融合EMA注意力机制的轻量级玉米雄穗检测  

Lightweight Maize Tassel Detection Incorporating EMA Attention Mechanism

作　　者：吴伟 陈伟 孙容 刘路 Wu Wei;Chen Wei;Sun Rong;Liu Lu(School of Engineering,Anhui Agricultural University,Hefei,Anhui 230036,China;Anhui Provincial Engineering Research Center of Intelligent Agricultural Machinery,Hefei,Anhui 230036,China)

机构地区：[1]安徽农业大学工学院,安徽合肥230036 [2]安徽省智能农机装备工程实验室,安徽合肥230036

出　　处：《黑龙江工业学院学报（综合版）》2024年第9期93-98,共6页Journal of Heilongjiang University of Technology(Comprehensive Edition)

基　　金：安徽省重点实验室建设项目“玉米中后期基于表型探测的精准施药关键技术研究”(项目编号:MAET202102)。

摘　　要：玉米雄穗对于玉米产量以及优良育种有着重要作用,为提升复杂田间环境下对抽雄期玉米雄穗的检测精度和速度,提出一种融合高效多尺度注意力机制的轻量级目标检测模型YOLOv5-EMM。通过嵌入高效多尺度注意力机制(EMA)模块、替换MobileNetV3轻量级骨干网络、优化MPDIoU损失函数及采用BiFPN优化颈部网络改进YOLOv5基础模型。试验结果表明,YOLOv5-EMM网络平均精确度达到97.8%,比改进前提高了2.7%,模型参数量低至3.31M,计算量也降低了51%。改进模型对复杂田间环境玉米雄穗有着较好的检测效果,可为玉米生长状态监测提供技术支持。Maize tassels play a crucial role in maize yield and excellent breeding.To improve the detection accuracy and detection speed of corn tassels during the tasseling period in complex field environments,this paper proposes a lightweight object detection model YOLOv5-EMM that integrates efficient multi-scale attention mechanism.By embedding an efficient multi-scale attention mechanism(EMA)module,replacing the MobileNetV3 lightweight backbone network,optimizing the MPDIoU loss function,and using BiFPN to optimize the neck network,the YOLOv5 model is improved.The experimental results show that the YOLOv5 EMM network has an average accuracy of 97.8%,which is 2.7%higher than before improvement.The number of model parameters is as low as 3.31M,and the amount of computation is also reduced by 51%.The YOLOv5 EMM model has a good detection effect on maize tassels in complex field environments,and can provide technical support for monitoring maize growth.

关 键 词：目标检测 深度学习 玉米雄穗 复杂田间环境 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]