Estimating soil heavy metal concentration using hyperspectral data and weighted K-NN method

Abstract

The potential hazard of heavy metals in reclaimed mine soil has influenced on the human health. The inversion analysis of hyperspectral data can be used to estimate heavy metal content of the soil effectively. In this paper, the characteristic bands are extracted by spectral pretreatment, including Savitzky-Golay (SG), Standard Normal Variety (SNV), First Derivative (FD), Second Derivative (SD), or Continuum Removal (CR) etc. Then, the weighted k-Nearest Neighbor (weighted k-NN) method is applied in the heavy metal inversion modeling to estimate the content of heavy metal with hyperspectral data. Compared with the widely used partial least squares regression (PLS), support vector machine (SVM) and k-Nearest Neighbor method (k-NN), the experimental results shown that the accuracy of weighted k-NN method was higher than other methods in the inversion of heavy Zinc (Zn), Chromium (Cr) and Plumbum (Pb).