Heavy metal content estimation in leaf by spectrum features of plant in De-Xing copper mining area

The estimation of heavy metal content in leaf is important to the integration of remote sensing into evaluation the ecological conditions in mining area. In this paper, correlation analysis and multivariable statistical methods were used to build hyperspectral models for the heavy metal (e.g., Cu) estimation with independent variables such as spectral reflectance, derivatives and ratio indices. Results showed that the heavy metals often display effects on plants as they changed plant moisture content, the pigment content, the leaf structure, and so on. Stepwise Multiple Regression Model predicted value and the actual value comparison showed that the model is stable, and the relative deviation about single plant mostly below2%. The first and second order differential spectrums were employed on three kinds of herbs synthesized also, the first order differential model proved better, and its relative deviation is lower than 15%.