偏最小二乘回归
近红外光谱
均方误差
吸光度
相关系数
平滑的
光谱学
分析化学(期刊)
内容(测量理论)
小波
数学
透射率
遥感
叶绿素
材料科学
决定系数
生物系统
化学
光学
计算机科学
统计
人工智能
物理
色谱法
光电子学
数学分析
有机化学
量子力学
生物
地质学
作者
Qingbo Li,Yan-Wen Huang,Guangjun Zhang,Qianxuan Zhang,Xiang Li,Jinguang Wu
摘要
In the present paper a method based on Vis/NIR spectral analysis technology was applied to the nondestructive measurement of plant chlorophyll content. Firstly, the Vis/NIR spectra in the wavelength range from 500 to 900 nm of the plant leaves were acquired from 35 samples by transmittance and reflectance method, and then three different mathematical treatments were used in original spectra data pretreatment to decrease the noise: moving average smoothing with the segment size 5, first Savitzky-Golay derivative, and wavelet transform (WT) way. Secondly, a total of 35 samples were examined in the test, in which 23 samples were selected randomly for model building and the other 12 for model prediction, then partial least squares (PLS) method was used to develop the quantitative analysis model for chlorophyll content with absorbance spectroscopy, and 7 principal components (PCs) were selected. Finally, this model was used to predict the chlorophyll content of 12 unknown leaf samples in prediction collection. The experiment result indicated that the better prediction performance was achieved with the correlation coefficient between the prediction values and the truth values being 0.93, and the root mean squared error of prediction is about 1.1 SPAD. It could be concluded that it is feasible to measure plant chlorophyll content based on visible/near infrared (Vis/NIR) spectroscopy. And it is also significant in realizing rapid and nondestructive measurement of chlorophyll content in the future.
科研通智能强力驱动
Strongly Powered by AbleSci AI