Detection of Monochromatic Foreign Matter in Curry Fried Rice Based on Hyperspectral Imaging

Abstract

Curry fried rice (CFR) is widely popular in China, however, in the industrial production process, foreign matter (FM) could be mixed into CFR inevitably, which threatened human health and enterprise disputes. Here reflectance hyperspectral imaging (HSI) technology combined with machine learning was used to detect monochromatic FM in CFR. After optimizing SG as pretreatment method and screening the feature information by PCA and CARS, the models BP-ANN, KNN, LDA, and RF were applied to distinguish FM and CFR. It turned out that BP-ANN based on CARS selected variables proved greater performance with RP2 of 98.08% and RMSEP of 0.00. Based on the optimal model, the results of the visualization provided excellent identification for the FM and CFR. The study demonstrated that HSI technology is an effective way of detecting monochromatic FM in CFR and also set a technical example for FM detection in complex food matrix.

Keywords: Curry fried rice, Monochromatic foreign matter, Hyperspectral imaging technology, Machine Learning