基于改进YOLOv8n的设施高垄草莓识别方法  

Strawberry identification method based on improved YOLOv8n

作　　者：李娜 陈丰 张华 苏祥祥 吴镛 朱婷倩 张运来 LI Na;CHEN Feng;ZHANG Hua;SU Xiangxiang;WU Yong;ZHU Tingqian;ZHANG Yunlai(College of Mechanical Engineering,Anhui Science and Technology University,Fengyang 233100,China)

机构地区：[1]安徽科技学院机械工程学院,安徽凤阳233100

出　　处：《安徽科技学院学报》2025年第1期38-47,共10页Journal of Anhui Science and Technology University

基　　金：安徽省科技特派员农业物质技术装备领域揭榜挂帅项目(2022296906020001);安徽省高校自然科学研究重大项目(KJ2021ZD0110);2022年度合工大智能院“科技成果培育专项”(IMIPY2022019);安徽省智能农机装备研究院建设协同创新项目(GXXT-2023-110)。

摘　　要：针对草莓生长环境中果实堆叠、叶片枝干遮挡和光线不均等问题导致的低识别率,实现设施高垄草莓的识别检测,改变草莓采摘依赖人工的现状。本文提出基于改进YOLOv8n的草莓识别模型,构建MBCA模块作为YOLOv8n的主干网络特征提取模块;构建AVCStem模块替换颈部网络的3个C2f模块,并将GSConv替换颈部网络的普通卷积,保持轻量化并进一步提升精度。改进后YOLOv8n模型mAP为96.8%,R为93.8%,P为92.4%。该研究可实现成熟草莓的识别,有助于进一步推动智能化草莓采摘机器人的研发与应用。To realize the identification and detection of strawberry in the facility ridge in order to avoid the low recognition rate caused by fruit stacking,leaf and branch shading,and uneven light in the strawberry growing environment,and change the status quo of strawberry picking relying on manual labor.A strawberry recognition model based on improved YOLOv8n was proposed.The MBCA module was constructed as the backbone network feature extraction module of YOLOv8n.The AVCStem module replaced the three C2f modules of the neck network,and the GSConv replaced the common convolution of the neck network,keeping the lightweight and further improving the accuracy.The experimental results showed that the R,P and mAP of the improved YOLOv8n model were 96.8%,93.8%,92.4%,respectively.This study could realize the identification of ripe strawberries,which was helpful to further promote the development and application of intelligent strawberry picking robots.

关 键 词：设施高垄草莓 YOLOv8n 图像识别 MBCA模块 

分 类 号：S225[农业科学—农业机械化工程]