Authors

Abstract

Apple harvesting in orchards is a challenging task due to its dependence on manual labor. In addition, the reduction in skilled farmers and increasing employee costs have popularized mechanical harvesting. As a highly optimal apple picking method, apple harvesting robots integrate machine vision, image processing, robot kinematics, and multi-sensor fusion. This article reviews the vision system and mechanical structure of apple harvesters and evaluates the performance of robotic apple harvester prototypes from 2010 to 2018. Moreover, horticultural adaptability is also discussed in order to facilitate the expansion of orchard structures suitable for mechanized operations. We find that to solve the difficulties faced by apple harvesters, the development of mechanized apple harvesting and modern orchard structure applications must be accelerated. Furthermore, research into anthropomorphic control strategies has the potential to optimize picking patterns, while improvements in environment reconstruction and semantic segmentation can improve harvesting efficiency. Finally, the challenges and strategies based on the development status of robotic apple harvester are also analyzed. The review is intended to assist researchers in structure design, sensor choice and adaptability improvement of agricultural machinery and horticulture, and to influence the direction of the development of robotic apple harvester.

Abstract in Chinese

苹果收获是一项极具挑战性的工作，并具有劳动密集型的特点。劳动力减少和劳动力成本的增加促进了机械化收获的发展。作为一种高度优化的收获方法，苹果收获机器人集成了机器视觉、图像处理、机器人运动学以及多传感器融合。本文对苹果收获机器人的视觉系统和机械结构进行了综述，并对2010至2018年间的苹果收获机器人的样机性能进行了评估。此外讨论了苹果收获机器人的园艺适应性，以促进果园结构适应机械化作业的发展。为解决苹果收获机器人所面临的困难，除园艺与农机适应性之外，拟人化的控制策略具有优化采摘方法的潜力，而环境重建技术和语义分割的应用可能提高采摘效率。最后，根据苹果收获机器人的发展现状，分析了其面临的挑战和策略。本文旨在为研究人员在机器人的结构设计、传感器选择以及园艺适应性改进方面提高参考，并对苹果收获机器人的发展方向产生一定影响。