机电工程技术2025，Vol.54Issue(14)：46-50,134,6.DOI:10.3969/j.issn.1009-9492.2025.14.008
基于改进YOLOv8-seg的螺丝及孔分割方法
Screw and Hole Segmentation Method Based on Improved YOLOv8-seg
摘要
Abstract
A method for screw and hole instance segmentation based on YOLOv8-seg is proposed to address the challenges of low positioning accuracy and assembly efficiency in the screw and hole assembly process within the manufacturing industry.Firstly,a multi-level feature fusion module(SDI)is introduced,and features from different levels are fused to enhance the semantic and detail information of screws and holes in the images.Secondly,a diverse branch block(DBB)is incorporated to replace standard convolution operations.By adding multiple parallel branches and reconfiguring parameters,multi-scale feature information is enhanced.Finally,the CIoU loss function is replaced with the DIoU loss function to reduce the distance between the predicted and ground truth box centers,better locate the target's position and edge features,and improve the accuracy of screw and hole segmentation and localization.Ablation and comparative experiments are conducted based on the WEEE screw disassembly dataset and a self-constructed screw-hole assembly dataset.The experimental results demonstrate that the YOLOv8-SBD-seg algorithm achieves a 4.59%improvement in precision compared to the traditional YOLOv8-seg method.Therefore,the proposed method significantly improves the positioning accuracy and assembly efficiency of screws and holes in assembly scenarios,demonstrating broad application prospects in the field of industrial assembly.关键词
螺丝及孔/实例分割/YOLOv8-seg/损失函数Key words
screws and holes/instance segmentation/YOLOv8-seg/loss function分类
机械制造引用本文复制引用
刘佳月,陆恒宇,顾兆旭,李绪臣,钱嘉力,巢渊..基于改进YOLOv8-seg的螺丝及孔分割方法[J].机电工程技术,2025,54(14):46-50,134,6.基金项目
国家自然科学基金(51905235) （51905235）
江苏省大学生创新创业训练计划项目(202411463001Z) （202411463001Z）
江苏理工学院研究生实践创新计划项目(XSJCX24_26) （XSJCX24_26）
