Le Matematiche (Nov 1993)

A neural network approach to discrimination between defects and calyces in oranges

  • Salvatore Ingrassia,
  • Enrico Commis

Journal volume & issue
Vol. 48, no. 2
pp. 273 – 285

Abstract

Read online

The problem of automatic discrimination among pictures concerning either defects or calyces in oranges is approached. The method here proposed is based on a statistical analysis of the grey-levels and the shape of calyces in the pictures. Some suitable statistical indices are considered and the discriminant function is designed by means of a neural network on the basis of a suitable vector representation of the images. Numerical experiments give 5 misclassifications in a set of 52 images, where only three defects have been classified as calyces.