Segmentation and classification of fruit images independent of image orientation using height width vectors

Abstract

Detecting fruits automatically through image processing is a bewildering job because of certain properties of several types of fruits. For each image in the database preprocessing removes background and separates the foreground layer.The images are segmented using image threshold method. The classification is done by using a classifier named KNN with the desirable aim of exact and fast classification of fruit.We have considered 170 images examining various cases like single fruit and multiple fruits of same type, different orientation of same type,cut fruit and full fruit images of same type,different colored fruit images of same type, multiple fruits of different types.. Our system successfully recognizes fruit images with 97.6% accuracy and provides an F-measure of 88%.