Particle Swarm Optimized Fuzzy Model for the Classification of Banana Ripeness

Abstract– Ripeness classification is an important task in the
post harvest of banana (Musa sp.) to regulate the ripening
treatment. In this paper, a fuzzy model is formulated to classify
the level of banana fruit into unripe, ripe and overripe stages.
Peak hue and Normalized Brown Area are the features of the
region of interest extracted from hue channel and opponent colors
of CIELa*b*. In a fuzzy modeling, defining the linguistic labels,
mapping it to the appropriate intervals and constructing the rule
base play vital roles. Classification and Regression Tree (CART)
algorithm is applied to model the intervals of feature space and
rule base for the fuzzy system. The parameters of fuzzy model are
tuned with Particle Swarm Optimization technique (PSO). The
proposed work is evaluated on MUSA database comprising of
banana samples at different ripening stages. Experimental results
show that the fuzzy model achieves the average classification rate
of 93.11% which outperforms state of the art algorithms.