Rapid Detection of Lambda-cyhalothrin Dosage in Tobacco Field Based on Hyperspectral Imaging Technology
In order to effectively prevent pesticide residue hazards caused by excessive use of pesticides in tobacco field and strengthen pesticide dosage control in tobacco production,the hyperspectral curve characteristics and discrimination model of fresh tobacco leaves treated with different lambda-cyhalothrin doses were studied.The visible near-infrared hyperspectral imaging system(400～900 nm)was used to collect hyperspectral images of fresh tobacco leaf samples treated with different concentrations of lambda-cyhalothrin(CK,1∶2000,1∶1000 and 1∶350)for 24 hours.Four commonly used spectral preprocessing methods were used for preprocessing,followed by continuous projection(SPA)and competitive adaptive reweighted sampling(CARS)to reduce the dimensionality of the data.Then,the least squares support vector machine(LSSVM)and random forest(RF)were used to establish discrimination models.The results show that:(1)After preprocessing with the second derivative(D2),the accuracy of both LSSVM and RF test sets reaches 100%.After preprocessing with the first derivative(D1),the accuracy of the RF test set also reaches 100%.(2)The number of SPA feature selections is less than that of CARS,and the number of wavelengths selected by SPA feature selection after D1 preprocessing is the least.(3)Taking into account the complexity and stability of the model,D1-SPA-RF was ultimately chosen as the optimal discrimination model,with 11 feature wavelengths selected and a test set accuracy of 97.70%.The D1-SPA-RF based on hyperspectral images has superior discriminative performance,and hyperspectral technology can be used for rapid detection of lambda-cyhalothrin doses in tobacco field.
国家科技期刊平台
NSTL
万方数据
