Selection of Agronomic Parameters and Construction of Prediction Models for Oleic Acid Contents in Rapeseed Using Hyperspectral Data
2023
Junwei Lu | Rongcai Tian | Shuangya Wen | Chunyun Guan
High oleic acid oilseed rape is a hot research area in the development of functional oilseed rape. At present, the model of predicting the oleic acid content in rapeseed at the early growth stage based on hyperspectral technology lacks a mechanistic explanation. In this study, based on the data collected at the 5&ndash:6 leaf stage of oilseed rape, a one-dimensional linear regression prediction model of the oleic acid content in leaves (x) and the oleic acid content in rapeseed (y) was constructed with the regression equation y = 1.83x + 75.26, and the R2, RMSE, and RPD of the testing set were 0.96, 0.23%, and 4.86, respectively. Then, a support vector regression prediction model of the spectral standard normal transformed feature parameters and the oleic acid content in leaves was constructed, and the R2, RMSE, and RPD of the testing set were 0.74, 0.21%, and 2.01, respectively. Finally, the sensitive parameter transfer model for the prediction of &ldquo:spectral standard normal transform feature&mdash:oleic acid content in leaves&mdash:oleic acid content in rapeseed&rdquo: was validated, and the R2, RMSE, and RPD of the full sample test were 0.71, 0.54%, and 0.54, respectively. The results show that although the accuracy of the prediction model after the introduction of the agronomic parameters was reduced compared with the performance of direct prediction by using spectra, the oleic acid content in oilseed rape leaves, as an important intermediate variable, could better explain the relationship between the reflection spectrum of the leaf and the oleic acid content in rapeseed. This study provides a theoretical basis and technical support for hyperspectral remote sensing technology in the quality prediction of rapeseed.
Show more [+] Less [-]Bibliographic information
This bibliographic record has been provided by Multidisciplinary Digital Publishing Institute