农业工程2026，Vol.16Issue(2)：17-23,7.DOI:10.19998/j.cnki.2095-1795.202508039
基于改进YOLOv8的烟草移栽识别系统设计
Design of tobacco transplanting recognition system based on improved YOLOv8
摘要
Abstract
To improve of tobacco transplanting operations quality and optimize transplanting work,a tobacco transplanting recognition system based on improved YOLOv8 was proposed.Firstly,software and hardware design of tobacco seedling removal,picking,and transplanting device overall structure was designed.Key mechanisms such as conveying mechanism,information collection mechanism,and seedling picking and transplanting mechanism were analyzed for their operational principles and component selection.Then,a to-bacco transplanting recognition method based on YOLOv8s-MS-DCNv2-ILOSS(YOLOv8s-MSDI)was proposed to achieve accurate recognition of smaller targets.Finally,this method was applied into system to complete automatic tobacco transplanting operations.Results showed that YOLOv8s-MSDI model had an average accuracy across all classes of 98.03%and an F1 value of 99.45%in tobacco seedling recognition tasks,which were higher than those of YOLOv3-FDN,YOLOv5s,and Faster R-CNN.This analysis showed that YOLOv8s-MSDI improved accurate extraction and recognition of tobacco seedling characteristics,thereby reducing missed detection rate,improving tobacco transplanting quality and efficiency,and further enhancing tobacco transplantation operations informatization level.关键词
YOLOv8/烟草移栽/识别系统/深度学习/多尺度注意力/可变形卷积Key words
YOLOv8/tobacco transplanting/recognition system/deep learning/multi-scale attention/deformable convolution分类
农业科技引用本文复制引用
张四伟,陈刚,彭支光..基于改进YOLOv8的烟草移栽识别系统设计[J].农业工程,2026,16(2):17-23,7.基金项目
云南省烟草公司文山州公司2024年科技计划项目(20245326001) （20245326001）
