Establishment of hyperspectral prediction model for cadmium content in flue-cured tobacco leaves

• Main Authors: CHEN Nan, FENG Hui-lin, YANG Yan-dong, CHEN Ping, REN Tian-bao, JIA Fang-fang, LIU Guo-shun
• Format: Article
• Language: zho
• Published: Agro-Environmental Protection Institute, Ministry of Agriculture 2021-07-01
• Series: Journal of Agricultural Resources and Environment
• Subjects: tobacco; cadmium; hyperspectral; model; bp neural network
• Online Access: http://www.aed.org.cn/nyzyyhjxb/html/2021/4/20210404.htm
• ISSN: 2095-6819; 2095-6819

Heavy metal pollution has become the focus of environmental biology and crop quality and safety. In order to obtain the cadmium content of tobacco leaves quickly and accurately, four cadmium pollution levels were simulated. The spectral reflectance of tobacco leaves in each treatment was obtained using an American ASD spectrometer, and the cadmium content of tobacco leaves in different periods was measured. The sensitive bands with the best correlation with the cadmium content were selected, and the spectral parameters were used as input factors to establish a back propagation(BP) neural network model of the cadmium content in tobacco leaves. The results showed that as the cadmium content increased, the reflectance in the visible light and near-infrared range(400~910 nm) first decreased and then increased. In the wavelength range of 930~1 000 nm, the leaf reflectivity was positively correlated with the cadmium content in the tobacco leaf. The reflectance in the 1 000~2 500 nm broadband range first increased and then decreased. The selected spectral indexes ratio vegetation index(RVI) and normalized difference vegetation index(NDVI) were RVI(520, 710) and NDVI(530, 710), respectively. The <i>R</i>² of the BP neural network model was 0.681 and the root mean square error(RMSE) was 8.001, the model was tested, and the test results showed that the <i>R</i>² was 0.801 and the RMSE was 4.430. The results showed that the BP neural network model could provide a good prediction of the cadmium content in tobacco leaves.