Support vector machine model for predicting the cadmium concentration of soil-wheat system in mine reclamation farmland using hyperspectral data

Data for the spectral reflectance of soil and wheat were collected using an ASD field spectrometer in the laboratory, and the soil samples and wheat samples were collected for chemical analysis of Cadmium concentrations. A normalization spectral pre-processing method such as the weighted smoothing, first derivative, continuum removal and logarithm of reciprocal transform spectrometer were employed. On this basis, choosing the sensitive wave band which has significant correlations with Cadmium pollution in soil and wheat as the correlation factors, and establishing the cadmium pollution content in soil-wheat system prediction model. The result shows that both of the content of Cd in reclaimed soils tested on the sites by filling mining coal gangue and fly ash are qualified for the third level criterion of Environmental quality standards for soils, but neither of the wheat planted on it does. The correlation coefficient of prediction model of soil is 0.974, and the correlation coefficient of prediction model of wheat is 0.782, which prove that the model can be ideal for estimate the cadmium content of the soil and wheat in mine reclamation farmland. The study can provide new method for monitoring heavy metals pollution level of soil and crop timely, dynamically, widely and speedy by using hyperspectral data, and providing constructive idea for guarantee of food security.