A LOW-COST AI-BASED VISION-GUIDED ROBOTIC SYSTEM FOR PRECISION VEGETABLE HARVESTING

Abstract

The global agricultural sector is currently grappling with critical challenges, including acute labor shortages, escalating operational costs, and the inherent inefficiencies associated with traditional manual harvesting. To address these systemic issues, this study presents the design and implementation of a sophisticated AI-powered vegetable harvesting system that synergizes machine learning, advanced image processing, and robotics. The proposed system utilizes a hybrid machine learning approach implemented using MATLAB to process the visual data and execute real-time decision-making. The core of the recognition engine was developed using a comprehensive dataset of 1,200 images, partitioned into 800 training and 400 testing samples, to classify crops into three distinct growth stages: Ripe, Semi-ripe, and Unripe. Once a target is identified and classified, the system communicates with a low-cost hardware control unit composed of an Arduino Uno, L293D motor drivers, and a servo-driven plucking gripper to perform precise, "unmanned" harvesting.