Abstract
The objective of this research was to develop a hyperspectral imaging system for estimating copper concentration in soils as an alternative to standard chemical analyses and to evaluate the analytical accuracy of the system using the visible–near-infrared and near-infrared regions. Hyperspectral imaging is a complex technology providing elevated information content. This work was carried out on air-dried <2-mm soil fraction contaminated by adding 20 mL of copper sulfate at concentrations ranging from 0 to 1000 mg of copper per kg of soil. The samples were scanned in random order and with orientation using visible–near-infrared and near-infrared spectrophotometers. A range of partial least squares regression models derived from the spectral arrays were tested on their ability to predict copper concentration. Significant correlations between predicted and known chemical concentrations were achieved with a correlation coefficient of 0.93 for the visible–near-infrared and 0.77 for the near infrared.
Acknowledgments
This study was supported by the project PROVISEBIO funded by the Ministero Italiano per le Politiche Agricole e Forestali (Law 41/82).