The Development of Machine Vision System for Sorting Passion Fruit using MultiClass Support Vector Machine

Abstract

This research aims to develop a machine vision system for sorting passion fruit based on the classification of the ripeness level. For years in the food processing industry, the sorting process has been done manually which is time-consuming and produces unreliable classification. To cope with this problem, this research proposed a machine that can sort passion fruit according to the ripeness level automatically. The system is equipped with a pneumatic drive, gripper collector, camera and bowl selector. Passion fruit is taken by the gripper collector and rotates 360 ° in front of the camera so that all the passion fruit surfaces can be captured. The camera feeds the images for the sorting process in three categories, i.e., ripe, nearly ripe and unripe using a computer vision-based intelligent system. The used computer vision method is K-Means Clustering as feature extraction and Multi-Class Support Vector Machine (MSVM) for classification of passion fruit ripeness level. The results show that Fruit Passion Sorting Machine can achieve 93.3% accuracy with an average time to sort each fruit is 0.94128 seconds with RBF kernel function parameters C = 25 and γ = 1e-5.