Design, integration, and evaluation of a low-cost system for automatic apple picking and infield sorting

Abstract

Fresh apple picking and sorting are continuous processes, and both are inefficient, costly, and labor-intensive. Existing research has predominantly focused on automating individual processes, resulting in discontinuous workflows, which have limited the advancement of overall orchard mechanization and lowered growers’ profitability. To address these challenges, a simple, compact, and low-cost system for automatic apple picking and infield sorting was integrated. The system hardware consists of a mobile platform, a new dual-arm picking subsystem integrating RGB-D cameras, low-cost five degree-of-freedom manipulators, vacuum suction cups, fruit receiving and transport components, and an infield sorting subsystem with a pitch-variable screw conveyor, an imaging chamber, and paddle sorters. Corresponding software algorithms were developed as well. In particular, a multi-frame fusion approach was proposed to minimize single-frame perception errors, and a motion control strategy was devised to improve the attachment success rate. Field experiments conducted in a spindle-canopy apple orchard generated promising results: a 52.5 % picking success rate, a 4.3 s average picking cycle time, a 23.8 % stem damage rate, and a 91 % sorting success rate with the speed of 2.4 apples per second. This developed system provides a new solution to automate both apple picking and infield sorting, which has a potential for practical application.