改进YOLOv5的棉田杂草检测  被引量:1

Improved YOLOv5 detection of weeds in cotton fields

在线阅读下载全文

作  者:杨明轩 陈琳[1] YANG Mingxuan;CHEN Lin(College of Computer Science,Yangtze University,Jingzhou 434000,China)

机构地区:[1]长江大学计算机科学学院,湖北荆州434000

出  处:《现代电子技术》2024年第24期60-67,共8页Modern Electronics Technique

基  金:国家自然科学基金项目(62006028);湖北省自然科学基金项目(2022CFB132);湖北省教育厅自然科学研究计划项目(B2022038)。

摘  要:针对复杂环境下棉田杂草检测与识别困难等问题,提出一种改进YOLOv5的棉田杂草检测算法——CSTYOLOv5。首先,通过数据增强算法解决棉田杂草样本分布不均匀导致的模型训练效果不充分问题;其次,考虑到通道信息和方向位置信息,在主干网络中加入了坐标注意力机制;最后,在颈部网络中将Swin Transformer Block引入C3模块,得到新的C3STR模块,以保留全局上下文信息和多尺度特征。实验结果表明,CST-YOLOv5模型的mAP值达到95.1%,F1值达到90.4%,比原YOLOv5模型提高了4.8%、3.2%。所设计算法具有良好的鲁棒性,能精确识别多类杂草。In allusion to the difficulty of detecting and identifying weeds in cotton fields in complex environments,a cotton field weed detection algorithm CST-YOLOv5 is proposed to improve YOLOv5.The data enhancement algorithm is used to solve the problem of insufficient model training effect due to the unbalanced distribution of weed samples in cotton fields.A coordinate attention mechanism is added to the backbone network by considering channel information and direction location information.The Swin Transformer Block is introduced into the C3 module in the neck network to obtain a new C3STR module to preserve global context information and multi-scale features.The experimental results show that the mAP value of the CST-YOLOv5 model can reach 95.1%,and the F1 value can reach 90.4%,which are respectively increased by 4.8%and 3.2%compared with the original YOLOv5 model.It verifies that the designed algorithm has good robustness and can accurately identify many types of weeds.

关 键 词:杂草检测 YOLOv5 深度学习 目标检测 注意力机制 棉花保护 

分 类 号:TN911.23-34[电子电信—通信与信息系统] TP391[电子电信—信息与通信工程]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象