Apples shape grading by fourier expansion and genetic program algorithm

Abstract

Apple is a very important fruit in China. The shape of the apple is important indices for classifications. An image collecting system was developed and 4 images are obtained from each apple in this study. In order to describing the irregular shapes of apples, Fourier expansion was developed to reduce the dimensionality of the edge points of an image to a set of 33 Fourier coefficients. These coefficients and variables were used as feature parameters. A new method called organization feature parameter based on formulae expression tree by using genetic algorithm was proposed in this paper. It could solve the problem how to getting optimum feature parameters. By applying "step decision tree" descriptor in combination with the new method to identify the shape of apples, the grade judgment ratios for "extra", "categories II" and "reject" are high, but the ratio for "category I" is not high. © 2008 IEEE.