基于机器视觉的轻量化钢材表面缺陷检测模型

刘文钊; 张丹

基于机器视觉的轻量化钢材表面缺陷检测模型

引用本文：刘文钊，张丹.基于机器视觉的轻量化钢材表面缺陷检测模型[J].计算技术与自动化,2024,(3):43-49

摘要点击次数: 416

全文下载次数: 0

作者	单位
刘文钊，张丹	（南京林业大学信息科学技术学院，江苏南京 210037）

中文摘要:针对当前主流缺陷检测模型参数量大、计算复杂度高，难以在计算资源有限的嵌入式设备上部署的问题，提出了一种轻量化钢材表面缺陷检测模型YOLO-LSNet。首先，为了降低模型的参数量和计算复杂度，提出了一种轻量化卷积模块MSConv。其次，提出M-BiFPN网络用于深浅层特征信息的融合。最后，用SIoU损失函数替换CIoU损失函数,加快网络的收敛速度。实验结果表明，YOLO-LSNet模型在NEU-DET 数据集上相较于基线网络YOLOv5，mAP提升了1.8%，模型参数下降了43.4%，计算量降低了36.1%。完成模型轻量化设计的同时，保证了模型的检测精度，具有良好的应用前景。

中文关键词:缺陷检测深度可分离卷积多尺度特征融合轻量化注意力机制

Lightweight Steel Surface Defect Detection Model Based on Machine Vision

Abstract:Aiming at the problems of large number of parameters and high computational complexity of current mainstream defect detection models, it is difficult to deploy on embedded devices with limited computing resources. A lightweight steel surface defect detection model YOLO-LSNet is proposed. Firstly, in order to reduce the parameter number and computational complexity of the model, a lightweight convolution module MSConv is proposed. Secondly, M-BiFPN network is proposed for deep and shallow layer feature information fusion. Finally, CIoU loss function is replaced by SIoU loss function to speed up the convergence of the network. The experimental results show that compared with the baseline network YOLOv5 in NEU-DET data set, the mAP of YOLO-LSNet model increases by 1.8%, model parameters decreases by 43.4%, and computation cost decreases by 36.1%. At the same time, the lightweight design of the model is completed, the detection accuracy of the model is guaranteed, and it has a good application prospect.

keywords:defect detection depthwise separable convolution multi-scale feature fusion lightweight attention mechanism

查看全文 查看/发表评论 下载pdf阅读器