Grading method of apple by maximum cross-sectional diameter based on computer vision

Given the issued GB/T 10651-2008 “Apples”, an apple sorting system based on machine vision technology was designed according to the realistic condition that the commodity value was devalued because of the low rate and accuracy of sorting. For Fuji apple, the process of pretreatment was conducted with threshold segmentation of apple image by using R-B channels under the RGB color model and mean filter. And then the contour of the apple was extracted by line scanning. Two theoretical models were established for the classification of apple size: model one took the maximum distance between two points of the contour line as the grading standards, while model two took the diameter of apple maximum cross-section which was obtained by curve fitting. The two algorithms of classification were programmed by using VC 6.0. The test of forty apple samples indicated that the classification accuracy of model one was 93.3% while model two was 87.1%. The highest classification efficiency of two channels was 12 apples per second which satisfied the online commercial application requirement. It can provide references for the automatic sorting industry of the nearly spherical fruit and vegetable according to the industry standard.