and fertilizer management. This study aimed to investigate the effects of different water and phosphorus fertilizer treatments on seed yield and seed quality of alfalfa. The experiment was designed to combine treatments with different irrigation levels （600， 900， and 1200 m3·hm-2） and phosphorus fertilizer rates （100， 200， and 300 kg·hm-2）. Seed yield， yield components， soluble protein content， and starch content were measured. Additionally， multispectral information of 19 bands was acquired， and an nCDA-CNN model was constructed to evaluate the potential of multispectral image technology for seed detection under different water and fertilizer treatments. The results showed that the appropriate water and phosphorus fertilizer application significantly increased seed yield， protein content， and starch content （P<0.05）. Among all treatments， the combination of 600 m3·hm-2 irrigation and 200 kg·hm-2 phosphorus fertilizer achieved the highest seed yield （885.44 kg·hm-2）， representing an increase of 287.96 kg·hm-2 compared to the control. Higher soluble protein and starch contents of seeds were observed under the treatment with 600 m3·hm-2 irrigation and 100 kg·hm-2 phosphorus fertilizer. Path analysis indicated that the number of seeds per pod had a greater influence on seed yield. Multispectral image analysis revealed that the nCDA-CNN model accurately predicted alfalfa seeds subjected to different water and fertilizer treatments， with a recognition accuracy of 95%. This provides a high-throughput， non-destructive detection method for alfalfa seed production. Optimized water and fertilizer manag-ement can significantly enhance alfalfa seed production. Multispectral image technology combined with deep learning models holds important application prospects for quality detection and sustainable development in the grass seed industry.

Key words: Water and fertilizer Management, Alfalfa, Seed yield, Multispectral detection

中图分类号: