Feature selection and broad band bitumen content estimation of Athabasca oil sand from infrared reflectance spectra

Abstract

The estimation of bitumen content in oil sands in feed stock is critical to improve the processability of ore and for effective bitumen extraction. Broad band reflectance spectroscopy has the potential to achieve this goal in a non-destructive manner but spectral variability is known to be influenced by the water content and mineralogy of oil sands. This study addresses these issues and defines spectral features sensitive to bitumen and water content using wavelet analysis. A reliable model is then established to predict bitumen content based on spectral indices that minimize the influence of these factors.