基于光谱参数对小白菜叶片镉含量的高光谱估算

为实现利用高光谱技术快速、准确、无损地检测叶类蔬菜叶片重金属镉污染情况,通过采用室内盆栽试验,检测了小白菜在6个不同的镉浓度梯度0 mg/kg(CK)、0.5 mg/kg(T1)、1 mg/kg(T2)、5 mg/kg(T3)、10 mg/kg(T4)和20 mg/kg(T5)下的叶片高光谱反射率及其镉含量。利用相关分析和逐步回归的统计方法对叶片原光谱、一阶导数光谱和光谱参数与镉含量进行统计分析,确定了反演叶片镉含量的敏感光谱参数,并建立了估算叶片镉含量的参数模型。结果表明:(1)在540 nm附近和红外区域,叶片光谱反射率随着处理浓度的增加呈下降趋势。T1组叶片光谱与对照组的光谱没有明显的变化差异;(2)原光谱与镉含量的敏感波段主要在690—1300 nm,相关系数最高的波段是782 nm。一阶微分光谱与镉含量的敏感波段在黄边、红外、近红外和远红外范围均有分布;(3)反映植物色素、水含量和细胞结构的参数MCARI(叶绿素吸收反射修正指数Modified Chlorophyll Absorption Reflectance Index),SDy(黄边面积Yellow Edge Area),WI(水质指数Water Index),DCWI(病态水分胁迫指数Disease Water Stress Index),SDr(红边面积Red Edge Area)和Dr(红边幅值The Amplitude of the Red Edge)可分别作为反演镉含量的敏感光谱参数,其倒数回归模型能够较好地反演镉污染下小白菜叶片的镉含量;(4)镉胁迫处理15 d时,建立的SDr的倒数模型估算处理30 d时小白菜叶片镉含量的效果最优。研究表明红边面积参数可以用于估算小白菜叶片的镉含量,可为评价小白菜的食用安全提供科学方法。

Hyperspectral estimation of the cadmium content in leaves of Brassica rapa chinesis based on the spectral parameters

As one of the most phytotoxic heavy metals, cadmium (Cd) is easily taken up by vegetables. However, more than 0.2 mg/kg Cd content in leafy vegetables will seriously impact human health according to the standard of maximum levels of contaminants in food (GB2762-2012). Thus, it is of great significance to closely monitor the content of Cd content in leaves of the vegetables. Hyperspectral remote sensing (RS) techniques could monitor the content of metal and non-metal in crops, through a rapid and non-destructive way compared with traditional methods. In order to explore a suitable method for monitoring the heavy metal contents in the leaves of Brassica rapa chinesis under different Cd contaminations, hyperspectral remote sensing techniques were adopted for this research. In the mean time, the most sensitive parameter for certain Cd content in the leaves could also be explored. Six treatments including 0 (CK), 0.5 (T1), 1 (T2), 5 (T3), 10 (T4), 20 (T5) mg/kg of Cd in soils (calculated according to dry weight) were applied for growing B. rapa chinesis. ASD portable field spectrometer was utilized to scan the hyperspectral reflective rate of leaf samples, and Flame atomic absorption spectrometer was used to measure the Cd concentrations, on the 15th and 30th day after the beginning of treatments, respectively. After correlation analysis and stepwise regression between original spectral datum, first derivative spectral datum, spectrum parameters and Cd contents, the sensitive parameters were determined. According to these sensitive parameters, fitting models used to estimate the Cd content in vegetable leaves were established. Results showed that: (1) Both near wavelength of 540 nm and near infrared bands, the spectral reflectance of leaves were generally decreased with increasing of Cd concentration, while no significant difference was detected between the graphs of B. rapa chinesis under T1 and CK. (2) Sensitive bands of the original spectrum were mainly distributed from 690 nm through 1300 nm and the correlation coefficient of wavelength 782 nm was the highest. For first derivative spectra, a range of sensitive bands that was correlative with the Cd content located in the yellow edge, infrared, near infrared and far infrared; (3) Sensitive parameters MCARI (Modified Chlorophyll Absorption Reflectance Index), SDy (Yellow Edge Area), WI (Water Index), DCWI (Disease Water Stress Index), SDr (Red Edge Area) and Dr (The Amplitude of the Red Edge), which reflected the changes of pigment, water content and cell structure, could be used to estimate the Cd content of the leaves. The nonlinear inverse fitting models of the 6 sensitive parameters can well predict the Cd contents of leaves of B. rapa chinesis under different Cd stresses; and (4) The nonlinear inverse fitting models of the SDr derived from the data collected on the 15th day was best fitted for the Cd content in B. rapa chinesis leaves on 30 d. This study showed that the red edge area parameters can be used to estimate the Cd content in B. rapa chinesis leaves. Hyperspectral remote sensing technique is suitable for evaluating the edible security of B. rapa chinesis, and can provide fundamental information for the detection of Cd content in vegetables.