长春工程学院学报（自然科学版）2025，Vol.26Issue(3)：110-115,6.DOI:10.3969/j.issn.1009-8984.2025.03.017
YOLOv8-UDS:一种高精度的复杂场景下屋顶缺陷智能检测模型
YOLOv8-UDS:A High-Precision Intelligent Detection Model for Roof Defects in Complex Scenes
摘要
Abstract
In the detection of roof defects,existing detection methods have low accuracy and single detection types.Therefore,an improved YOLOv8 defect detection model YOLOv8-UDS is proposed.The model a-dopts UniRepLknet feature extraction network and variable attention DAttention(DAT)module for fea-ture extraction.The extracted features are efficiently fused through a multi-level feature fusion module(Semantic Detail Infusion,SDI)to improve the detection ability and accuracy of the model.The experimen-tal results show that the accuracy of YOLOv8-UDS reaches 87.9%,which is 10.2%higher than the origi-nal YOLOv8 model and can meet the current demand for roof defect detection.关键词
屋顶缺陷检测/可变注意力/多层次特征融合Key words
roof defect detection/variable attention/multilevel feature integration分类
信息技术与安全科学引用本文复制引用
王朝辉,赵佳..YOLOv8-UDS:一种高精度的复杂场景下屋顶缺陷智能检测模型[J].长春工程学院学报（自然科学版）,2025,26(3):110-115,6.基金项目
吉林省科技发展计划杰出青年人才项目(20240602004RC) （20240602004RC）
