Discrimination of heavy metal crop contamination using measurements of leaf spectra

Discriminating the type of heavy metal contamination of crops with Cu (Copper) or Pb (Lead) via hyperspectral remote sensing techniques is an effective way. A new method, the TDSF (Three-Dimensional Spectral Feature) discrimination method, is proposed to aggregate multiple spectral features of crop leaves to visually discriminate Cu and Pb contamination types. First, we processed the leaf spectra by using CR (Continuum Removal) and FOD (Fractional Order Derivative), and obtained CF (i) (CR and FOD, i is the order of FOD) spectra, and extracted the sensitive bands of each CF i spectrum. Second, we established the CPCI i (Cu and Pb Contamination Index) by using the sensitive band. We calculated the |r| of CPCI i with the contamination type. Then, the highest three |r| CPCI i (CPCI0.5, CPCI0.8 and CPCI0.6) were used to construct the TDSF of the leaf. Finally, we partitioned the TDSF in a three-dimensional space by using a discrimination plane to discriminate the heavy metal contamination type of TDSF corresponding leaves. The results showed that the optimal DA (Discriminating Accuracy) of the TDSF discrimination method was 95.24% in the experimental group and 100% in the validation group.