Enhancing fruit recognition with robotic automation and salp swarm optimization for random forest classification

Abstract

In response to the growing demand for automation and labor-saving solutions in agriculture, there has been a noticeable lack of advancements in mechanization and robotics specifically tailored for fruit cultivation. To address this gap, this work introduces a novel method for fruit recognition and automating the harvesting process using robotic arms. This work employs a highly efficient and accurate model utilizing a single shot multibox detector (SSD) for detecting the precise fruit position. Once the fruit's position is identified, the angles of the robot arm's joints are calculated using inverse kinematics (IK). Finally, the optimal path planning is ensured by the salp swarm optimization (SSO) assisted random forest (RF) classification. This approach enables the precise management of robotic arms without any interference with either the fruits themselves or other robotic arms. Through meticulous consideration of these factors, our method ensures seamless operation in agricultural environments. Experimental validation demonstrates the effectiveness of these techniques in detecting apple fruits outdoors and subsequently automating their harvesting using robotic arms. This successful implementation underscores the potential for widespread application of our approach in enhancing efficiency and productivity in fruit cultivation.