Hyperspectral Properties of Rice Varieties with Varying Resistance Under Brown Planthopper(Nilaparvata lugens)Infestation
[Objective]The objective of this study is to investigate changes in hyperspectral reflectance curves and sensitive spectral features in rice varieties displaying varying resistance levels to brown planthopper infestation.Additionally,the study aims to examine changes in hyperspectral values in different parts of the rice plant.The obtained results are utilized to develop a machine learning model for identifying brown planthopper resistance,providing essential fundamental data for the development of intelligent technologies in identifying brown planthopper resistance.[Method]Three rice varieties(TN1,Mudgo,and RHT),each with varying resistance levels to brown planthoppers,were selected.The differences in hyperspectral values and vegetation indices were analyzed,and a random forest model was established to predict their resistance level.[Results]The study revealed that the number of significant spectral bands and the number of significant differences in spectral bands,significantly correlated with the duration of brown planthopper infestations,decrease with increasing resistance levels of the rice plant.At around 680 nm,the correlation with the duration of brown planthopper infestations was strongest for all three varieties.The analysis of vegetation indices showed that SIPI,SR605/760,and PSNDb had higher absolute values of correlation coefficients with resistance levels than those beyond 680 nm.Differences in different plant parts appeared first in TN1,a variety sensitive to brown planthoppers,followed by Mudgo,a moderately resistant variety,and finally RHT,a highly resistant one.The differences first appeared in the sheath of the first leaf,followed by the sheath of the second leaf,and the sheath of the third leaf.The results of the prediction model showed that the model with all spectral bands as input performed better than the random forest model built with a single vegetation index,SIPI,and achieved an accuracy of 85.9%.[Conclusion]The study highlights spectral changes associated with brown planthopper resistance among rice varieties and different plant parts.It confirms the suitability of machine learning technology for predicting the resistance level of rice.
brown planthopperrice resistancehyperspectravegetation indexmachine learning
国家科技期刊平台
NSTL
万方数据
