Identification of biodiesel feedstock in biodiesel/diesel blends using digital images and chemometric methods
Abstract
This study aims to identify the biodiesel feedstock (cottonseed, sunflower, corn or soybean oil) in biodiesel/diesel blends using digital images and chemometric methods. For this purpose, colour histograms (extracted from digital images) coupled with supervised pattern recognition techniques: Soft Independent Modelling of Class Analogy (SIMCA), Partial Least Squares Discriminant Analysis (PLS-DA) and the Successive Projections Algorithm for variable selection associated with Linear Discriminant Analysis (SPA-LDA) were used. SPA-LDA coupled with intensity histograms provided better results by selecting 12 variables alone, achieving only one error of classification in the external validation (test) set. Thus, the proposed methodology presents a noteworthy eco-friendly approach for identifying the biodiesel feedstock in biodiesel/diesel blends using a simple, fast, inexpensive and non-destructive analytical tool.