农机化研究2025，Vol.47Issue(7)：251-256,293,7.DOI:10.13427/j.issn.1003-188X.2025.07.038
基于改进YOLOv5的大白菜快速识别方法研究
A Rapid Recognition Method of Napa Cabbage Based on Improved YOLOv5
摘要
Abstract
The addition of deep learning object detection technology in napa cabbage harvester is an important way to pro-mote the automation of napa cabbage harvest.However,the common deep learning object detection methods rely too much on GPU hardware and are difficult to be applied in ordinary mobile terminals.In order to solve this problem,the model YOLOv5 was improved.Firstly,the backbone part of the original model was replaced with the down-sampling units and the basic units of the ShuffleNetV2 in series.The parameters of the model were greatly reduced to achieve lightweight.Then,the Convolutional Block Attention Modules(CBAM)were embedded in the neck part to detect,compensating for the model accuracy decline after lightweight.The final model was named YSCNet.The experimental results showed that compared with other YOLO series models,the improved lightweight attention mechanism model YSCNet model had the least parameters.Its precision,recall and mean average precision were 97.32%,97.38%,and 98.08%,respectively.The real-time detection test showed that YSCNet performed better on the CPU platform compared to the original model,and its FPS could reach 50.51 fps.关键词
大白菜/目标检测/YOLOv5/轻量化/注意力机制Key words
napa cabbage/object detection/YOLOv5/lightweight/attention mechanism分类
信息技术与安全科学引用本文复制引用
郑文峰,龚中良,刘强,文韬,管金伟,廖舒怀..基于改进YOLOv5的大白菜快速识别方法研究[J].农机化研究,2025,47(7):251-256,293,7.基金项目
湖南省林业杰青培养科研项目(XLK202108) （XLK202108）
