Detection of Spinach Juice Residues on Stainless Steel Surfaces Using VNIR Hyperspectral Images

Abstract

Spinach is one of the most commonly consumed fresh-cut vegetables. Hygiene and sanitation in automated processing facilities have been an important issue. This research aimed to develop a line-scan hyperspectral imaging technique for detecting spinach droplets on a stainless steel surface. The hyperspectral imaging system uses UV-A (365 nm) light sources to obtain 3D hypercube data with spatial and spectral data in the visible and near-infrared (VNIR) region ranging from 400 to 1000 nm. Freshly made 100% spinach juice and distilled water were used to prepare juice dilutions at 20%, 10%, 5%, 2%, and 1% juice. For each of the six juice concentrations, fifteen droplets were placed on a stainless steel sheet, and VNIR hyperspectral image data was collected for the 6 × 15 array of droplets on the metal sheet. To detect and classify the diluted droplets on the spectral domain, three classification models (support vector machine, partial least square discriminant analysis, and random forest) and six pre-processing methods were implemented. Among them, support vector machine (SVM) showed the best classification accuracy with A = 0.95. Besides, the classification model used to reduce the number of wavelengths and calculation time, the genetic algorithm (GA) applied to the SVM showed the most accurate result as A = 0.90 among three methods. The developed classifier demonstrated potential for detecting and classifying spinach juice droplets on the surface of stainless steel sheet metal.