Identification of Black Bean Geographical Origins Using Portable NIR Spectroscopy Combined with Fuzzy Orthogonal Linear Discriminant Analysis and Its Kernel Extension

Since different geographical origins of black beans have different nutritional and economic values, their classification is of great significance in the field of agriculture. Therefore, this study aims to develop a classification model for black bean origin identification based on chemometric analysis of near-infrared (NIR) spectral data. Due to the overlap and complex distribution characteristics of raw NIR spectral data, traditional linear discriminant analysis algorithms are unable to extract effective discriminative information from them. To overcome these limitations and improve the performance of the algorithms, fuzzy set theory was incorporated into orthogonal linear discriminant analysis (OLDA), and the fuzzy OLDA (FOLDA) was proposed to extract discriminant features from the NIR absorption spectra of black beans. FOLDA enhances sample diversity and better handles noisy data. Furthermore, by combining FOLDA with a linear kernel function, the kernel FOLDA (KFOLDA) algorithm was proposed to address the challenge of linear algorithms in processing nonlinear data. KFOLDA not only effectively captures and processes nonlinear information within the data but also significantly improves the model’s classification accuracy. The experimental results show that the K-nearest neighbor (KNN) classifier achieved classification accuracies of 86.4% and 95.5% for FOLDA and KFOLDA, respectively. Thus, the proposed model demonstrates satisfactory classification accuracy and performance metrics for distinguishing different geographical origins of black beans.