Application of neural networks in quality control of 'Jonagold' apples

This paper studies the possibilities to grade the 'Jonagold', a bicolour apple, using neural networks and Fisher"s linear discriminant analysis. In a first step the pixels are sorted into three areas, the blush colour area, the intermediate colour area and the ground colour area; the accuracy reached 95% with the neural networks. Int the second step the fruit are graded into four categories on their ground colour basis; the accuracy reached 68%whatever the method used. It si also shown that there is no need to separate the ground colour and the intermediate colour to compute the ground colour classification parameters.