Development of calibration models to predict texture and total soluble solids in jelly using hyperspectral imaging

Onnom, Poonnada; Teerachaichayut, Sontisuk

doi:10.1063/1.5066852

Development of calibration models to predict texture and total soluble solids in jelly using hyperspectral imaging

Author(s)

Onnom, Poonnada

Teerachaichayut, Sontisuk

Date Issued

November 9, 2018

Type

Conference Paper

DOI

10.1063/1.5066852

Abstract

The quality of foods for the elderly is important which must be controlled in the manufacturing process. A nondestructive technique that is fast, accurate and reliable is required in order to produce the acceptable products for consumers. Hyperspectral imaging technique was used to establish the calibration models for texture and total soluble solids (TSS) of jelly using partial least squares regression. A set of 99 samples were used for calibration and a set of 49 samples were used for prediction. Spectral pretreatments were investigated in order to develop the highest efficiency of the calibration models. The standard normal variate (SNV) spectral pretreatment was selected for development of the model for texture while the smoothing spectral pretreatment was selected for the TSS model. The accuracy of the calibration models for texture obtained the coefficient of determination (R²) of 0.882 and the root mean square error of prediction (RMSEP) of 0.04N and the accuracy of the calibration models for TSS obtained R² of 0.969 and RMSEP of 1.32°Bx. The results showed that HSI can be used for nondestructive determination of texture and TSS of jelly. It can be applied for an online sorting system for the manufacturing process.

Citation

Aip Conference Proceedings, 2030, 2018