Review: The evolution of chemometrics coupled with near infrared spectroscopy for fruit quality evaluation

Abstract

Short wave near infrared (NIR) spectroscopy operated in a partial or full transmission geometry and a point spectroscopy mode has been increasingly adopted for evaluation of quality of intact fruit, both on-tree and on-packing lines. The evolution in hardware has been paralleled by an evolution in the modelling techniques employed. This review documents the range of spectral pre-treatments and modelling techniques employed for this application. Over the last three decades, there has been a shift from use of multiple linear regression to partial least squares regression. Attention to model robustness across seasons and instruments has driven a shift to machine learning methods such as artificial neural networks and deep learning in recent years, with this shift enabled by the availability of large and diverse training and test sets.

Model development and update of portable NIRS instrument for assessment of internal quality attributes of two navel orange varieties

“…Generally, the derivative spectra were applied to build models for removing baseline shift in fruit quality assessment by the portable NIRS instrument ( 3 ). The spectra of this study were transformed into second derivative (2D) spectra by second Savitzky-Golay derivative with window width of 13 points and fitting order of 2.…”

“…Currently, partial least square regression (PLSR) algorithm is adopted widely to build a robust model combined with pretreatment methods ( 3 ). The first or second derivative spectra were frequently applied to develop PLSR models rather than raw absorption spectra because of baseline correction with derivative pretreatment.…”

“…Variable selection also is a critical step for extracting useful variables or eliminating variables containing mostly noise during model development ( 3 ). Recently, variable selection algorithms were explored such as variable combination population analysis (VCPA) ( 8 ), iteratively retaining informative variables (IRIV) ( 9 ), Monte Carlo uniformation variable elimination (MC-UVE) ( 10 ), successive projections algorithm (SPA) ( 11 ) and window search ( 6 ).…”

The knowledge domain and emerging trends in apple detection based on NIRS: A scientometric analysis with CiteSpace (1989–2021)

“…Commonly used algorithms for modeling include multiple linear regression (MLR), partial least squares (PLS), artificial neural network (ANN), support vector machine (SVM), and so on. Novel algorithms like the firefly algorithm (FA), improved particle swarm optimization‐extreme learning machine (IPSO‐ELM), etc., have appeared in recent years (Anderson & Walsh, 2022 ; Bobelyn et al, 2010 ; Fan et al, 2019 ; Zhang, Huang, et al, 2022 ).…”

OpenVNT: An Open Platform for VIS-NIR Technology

“…At present, partial least squares regression (PLSR) is considered the standard regression method [ 22 , 23 ]. Parameterising PLSR models requires data preprocessing to minimise physical effects, e.g., scattering, etc., on spectral data that can hamper correlation.…”