摘要
针对当前农作物种子分拣过程中,人工分拣效率低下及现有机器视觉识别系统存在的检测速度慢、识别准确率低等问题,提出了一种基于改进YOLOv8n的种子目标检测方法。首先在C2f中用轻量化网络FasterNet替换Bottleneck,访问提升网络运行速度与能力,减少冗余计算和内存。其次,引入EMA注意力机制,通过并行子网结构和跨空间信息聚合更好地关注多尺度特征,提高种子识别精度。最后,采用Wise-IOU-v3损失函数,在减少低质量标注影响的同时,也更好地提升了网络收敛速度。实验结果显示,与标准的YOLOv8n算法相比,改进的YOLOv8n算法精确率、召回率、mAP(0.5)和mAP(0.50.95)分别提高了2.2%、3.4%、1.4%、4.4%,FLOPs减少了20.2%,参数量减少了27.2%。改进的YOLOv8n在精度和速度权衡方面展现出明显优势,可为农业自动化分拣设备提供更可靠的技术支持。
Aiming at the problems of low manual sorting efficiency and slow detection speed and low recognition accuracy of existing machine vision recognition systems in the current process crop seed sorting,a seed target detection method based on improved YOLOv8n was proposed.First,in C2f,the lightweight network FasterNet was used replace Bottleneck,which reduces redundant calculations and memory access,and improves the network operation speed and capability.Second,the EMA attention mechanism was introduced,which uses parallel sub structure and cross-space information aggregation to focus better on multi-scale features,and improve the accuracy of seed recognition.Finally,the Wise-IOU-v3 loss function was used,which reduces the impact of low-quality annotations and accelerates the convergence speed of the network.The experimental results show that compared with the standard YOLOv8 algorithm,the improved YOLOv8n algorithm has improved by 2.2%,3.4%,1.4%,and 4.4 in precision,recall,m AP(0.5),and m AP(0.50.95),and the FLOPs are reduced by 20.2%,and the number of parameters is reduced by 27.2%.The improved YOLOv8n shows significant advantages in the trade-off precision and speed,and it can provide more reliable technical support for agricultural automatic sorting equipment.
作者
王影
梁秋阳
常广良
刘麒
WANG Ying;LIANG Qiuyang;CHANG Guangliang;LIU Qi(School of Information and Control Engineering,Jilin University of Chemical Technology,Jilin City 132022,China;Kesshida(Changchun)Automobile Electrical Appliance Co.,LTD.,Changchun 130000,China)
出处
《吉林化工学院学报》
2025年第7期17-23,共7页
Journal of Jilin Institute of Chemical Technology
基金
吉林市科技成果项目(20190502118)
吉林化工大学科研项目(2018064)
吉林化工大学重大科技项目(2018017)。
