Development of Mangosteen Maturity Classification Model on Color Based Using Fuzzy Neural Network.

Abstract

Fuzzy Neural Network (FNN) has a capability to classify a pattern located within two different classes where a classical Neural Network (NN) is failed to do so. The fuzzy pattern classification is using membership degree on output of neuron as learning target. Objective of this research is to develop an artificial intelligence system model for non-destructive classification of fresh mangosteen using Fuzzy Neural Network. Component of color result in from image processing that influential against level of mangosteen’s maturity is used as input parameter. Percentage accuracy ratio of FNN model compare to NN for five, three, and two classification classes is 70:40, 86:65 and 90:90 respectively. The best result of FNN modeling is achieved on three class target classification (unripe, export and local) with green color index, value, a* u*, v*, entropy, contrast, energy and homogeinity as predictor parameters and 15 neurons hidden layer. Comparison of percentage capability of FNN against NN to identify the class is 100:0, 100:87 and 63:75.