Detection of Bemisia tabaci Infestation Degrees on Tomato Leaves Based on Visible/Near-Infrared Hyperspectral Imaging
Fan Yangyang
Song Hualu
Liang Zhichao
Shen Tingting
Wang Shuai
Abstract：Bemisia tabaci(Aleyrodidae)is a widely distributed and highly destructive pest.Early detec-tion is of crucial significance to its effective control.In this study,the visible/near-infrared hyperspectral ima-ging technology was used to acquire the hyperspectral images of tomato leaves at five early infestation stages of B.tabaci.The feature wavelengths were extracted using four methods:successive projections algorithm(SPA),variable combination population analysis(VCPA),Monte Carlo uninformative variable elimination(MCFS),and principal component analysis loading(PCA-loading).These feature wavelengths were then used as input variables to construct detection models employing linear discriminant analysis(LDA),support vector machine(SVM),and backpropagation neural network(BPNN).The results indicated that the SVM model built with feature wavelengths extracted by SPA achieved the best detection performance,with an accu-racy of 86.21％.Subsequently,texture features were extracted from the images corresponding to the SPA-se-lected wavelengths using the gray-level co-occurrence matrix(GLCM).Ten texture features were selected through analysis of variance and used as input variables independently or in combination with the feature wave-lengths for model construction.The results showed that the models using only texture features as input variables had significantly lower detection accuracy compared to those using only feature wavelengths.Furthermore,combining feature wavelengths with texture features as input variables did not lead to significant improvement in model detection accuracy.In conclusion,detecting early B.tabaci infestation on tomato leaves is achievable by acquiring images via visible/near-infrared hyperspectral imaging,extracting feature wavelengths using the SPA method,and inputting them into an SVM model for detection.This approach offers technical supports for early detection and management of B.tabaci.
Keywords：Bemisia tabaciTomato leavesHyperspectral imaging technologyFeature wavelengthTexture featureMachine learning
Publication Date：2026-03-30
Online Publishing Date：2026-05-22(First online date of this platform, not the publication date of the document)
Pages：9( 151-159 )