高光谱遥感监测农作物重金属污染已成为遥感研究的重要内容之一。 受污染的作物叶片中重金属含量映射到光谱上的信息量差异较微弱, 如何灵敏地挖掘其所包含的价值信息具备一定挑战性。 以农作物叶片光谱为研究对象, 通过多个光谱特征波段组合的方式, 提出了一种铜污染植被指数(CPVI)的污染程度探测模型, 来表征重金属Cu对农作物的污染程度。 首先设置盆栽实验, 将不同浓度梯度的CuSO4·5H2O粉末添加到土壤中, 模拟Cu污染土壤环境, 胁迫玉米生长。 采集玉米穗期的老、 中、 新叶片光谱, 测定叶片中Cu2+含量及相对叶绿素浓度。 而后利用随机选取的58组玉米叶片光谱作为实验数据, 在380～900 nm波长范围内选取波长λ1和λ2的两组叶片光谱反射率并计算相应的CPVI［λ1, λ2］模型指数及其与对应叶片中Cu2+含量的皮尔逊相关系数, 得到相关性特征绝对值矩阵。 其次, 根据得到的相关性特征绝对值矩阵, 提取皮尔逊相关系数较高的光谱特征波段690和465 nm, 并结合波段850 nm建立针对玉米叶片的铜污染植被指数(CPVIm)。 之后, 利用另外26组数据对CPVIm指数进行检验, 同时将该指数与归一化植被指数(NDVI)、 陆地叶绿素指数(MTCI)等常规植被指数进行比较以验证CPVIm的有效性与优越性。 结果表明, NDVI, MTCI, REP和DVI与叶片中Cu2+含量相关系数最高仅为0.68, 残差平方和RSS最低为70.99, 而CPVIm与叶片中Cu2+含量显著负相关, 相关系数达-0.80, 残差平方和为48.52, 均优于NDVI和MTCI等常规植被指数, 证明CPVIm对重金属胁迫更敏感。 同时利用两期不同年份不同品种的玉米光谱数据进行CPVIm指数的鲁棒性验证, CPVIm与叶片Cu2+含量的相关系数r分别为-0.90和-0.96, 均显著相关, 说明该指数对于不同品种的玉米污染程度探测仍具有良好的适用性。 另外, 利用玉米叶片中Cu2+含量、 CPVIm和叶片中叶绿素相对浓度构建三维分析模型, 从空间角度直观地反映了三者之间具有一定的相关关系。 通过光谱特征波段组合方式构建的CPVI探测模型可作为评价农作物重金属污染程度的参考方法, 基于该方法构建的CPVIm指数可有效甄别玉米受重金属Cu2+污染的程度。

The use of hyperspectral remote sensing to monitor heavy metal pollution in crops has become an important part of remote sensing research. The difference in the amount of heavy metal content in the contaminated crop leaves mapped to the spectrum is weak, so it is challenging to dig sensitively the value information contained in it. In this paper, based on the spectrum of crop leaves, a pollution detection model of copper pollution vegetation index (CPVI) was proposed by combining multiple spectral feature bands to characterize the pollution degree of heavy metal copper on crops. Firstly, a pot experiment was conducted to add CuSO4·5H2O powder with different concentration gradients to the soil to simulate copper-contaminated soil environment and stress corn growth. The spectra of old, middle and new leaves at the ear of corn were collected, and the Cu2+ content and relative chlorophyll concentration in the leaves were determined. Then, using 58 randomly selected corn leaf spectra as experimental data, the spectral reflectances of the two groups of wavelengths λ1 and λ2 were selected in the wavelength range of 380~900 nm. The Pearson correlation coefficient between CPVI ［λ1, λ2］ and Cu2+ content in the corresponding leaves was calculated, and the absolute value matrix of correlation characteristics was obtained. Secondly, according to the obtained correlation feature matrix, the characteristic band of 690 and 465 nm with high correlation coefficient was extracted and combined with the band 850 nm to establish the Copper pollution index of maize (CPVIm). After that, CPVIm index was verified based on 26 other groups of data, and Normalized difference vegetation index (NDVI), MERIS terrestrial chlorophyll index (MTCI) and other conventional vegetation indexes were compared to verify the effectiveness and superiority of CPVIm. The results showed that the highest correlation coefficient between NDVI, MTCI, REP, DVI and Cu2+ content in leaves was 0.68, and the lowest residual sum of squares was 70.99. However, CPVIm was significantly negatively correlated with Cu2+ content in leaves. The correlation coefficient was -0.80, and the residual sum of squares was 48.52, which were better than conventional indexes such as NDVI and MTCI. It proved that CPVIm is more sensitive to heavy metal stress. At the same time, the robustness of CPVIm index was verified by using the spectral data of different varieties of maize in different years. The correlation coefficient of CPVIm and Cu2+ content were -0.90 and -0.96, respectively, which were significantly correlated. It shows that CPVIm is still suitable for detecting the pollution degree of different maize varieties. In addition, using Cu2+ content, CPVIm and chlorophyll relative concentration in maize leaves, a three-dimensional analysis model was constructed, which reflected the correlation between them from a spatial point of view. The CPVI detection model based on the combination of spectral characteristic bands can be used as a reference method to evaluate the pollution degree of heavy metals in crops. The CPVIm index based on this method can effectively identify the degree of heavy metal Cu2+ pollution in maize.