基于深度学习的烟苗移栽机智能控制系统设计与试验

打开文本图片集

中图分类号：S223.9 文献标识码：A 文章编号：2095-2953（2025）12-0065-05

Design and Experiment of an Intelligent Transplanting Machine Control System Based on Deep Learning

LI Kun-zhen  ，CHEN Wen-hong， YANG Lu，LI Ling-mei ， JIANG Tao

WU Jian， WANG Feng

1.CollegeofMechanicalEngineeingandTransportation，SouthwestForestry University，Kunming Yunnan 65O224，China; 2. Qujing Company，Yunnan Provincial Tobacco Co.，Ltd.，Qujing Yunnan 655ooo，China）

Abstract：Aiming atthe problem that traditional transplanters have dificulty in precisely locating seedling holes under the film-covered tobacco transplanting agricultural mode in Yunnan tobacco-growing areas，this paper designs an intelligent transplanting control system based on deep learning and STM32 controler.The system uses abinocular camera tocollectdepth images of the seedling holesand realizes real-time recognition of seedling holes through the YOLOv8n model.Combined with the ranging principle of the binocular camera，the relative position deviation between thecenterof the seedling hole and the transplanting mechanism is calculated.The STM32 controler perceives thedisplacementofthe driving wheels via encoder signalsandcoordinateswith visual information to guidethe transplanter to locate seedling holes. Experimental results show that the positioning accuracy of the system is ±50mm ， meeting the agricultural requirements for tobaco seedling transplanting and providing an effective solution for precise operations in complex environments.

Key words：binocular vision positioning；STM32；agricultural machinery ；deep learning

在烟草的种植过程中，烟苗的移栽是一个重要的环节[1]，直接关系到烟苗的生长质量、成活率以及后续烟叶的质量与产量[2]。（剩余4215字）