孟繁1,2　周馨曌2　吴烽云3　邹天龙2

（1.仲恺农业工程学院，广东 广州 510225

2.佛山市中科农业机器人与智慧农业创新研究院，广东 佛山 528010

3.广州商学院，广东 广州 511363）

摘要：水果采摘机器人是一种具有较大潜力的农业自动化技术，不仅需在复杂且不断变化的环境中连续作业，还面临地形、树木分布、天气变化、环境光照、遮挡等多种挑战。视觉同步定位与地图构建（SLAM）作为一种成本低廉且能够提供丰富语义信息的技术，有望提高水果采摘机器人的效率和自动化程度。近年来，视觉SLAM在水果采摘机器人领域取得了一系列重要进展，主要包括深度学习优化方法、基于点线特征的优化方法、基于RGB-D的视觉SLAM优化方法、动态环境中视觉SLAM优化方法、回环检测和后端优化方法等。未来，水果采摘机器人领域的研究将朝着更高的自动化程度和采摘效率方向发展；可能的发展方向包括更复杂的感知系统、更智能的决策算法、更强大的硬件支持等。此外，水果采摘机器人在多样化水果园中的适应性和鲁棒性研究也将引领这一领域的发展。通过不断推动视觉SLAM技术的创新，水果采摘机器人有望成为现代农业的重要工具，提高水果产量并减轻农业劳动力短缺的问题。

关键词：同步定位与地图构建；水果采摘机器人；机器视觉；自动化

中图分类号：TP391.4        　文献标志码：A      　   文章编号：1674-2605(2023)05-0002-07

DOI：10.3969/j.issn.1674-2605.2023.05.002

Research Progress in Fruit Picking Robots Based on Visual SLAM 

MENG Fan1,2  ZHOU Xinzhao2  WU Fengyun3  ZOU Tianlong2

(1.Zhongkai College of Agricultural Engineering, Guangzhou 510225, China 

2.Foshan Zhongke Innovation Research Institute of Intelligent Agriculture, Foshan 528010, China 

3.Guangzhou College of Commerce, Guangzhou 511363, China)

Abstract: Fruit picking robots are a promising agricultural automation technology, but they must operate continuously in complex and constantly changing environments, facing various challenges such as terrain, tree distribution, weather changes, environmental lighting, and occlusion. Visual SLAM, as a low-cost technology that can provide rich semantic information, has attracted widespread research interest as it has the potential to significantly improve the efficiency and automation of fruit picking robots. In recent years, visual SLAM has made a series of important progress in the field of fruit picking robots. Mainly including deep learning optimization methods; Optimization method based on point line features; Visual SLAM optimization method based on RGB-D; Visual SLAM optimization methods in dynamic environments; Loop detection and backend optimization methods. In the future, research in the field of fruit picking robots will move towards higher levels of automation and harvesting efficiency. Possible development directions include more complex perception systems, more intelligent decision algorithms, and stronger hardware support. In addition, research on the adaptability and robustness of robots in diverse water orchards will continue to lead the development of this field. By continuously promoting the innovation of visual SLAM technology, fruit picking robots are expected to become an important tool in modern agriculture, increasing production and alleviating the problem of agricultural labor shortage.

Keywords: simultaneous localization and mapping; fruit picking robots; machine vision; automation