Classification of tomatoes with different genotypes by visible and short-wave near-infrared spectroscopy with least-squares support vector machines and other chemometrics

Abstract

The feasibility of using visible and short-wave near-infrared (Vis/SW NIR) diffuse reflectance spectroscopic technique in the 400–1000 nm region to classify tomatoes with different genotypes was investigated. The discrimination performance of different chemometrics techniques, including least-squares support vector machines (LS-SVM), discriminant analysis (DA), soft independent modeling of class analogy (SIMCA), discriminant partial least-squares (DPLS) was compared. The results show that LS-SVM has the same performance, 100% of success rate, which is the same with DA but requires less time if a large variety of tomatoes are to be discriminated according to genotypes. Selecting the most important wavelengths that affect the classification rate most was proposed. The results show that the overall identification rate by LS-SVM method decreases from 100% to 96.81% in validation set using the selected wavelengths, which is acceptable. The importance of these conclusions may be helpful to transfer Vis/SW NIR technology from the laboratory to the industrial world for on-line and portable application.