Rapid identification of adulterated rice based on data fusion of near-infrared spectroscopy and machine vision

Rice is susceptible to mold and mildew during storage. Metabolites such as aflatoxin produced during mildew have great harm to the health of consumers. A rapid identification approach of contaminated rice was developed based on data fusion of near-infrared spectroscopy and machine vision to satisfy the need for rapid detection of normal rice adulterated with moldy rice. The successive projection algorithm (SPA) was merged with principal component analysis (PCA) and support vector classification (SVC) to create the SPA-PCA-SVC method, which was based on variable selection, feature extraction, and nonlinear modeling methodologies. K-fold cross-validation and the sum of predicted residual squares were used to find the optimal number of main components. The model parameters were tuned using a genetic algorithm. Identification models of adulterated rice was established based on NIR spectroscopy, machine vision, and their fusion data using this method. The identification accuracy of the training set was 92.81%, 86.27%, and 99.35%, and the identification accuracy of the test set was 69.23%, 82.69%, and 96.15%, respectively. Compared to near-infrared spectroscopy and machine vision alone, the identification performance of the model built by fusion data is significantly improved. The findings demonstrate the viability of the near-infrared spectroscopy and machine vision data fusion method for the detection of contaminated rice, providing a theoretical foundation for the creation of online adulterated rice identification tools.