Shape classification of Sunshine mango using machine vision

Abstract

This thesis presents the application of machine vision to classify the shape regularity of sunshine mango. The algorithm were successfully developed and programmed for image processing and image acquisition and then the regular and misshapen mangoes were able to classify using discriminant analysis. Using the acquired images from mangoes with different shapes, some essential geometrical features such as length, width, perimeter, area, major axis and minor axis were extracted from each image. Four size-shape parameter, area ratio, aspect ratio, circularity and compactness were used to analyse the mangoes between regular and misshapen. Based on discriminant analysis, three size-shape parameter (area ratio, aspect ratio, and circularity) were found to be effective in differentiate the regular and misshapen of mangoes. Overall the algorithm from discriminant analysis were able to classify 74% success rate to differentiate the regular and misshapen mangoes.