Fuzzy-based determination of model and parameters of dual-wavelength vision system for on-line apple sorting

Abstract

One of the basic tasks in developing a multi-wavelength machine vision system for on-line fruit sorting is to determine the sorting model and system parameters including the wavelength (or filter) combination and the sorting threshold. A fuzzy method was proposed to evaluate the performance of a dual-wavelength vision system developed for on-line sorting of apples. Three sorting models were developed based on the sizes of defect blobs extracted from apple images. Twelve dual-filter combinations were tested and compared under different sorting requirements. Experimental results showed the effectiveness of the fuzzy method in determining the model and system parameters for the on-line dual-wavelength vision sorting system to meet any user-specified sorting requirements for both good and defective apples.