Vegetable grading using tactile sensing and machine learning

Abstract

With the advent of e-Commerce, online automated fruit and vegetable grading has acquired significant research interests recently. To get an idea of freshness of the fruits and vegetables the softness or firmness estimation is one of the basic steps of the grading process which has its roots grounded in bio-mimicking. This paper proposes and implements a basic touch sensitive robotic system for ripeness classification of two vegetables using machine learning approaches. Two piezoresistive force sensors mounted on a robotic gripper, controlled by a PIC32 microcontroller, are used to receive tactile feedback on predefined palpation sequence of an object. Eight statistical parameters are generated from the force values from each instance of the dataset during tactile palpation. These parameters serve as the input for Support Vector Machine (SVM) and K-Nearest Neighbor (KNN) approaches. A study and analysis of these Machine Learning methodologies has been conducted in this paper. © Springer International Publishing Switzerland 2014.