Application of support vector machine to apple classification with near-infrared spectroscopy

An apple NIR spectroscopy acquisition device was developed to diminish experimental errors in apple clasification. To improve and simplify the prediction model of classification, a new machine learning method called Support Vector Machine(SVM) was used to build near infrared(NIR) spectrum classification models for apples from different production areas and of different varieties. By choosing RBF as the core function, the suitable preprocessing method, penalty coefficient C and normal coefficient γ, for the model were determined. The classification accuracies for training set and test set of the SVM model for different apple varieties were both 100%, while those of the apples from origin areas were 87% and 100%, respectively. Compared with the discrimination analysis model, the SVM models' accuracy increased by about 5%. The results show that SVM has a perfect performance in establishing the NIR models for apple classification.