Computer vision technologies for robotic harvesting and sorting automation : A reivew

다국어 초록 (Multilingual Abstract)

Global agriculture, including that of South Korea, is facing a structural labor crisis caused by the rapid decline of the working-age population and the severe aging of farmers. In particular, the automation of harvesting and sorting, the most labor-intensive stages of agricultural production, has emerged as a critical task for ensuring long-term sustainability. This study systematically reviews recent research trends in computer vision, one of the core technologies enabling agricultural automation, and proposes directions for advancing its practical application. The review focuses on publications from 2024 to 2025 in major academic databases such as Springer, MDPI, and Elsevier, with a comprehensive examination of essential computer vision techniques for automation system development. The scope spans the entire pipeline, from data acquisition to algorithm development and field deployment, including RGB, RGB-D, and multispectral/hyperspectral imaging for data collection; YOLO-based approaches for real-time object detection; 3D perception and pose estimation; quality and ripeness assessment algorithms; and model compression and optimization for deployment on edge devices.
By consolidating the current state of computer vision technologies for harvesting and sorting automation, this review provides strategic insights into future research directions, ultimately contributing to the realization of smart agriculture and the broader digital transformation of the agricultural sector.