Building Efficient Fruit Detection Model

Abstract

This manuscript conduct a deep routed survey of various existing fruit recognition system (FRS) for identifying different variety of fruits. From extensive survey it is seen the existing model does not perform well when color intensity and fruit size varies. Thus, it is important build an efficient feature extraction model to build a good training descriptor. Thus, this paper extract the fruit shape and color for establishing each fruit feature set. Our model is consisted of following phases, pre-processing (PP) phase, feature extraction (FE) phase, and testing phase. In PP stage, the image is resized. In FE stage, color, shape features, and scale invariant feature transform is used to build feature vector for each fruit variety. Then, in testing phase, we use K-Nearest Neighborhood classification algorithm to identify fruits. Our model can automatically recognize fruit along with calorie it can offer.