Identifying the authenticity and geographical origin of rice by analyzing hyperspectral images using unsupervised clustering algorithms

The present study evaluated the capability of the hyperspectral imaging system (HSI) as a rapid and non-destructive technique to identify the authenticity and origin of three Iranian rice varieties (Shiroudi, Khazar, and Hashemi) from three main origins. Captured raw spectral data were pre-processed using de-trending (DT), multiplicative scatter correction (MSC), and standard normal variant (SNV), and then were fed to principal component analysis (PCA) for the visual discrimination of the samples and data reduction. Next, since in real applications, the system might face unknown rice samples with unknown patterns, three unsupervised algorithms, self-organizing map (SOM), automatic clustering by artificial bee colony (ABC), and k-means algorithms were applied for clustering the samples in their original group. Results illustrated that SOM and k-means clustering algorithms led to the reliable grouping of the rice varieties. The models indicated that the high-yielding rice (Shiroudi and Hashemi) from two different origins with similar weather conditions were distinguished close to each other and completely separated from the third variety (Khazar). However, the automatic clustering method separated the varieties with less accuracy than the other two methods. Finally, the HSI system coupled with unsupervised algorithms provided satisfactory results and could be used as a reliable, out-lab, and fast method for ‎authenticating rice‎ varieties from different geographical origins.