首页|基于改进YOLO v5的机床刀具识别方法

基于改进YOLO v5的机床刀具识别方法

扫码查看
针对目前机床刀具分类应用较少、预处理复杂、目标检测适用范围小且识别精度不高的问题,提出基于改进的YOLO v5机床刀具图像识别算法,利用卷积神经网络在特征提取层加入CBAM注意力模块,可以更清晰地提取图像特征,在特征融合层加入CARAFE上采样模块,使刀具的表面特征恢复更好,可以减少特征融合时部分特征的丢失.实验结果表明,改进后的算法使机床刀具等小目标检测精度和检测速度明显提升,且改进后的模型平均精度为96.8%,比YOLO v4模型提高了 14.96%,比YOLO v5模型提高了 2%.本方法能对不同刀具进行识别,为工业制造中机械零件的识别提供了新的算法支持.
Tool Recognition Method for Machine Tools Based on Improved YOLO v5
The object detection has the problems of small application range and low recognition precision.An im-proved Yolo v5 tool image recognition algorithm is proposed.Based on the idea of convolutional neural network,CBAM at-tention module is added to the feature extraction layer to extract image features more clearly,and the CARAFE sampling module is added to the feature fusion layer.The experimental results show that the improved algorithm can obviously im-prove the detection accuracy and speed of small targets,such as machine tool and so on,the average accuracy of the im-proved model is 96.8%,which is 14.96%higher than that of the YOLO v4 model and 2%higher than that of the YOLO v5 model.The method in this paper can be used to identify different cutting tools and provide a new algorithm support for the identification of mechanical parts in industrial manufacturing.

tool detectionattention mechanismYOLO v5object detectionfeature extraction

闵筱萌、杜文华、段能全、曾志强、刘莞尔

展开 >

中北大学机械工程学院

机床刀具检测 注意力机制 YOLO v5 目标检测 特征提取

国家自然科学基金

52275139

2024

工具技术
成都工具研究所

工具技术

CSTPCD北大核心
影响因子:0.147
ISSN:1000-7008
年,卷(期):2024.58(3)
  • 16