摘要
视觉是采摘机器人感知外部环境的重要手段之一。采摘机器人的采摘效率与准确率很大程度上受到其视觉子系统识别与定位性能影响。近年来,研究者围绕采摘机器人的视觉系统开展了大量研究。伴随人工智能深度学习的发展,机器人视觉研究取得了较大进步,采摘机器人进入了实际应用阶段。本文针对采摘机器人视觉系统研究现状,主要从视觉数据获取方法、采摘点识别技术和采摘点定位技术3个方面进行总结分析,指出了当前采摘机器人采摘点识别与定位过程中面临的一些挑战,最后对未来采摘机器人视觉系统的发展进行了展望。
Vision is one of the important means for picking robots to perceive the external environment.The picking efficiency and accuracy of the picking robots are largely affected by the recognition and positioning performance of its vision subsystem.In recent years,a lot of researches have been carried out regarding the visual system of picking robots.With the development of artificial intelligence deep learning,research of robot vision has made great progress,which opened the stage of practical application of picking robots.In view of the research status of the picking robots visual system,this paper summarized and analyzed from three aspects,such as visual data acquisition method,picking point recognition technology and picking point positioning technology,and pointed out some challenges faced by the current picking robots in the process of picking point identification and positioning.Finally,the development of visual system for picking robots in the future was prospected.
作者
金寿祥
周宏平
姜洪喆
孙梦梦
JIN Shou-xiang;ZHOU Hong-ping;JIANG Hong-zhe;SUN Meng-meng(School of Mechanical and Electronic Engineering,Nanjing Forestry University,Nanjing 210037,China)
出处
《江苏农业学报》
CSCD
北大核心
2023年第2期582-595,共14页
Jiangsu Journal of Agricultural Sciences
基金
国家自然科学基金青年科学基金项目(32102071)
江苏省农业科技自主创新基金项目[CX(20)3040]
江苏省高等学校自然科学研究项目(21KJB220013)。
关键词
视觉数据
采摘点识别
三维定位
深度学习
visual data
picking point identification
three dimensions positioning
deep learning
