农机化研究2026，Vol.48Issue(3)：54-61,8.DOI:10.13427/j.issn.1003-188X.2026.03.008
基于改进YOLOv5s轻量化识别模型的智能采茶装置设计
Design of Intelligent Tea-picking Device Based on Improved YOLOv5s Lightweight Recognition Model
摘要
Abstract
Aiming at the problems of labor shortage and low picking efficiency in the tea-picking industry,a lightweight tea bud recognition algorithm based on the improved YOLOv5s was proposed,an intelligent tea-picking device was de-signed.Focusing on core issues including target recognition algorithms,3D precise positioning,and integrated picking and collection,and constructed a highly efficient and accurate intelligent tea-picking system by integrating 3D spatial positioning technology,mechanical structure innovations,and fluid mechanics advancements.Firstly,in view of the real-time and lightweight requirements for tea bud detection,an improved YOLOv5s model was proposed,which integrat-ed the GhostNet lightweight network,CARAFE up-sampling operator,and EIOU loss function.Secondly,to address the need for accurate positioning of tea-picking points,a 3D spatial positioning algorithm based on a binocular depth camera and image processing was proposed.Finally,a tea-picking device based on an XYZ three-axis motion platform,featu-ring integrated picking and collection functions was designed.Experimental tests were conducted on the tea-picking de-vice,and the results demonstrated that the proposed device was feasible and effective.The success rate of tea bud recoghition was(86.6±3.4)%,the success rate of picking after successful identification was(90±2)%.This paper provides an alternative approach regarding the concept,technical directions,and component selection for intelligent tea picking.关键词
采茶装置/机器视觉/YOLOv5s轻量化/采摘点定位/深度学习Key words
tea-picking device/machine vision/YOLOv5s lightweight/picking-point positioning/deep learning分类
信息技术与安全科学引用本文复制引用
Xu Jiahe,Gao Junran,Yu Fusheng,Zhang Xiao,Yu Zikun,Zhang Tongjia..基于改进YOLOv5s轻量化识别模型的智能采茶装置设计[J].农机化研究,2026,48(3):54-61,8.基金项目
山东省自然科学基金项目(ZR2024QE374) （ZR2024QE374）
