摘要
玉米是我国三大主粮作物之一,其表型参数的获取在作物遗传解析和新品种选育等研究领域意义重大。目前,玉米考种主要依靠肉眼和传统机器视觉方法完成,为提高玉米籽粒识别的速度和精度,建立了基于YOLOv5的籽粒目标检测模型,用不同环境下采集的数据训练并选择最优模型。试验结果表明,YOLOv5s的时间复杂度最低,其精确率、召回率和mAP@0.5的均值分别达到90.4%,85.9%和91.4%,实现了对密集黏连目标较理想的检测效果。
Currently,kernel detection relies mainly on the naked eye and traditional machine vision,in order to improve the speed and accuracy of kernel detection,this paper established a target detection model based on YOLOv5,trains and selects the optimal model with data collected in different environments.The experimental results showed that YOLOv5s has the lowest time complexity,with an average precision,recall and mAP@0.5 of 90.4%,85.9%and 91.4%,which achieved a ideal detection effect for dense adhe-sion targets.
作者
李毅
谢一民
樊智慧
闫伟茉
Li Yi;Xie Yimin;Fan Zhihui;Yan Weimo(Northeast Agricultural University,Harbin 150000,Heilongjiang,China)
出处
《农业技术与装备》
2024年第11期6-8,12,共4页
Agricultural Technology & Equipment
基金
东北农业大学省级大学生创新训练项目(S202310224145)。
