重庆科技大学学报（自然科学版）2025，Vol.27Issue(1)：107-114,8.DOI:10.19406/j.issn.2097-4531.2025.01.014
基于改进YOLOv7-tiny的轻量化带钢表面缺陷检测算法
A Lightweight Steel Strip Surface Defect Detection Algorithm Based on the Improved YOLOv7-tiny
摘要
Abstract
To solve the problems of low accuracy,slow speed,and difficult in deploying on embedded devices for surface defect detection of strip steel,a lightweight strip surface defect detection algorithm based on an improved YOLOv7-tiny is proposed.Firstly,the convolution in the backbone feature extraction module ELAN is replaced by the lightweight GhostNetV2 convolution,and the SimAM parameter-free attention module is introduced to reduce the weight and improve the detection speed of the model.Secondly,GD mechanism is introduced to design a new neck multi-scale feature fusion network structure to improve the model's detection capability for small-scale defec-tive targets.Finally,the SiLU activation function and the SIoU bounding box loss function are used to accelerate the convergence efficiency of the model.关键词
缺陷检测/YOLOv7-tiny算法/SimAM模块/聚散机制/轻量化Key words
defect detection/YOLOv7-tiny algorithm/SimAM module/GD mechanism/lightweight分类
信息技术与安全科学引用本文复制引用
彭杰,苏盈盈,杜谦,刘灿,张乐,阎垒..基于改进YOLOv7-tiny的轻量化带钢表面缺陷检测算法[J].重庆科技大学学报（自然科学版）,2025,27(1):107-114,8.基金项目
重庆市自然科学基金面上项目"热轧钢带智能检测中DeLiGAN和迁移学习的改进及应用"(CSTB2022NSCQ-MSX1425) （CSTB2022NSCQ-MSX1425）
重庆市教委科学技术研究项目"面向智能化工厂转型的通用指针式仪表识别方法及实现"(KJQN202101510) （KJQN202101510）
重庆科技大学硕士研究生创新计划项目"G-YOLOv8改进方法及其远距离车辆行人检测中的应用"(YKJCX2320403) （YKJCX2320403）
