Precautionary analysis of sprouting potato eyes using hyperspectral imaging technology

Abstract

Sprouted potatoes are not allowed for healthy diet. A good knowledge of the sprouting stage of potatoes can help manage the storage conditions and guide market distribution, thus enabling the quality assurance of potatoes on table. This article presented an intelligent method for precautionary analysis of potato eyes based on hyperspectral imaging technique. Potential potato eyes were classified into two categories according to the time gap to the sprouting date, i.e. by-sprouting and pre-sprouting potato eyes, representing eyes about to sprout and eyes that will take a while to sprout. Features used for classification were extracted by two methods, including successive projections algorithm (SPA) and a newly-developed sine fit algorithm (SFA). Then classifiers of fisher discriminant analysis (FDA) and least square support vector machine (LSSVM) were utilized for classification of potential sprouting potato eyes. Results showed that FDA was more effective than LSSVM in classifying pre-sprouting and by-sprouting potato eyes, and SFA performed well in FDA classifier with the recognition accuracy of 95.3% for prediction set. It is concluded that hyperspectral imaging has the potential for predicting the sprouting stages of potato eyes.