| 何鸿举1*
，王玉玲2
，乔 红3
，欧行奇2
，刘 红4
，王 慧1
，朱亚东1
，蒋圣启1.基于长波近红外光谱快速无接触评估小麦籽粒含
水率[J].海南师范大学学报自科版,2019,32(1):26-32 |
| 基于长波近红外光谱快速无接触评估小麦籽粒含
水率 |
| Rapid and Non-contact Evaluation of Water Content in Wheat byLong-wave Near-infrared Spectra |
| |
| DOI：10.12051/j.issn.1674-4942.2019.01.005 |
| 中文关键词: 光谱 检测 小麦 水分 |
| 英文关键词: spectrum detection wheat moisture content |
| 基金项目:河南省重大科技专项（151100110700）；新乡市重大科技专项（ZD18007）；河南科技学院高层次人才引进项目
（2015015，2015003）；河南科技学院重大科研培育项目（2015ZD02）；河南科技学院标志性创新工程项目（2015BZ03） |
|
| 摘要点击次数: 623 |
| 全文下载次数: 409 |
| 中文摘要: |
| 利用长波近红外光谱（900 ~ 1700 nm）联用偏最小二乘（Partial Least Squares，PLS）算法
快速评估小麦水分含量。通过采集7个不同品种小麦籽粒（百农201、百农207、百农307、百旱207、
AK-58、冠麦1号、周麦18）的近红外反射光谱信息，经高斯滤波平滑（Gaussian Filtering Smoothing，
GFS）、多元散射校正（Multiplicative Scatter Correction，MSC）和标准正态变量变换（Standard Normal
Variable Correction，SNV）三种预处理后，分别利用偏最小二乘法（Partial Least Squares，PLS）挖掘光
谱信息与小麦水分之间的定量关系。结果显示，经GFS预处理的近红外光谱（100个波长）构建的
全波段 PLS 回归模型（F-PLS）的预测相关系数（RP=0.927）、预测误差（RMSEP=1.596%）和鲁棒性
（ΔE = 0.064）均优于另外两种光谱。采用 Regression coefficient 算法筛选最优波长优化 F-PLS 模
型，以提高预测效率。结果显示，从GFS预处理光谱筛选的29个最优波长构建的O-PLS回归模型
预测精度及鲁棒性均较好（RP = 0.909，RMSEP=0.229%，ΔE = 0.078）。本试验表明，利用长波近红
外光谱技术来快速无接触评估小麦籽粒含水率的潜力巨大。 |
| 英文摘要: |
| The moisture content of wheat was rapidly evaluated by long-wave near-infrared spectroscopy (900～1700 nm)
combined with partial least squares (PLS) algorithm. By collecting the near-infrared reflection spectrum information of 7 different varieties of wheat grain (Bainong 201, Bainong 207, Bainong 307, Baihan 207, AK-58, Guanmai 1, Zhoumai 18),
following by the Gaussian filtering smoothing (GFS), multivariate scattering correction (MSC) and standard normal variable
(SNV) transformation pretreatment, the quantitative relationship between spectral information and wheat moisture was built
by PLS respectively. The results showed that the correlation coefficients of prediction in the full-band PLS regression mod⁃
el (F-PLS) constructed by GFS pretreatment (100 wavelengths) performed better (RP=0.927, RMSEP=1.596%, ΔE = 0.064)
than other two pretreated spectra. Regression coefficients algorithm was used to select the optimal wavelength for the optimi⁃
zation of F-PLS model to improve the prediction efficiency. The results showed that the O-PLS regression model built with
29 optimal wavelengths selected from GFS spectra had good prediction accuracy and robustness (RP = 0.909,RMSEP=
0.229%,ΔE = 0.078). The experiment showed that long-wave near-infrared spectroscopy technology has great potential to
evaluate the moisture content of wheat grains in a quick and contactless way. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
