药桑叶片分类及叶绿素含量预测

打开文本图片集

中图分类号：S88；TP181 文献标识码：A文章编号：1673-9868（2026）02-0001-16

Classification of Medicinal Mulberry Leaf and Prediction of Chlorophyll Content

LUO Yiwei， LIU Hongjiang， LI Wei，LI Li， YUAN Jianglian， HE NingjiaState Key Laboratory of Resource Insects，Southwest University，Chongqing 400716，China

Abstract：To advance the construction of a factory-based precision management system for medicinal mulberry，this study first established hyperspectral reflectance-based classification models for leaves under distinct physiological conditions. Subsequently，chlorophyll content prediction models were developed. During the model construction， 3 sample set division methods，5 data preprocessing methods，and 5 machine learning algorithms were systematically compared. Results of medical mulberry leaf classification indicated that the support vector machine （SVM） models using random database（RD） with either raw data or multiplicative scatter correction （MSC） for sample partitioning-specifically，the RD-RAW-SVM and RD-MSC-SVM-achieved the best R2 values of O. 954 on the test set. Confusion matrix analysis revealed that the clasification accuracies for healthy，infested，and shaded leaves were identical across the two different models，yielding 96% ， 93% ，and 98% ，respectively. Nutrient-deficient leaves correctly classified with accuracies of 96% and 97% in the two models. For chlorophyll content prediction， the partial least squares （PLS） model using RD combined with baseline correction （BC）-specifically，RD-BC-PLS-demonstrated optimal performance，with a testing set R2 of 0.895 and root mean square error （RMSE） of 3.461. Additionally，weight analysis identified two critical wavelengths （ 505nm and 734nm ） contributing maximally to chlorophyll content prediction，enabling the derivation of novel spectral indices. Comparative evaluation with 19 conventional vegetation indices revealed that the red-edge normalized diference vegetation index （NDVI705）and the newly developed MFD734-505 exhibited equivalent predictive eficacy for chlorophyll estimation，with a testing set R2 of O.864. This integrated modeling framework provides an important foundation for the precise monitoring and intelligent management of factory-based medicinal mulberry cultivation.

Key words： medicinal mulberry； hyperspectral reflectance; classification model; chlorophyll content; regression model

桑树作为古代丝绸之路的重要基石，数千年来在蚕桑业和传统医药中具有不可替代的地位，《本草纲目》对此早有记载。（剩余20334字）