Validation of the multivariate models for correlation between texture profile analysis, digital images, and NIR spectroscopy

Abstract

The present study evaluated the potential of near-infrared (NIR) spectroscopy, conventional digital images, and images with thermal and X-ray filters, coupled with partial least squares (PLS) for the determination of texture profile analysis (TPA), mainly hardness, adhesiveness, springiness, chewiness, gumminess, cohesiveness, and resilience. Furthermore, the approach can be classified as a multivariate multiproduct calibration model since different types of flour were incorporated into each single multivariate model. The models were validated through the parameters of merit estimation, considering accuracy, adjust, linearity, inverse of analytical sensitivity, and limits of detection and quantification. The variable importance in projection (VIP-scores) showed that hardness, gumminess, and cohesion share similar important variables in their modeling, while adhesiveness and elasticity also do. On the other hand, chewability and resilience share common variables in their respective models.