草地学报 ›› 2024, Vol. 32 ›› Issue (2): 527-534.DOI: 10.11733/j.issn.1007-0435.2024.02.021

• 研究论文 • 上一篇    下一篇

基于人工神经网络的狼尾草属牧草品质近红外光谱预测研究

朱瑞芬1,2, 徐远东1,2, 孙万斌1,2, 刘畅1,2, 姚博1,2, 陈积山1,2   

  1. 1. 重庆市畜牧科学院, 重庆 402460;
    2. 重庆市草业工程技术研究中心, 重庆 402460
  • 收稿日期:2023-07-06 修回日期:2023-10-25 出版日期:2024-02-15 发布日期:2024-03-06
  • 通讯作者: 陈积山,E-mail:cjshlj@163.com
  • 作者简介:朱瑞芬(1982-),女,汉族,甘肃礼县人,副研究员,主要从事牧草与草地管理研究,E-mail:75529693@qq.com
  • 基金资助:
    南方狼尾草品质的近红外光谱模型研究与应用评价(21521);重庆市现代农业产业体系(草食牲畜:CQMAITS202313)资助

Research on Nutritional Components of Pennisetum Rich. Forage by Near Infrared Spectroscopy Model based on Artificial Neural Network

ZHU Rui-fen1,2, XU Yuan-dong1,2, SUN Wan-bin1,2, LIU Chang1,2, YAO Bo1,2, CHEN Ji-shan1,2   

  1. 1. Institute of Pratacultural Science, Chongqing Academy of Animal Sciences, Chongqing 402460, China;
    2. Pratacultural Engineering and technology research center of Chongqing, Chongqing 402460, China
  • Received:2023-07-06 Revised:2023-10-25 Online:2024-02-15 Published:2024-03-06

摘要: 本研究利用近红外光谱通过人工神经网络(Artificial neural network,ANN)建立狼尾草属(Pennisetum Rich)牧草水分、粗蛋白、木质素、酸性/中性洗涤纤维及灰分含量的预测模型。结果表明:基于人工神经网络的狼尾草属牧草品质预测模型总体优于全光谱偏最小二乘法(PLS)模型效果。在人工神经网络的方向传播(BP)网络模型中,6项表征牧草品质指标的校正均方根误差(RMSEC)、预测均方根误差(RMSEP)均显著低于PLS模型,同时校正集决定系数(R2C)、预测集决定系数(R2P)显著提高,除灰分含量预测不理想外,其他预测效果均理想。同时人工神经网络的BP网络对于近红外光谱的非线性数据具有良好的拟合能力,其预测模型对于指导狼尾草属牧草品质预测和分级管理研究具有广阔的应用前景。

关键词: 狼尾草属, 常规营养成分, 近红外光谱, 人工神经网络

Abstract: In order to determine the nutrient content of Pennisetum Rich. forage rapidly by Near Infrared Spectroscopy,the prediction models of moisture content,crude protein content,lignin content,acid detergent fibe,neutral detergent fibe and crude ash content of Pennisetum Rich. forage were established by using artificial neural network (ANN). The results showed that the quality prediction model of Pennisetum Rich forage based on ANN was better than the partial least squares (PLS) model. In the back propagation (BP) network model of ANN,the calibration root mean square error and prediction root mean square error are significantly reduced,and the calibration set determination coefficient and the prediction set determination coefficient were significantly improved. Except the ash content prediction,the other prediction results were ideal. Meanwhile,the BP of ANN algorithm had good fitting ability for the nonlinear data of near-infrared spectrum. The prediction model in this study has application prospects for guiding forage quality classification management and scientific research.

Key words: Pennisetum Rich., Quality components, Near infrared spectroscopy, Artificial neural network (ANN)

中图分类号: