Development of near-infrared hyperspectral-based smart interface for glutinous rice quality detection

Abstract

Hyperspectral imaging (HSI) technology combined with chemometrics has offered a profound advancement in rice quality assessment. With the advent of this technology, glutinous rice quality such as moisture content, colour indices, protein, fat, and ash content are swiftly and accurately predicted without destroying the grains. The technology eliminates the laborious, time consuming, chemically demanding and expensive traditional method of grain quality determination. However, the complexity of HSI technology makes it more prominent in the research field because it requires high technical skills. Therefore, the development of a smart user interface (GUI) called HyperspecGlu in this study aids the rapid and nondestructive application of HSI data coupled with chemometrics for the determination of glutinous rice quality which includes colour change, golden index, moisture, protein, fat and ash content. The tool simplifies the HSI data processing and glutinous rice quality prediction, featuring data upload, preprocessing, model execution and result visualization through a click-and-run button. Employing three-stage processing techniques which include Savitzky-Golay first derivative techniques for spectral correction, redundant wavelength removal using variable importance space shrinkage approach and predictive model development gave a good prediction accuracy, which makes the HyperspecGlu reliable. Therefore, the HyperspecGlu toolbox is capable of swiftly detecting glutinous rice quality with high accuracy based on the HSI combined with chemometrics and the GUI makes the process available and accessible for users with little or no programming knowledge.