| Vegetables are essential food for human being and their safaty quality is extremely important to human health. But heavy metal contamination is one of factor to restrict the vegebales health quality. Cadumim (Cd) is one of the most toxic heavy metal elements, has strong water solubility and can easily be absorption by vegetables when compared to other corps. Thus, finding a sort of method can detect Cd content in vegetable through a fast, accurate, non-destructive way that is important significance to diet safety and industrial development of vegetable. Hyperspectral Remote Sensing Technology (HRST) has been widely used in the precision agriculture that can provide a feasible ability for accurate, nondestructive, dynamic diagnosis of crop heavy metal content. Currently, the study about the accurate prediction Cd content of the edible parts in vegetable is less. Especially, using the data of vegetable at growth early predicts heavy metal content of the plant organ or the edible part at later growth stage.In order to explore a suitable method including finding out the most sensitive parameter for monitoring the heavy metal contents in the edible organs of Brassica juncea Var. tumida Tsen et Lee. and Capsicum frutescens L. under different Cd contaminations, HRST were used in this research. Nine treatments including 0 (CK),0.1 (T1),0.2 (T2),0.3(T3),1 (T4),2 (T5),5 (T6),10 (T7) and 20 (T8) of Cd in mg/kg of soil dry weight were applied to B. juncea Var. tumida Tsen et Lee. and C. frutescens L. ASD FieldSpec 3 HR was utilized to scan the hyperspectral reflective rate of leaf samples, and inductively coupled plasma emission spectrometer (ICP-AES) was used to measure the Cd concentrations at the different stage after the beginning of treatments. In this study, accumulation and transfer characteristics of Cd in organs and leaf hyperspectral reflectance change rule were studied in the B. juncea Var. tumida Tsen et Lee. and C. frutescens L. under different Cd concentration. The sensitive growth period, sensitive bands and sensitive physiological parameters were selected by the correlation analysis and stepwise regression analysis between original spectral datum, first derivative spectral datum and wavelet transform spectrum with Cd content in leaves. According to these sensitive indexes, fitting models including simple regression models, multiple regression models and wavelet models were established to estimate the Cd content in vegetable edible organs. The research results show that:(1) With the Cd concentration increasing in soil, the content of Cd in lump and pepper gradually increased, and bio-concentration factor (BCF), translocation factor (TF) of lump and pepper had a first increasing then decreasing trend. TF values were higher than 1 at expending period under middle and high Cd concentration treatment groups of B. juncea Var. tumida Tsen et Lee., which meaned B. juncea Var. tumida Tsen et Lee. having the characteristics of hyperaccumulators. Relationship was better betweeb lump and pepper Cd content and leaves Cd content at four stage respectively. The correlation coefficients of B. juncea Var. tumida Tsen et Lee. expanding period and C. frutescens L. maturing period were best, and the value of R2 were 0.942, and 0.925 (P<0.01) respectively. Cd content in lump and pepper would exceed the national security standard, and will exist consumption risk when values of the soil Cd concentration higher than 0.1 mg/kg (T1),0.3 mg/kg (T3) respectively.(2) Spectral reflectance of leaves significantly changed at visible bands and near infrared bands of B. juncea Var. tumida Tsen et Lee. and C. frutescens L. under different Cd concentrations treatment. For B. juncea Var. tumida Tsen et Lee., reflectance change was the most obvious at expanding period. There was significant difference of leaf reflectance between CK group and other eight treatment groups. Reflectance of the visible bands (350 nm-680 nm) decreased, while reflectance of the near infrared and far infrared bands mainly increased with increasing of Cd concentration. Reflectance of budding stage in C. frutescens L. had significantly increased at near infrared bands of all treatment groups with increasing of Cd concentration.(3) Sensitive physiological parameters (GRI、GRN and NDVI) of B. juncea Var. tumida Tsen et Lee. had better potential to estimate the Cd content of lump, and the fitting coefficient between measured value with the predicted value of lump Cd content reached 0.719 (P<0.01). Quadratic models of sensitive bands R382 and R359 of the first derivative spectra could well predict the Cd contents of in lump, and correlation coefficient were 0.763 and 0.758 (P<0.01) respectively. The linear and the nonlinear inverse fitting model of DR and R2168, and the linear model of R464 also could well predict the Cd contents of in pepper fruit. Value of Cd content was predicted by linear model of sensitive band R2168 had best fitting ability with measured Cd content, and correlation coefficient was 0.771(P<0.01).(4) Wavelet function db5’s fitting model could be used to estimate Cd content of lump in B. juncea Var. tumida Tsen et Lee., the fitting precision between measured Cd content with the predicted Cd content was 0.618. The ability of wavelet function db3 to predict Cd content model was poor for C. frutescens L., and the fitting precision between the predictive value with the measured value was 0.567 (P<0.01).This study of correlation analysis between B. juncea Var. tumida Tsen et Lee. and C. frutescens L. leaf spectral reflectance with leaf the cadmium content found that there was certain regularity each other. And this regular will have change at different growth stage. Estimate ability of inversion Cd content of vegetable editable organs, fitting model of using stepwise regression method to establish was better than the wavelet analysis function’s. This study showed that models, based on spectral parameters and dimension, both can predict Cd content of lump and pepper fruit, but its accuracy is needed to optimize by increasing the sample quantity, etc. |
PDF Full Text Request