Development of Machine Vision System for Vegetable Quality Assessment

Abstract

Quality assessment of the vegetables and fruits is crucial because it impacts its market price, the customer’s preference, and most significantly human health. Efficient quality measurement may help to forestall the losses within the yield and quantity of the agricultural product. Evaluating tons of fruits and vegetables manually could be a tedious, costly, and an inaccurate process. Automating these processes using machine vision techniques will improve the productivity and accuracy of the system and also minimize the human intervention. In recent years, various techniques for effectively assessing food quality are successfully employed using computer vision. Image segmentation and classification are essential techniques used in machine vision of which color-based segmentation gives remarkable results when applied on vegetable and fruit images. This report gives a detailed description of varied techniques implemented for accurate grading of the vegetable. Color-based segmentation using K means clustering is adopted for image segmentation. While the deep learning method is used for accurate grading of vegetables. After training the model using the transfer learning approach, Convolutional Neural Network (CNN) provides promising results. The proposed machine vision system can accurately classify tomatoes into three categories based on their color and defects present on their exterior surface.