摘要
针对小目标路面病害检测中模型对小病害关注度不足、计算成本过高的问题,研究提出了基于全局注意力机制(Global Attention Mechanism, GAM)与幻影网络V2(GhostNetV2)优化的You Only Look Once Version 8(YOLOv8)模型。该模型通过嵌入GAM模块以增强对小目标特征的提取,并利用GhostNetV2重构骨干网络,实现模型的轻量化。实验结果表明,优化后模型的mAP@0.5达86.4%,较原始YOLOv8提升了8.9%,精确率和召回率分别提升至88.5%和83.2%。同时,模型参数量减少了37.5%,推理速度提升至59 f·s^(-1),有效平衡了检测精度与计算效率,为小目标路面病害检测提供了新的技术支持。
In view of the problems that the model pays insufficient attention to small pavement diseases and the calculation cost is too high in the detection of small target pavement diseases.The study proposes the You Only Look Once Version 8(YOLOv8)model optimized based on the Global Attention Mechanism(GAM)and GhostNetV2.This model embeds the GAM module to enhance the extraction of small target features and uses GhostNetV2 to reconstruct the backbone network,achieving the lightweight of the model.The experimental results show that the mAP@0.5 of the optimized model reaches 86.4%,which is 8.9%higher than that of the original YOLOv8.The precision and recall rates have increased to 88.5%and 83.2%respectively.Meanwhile,the number of model parameters was reduced by 37.5%,and the inference speed was increased to 59 f·s^(-1),effectively balancing detection accuracy and computational efficiency,providing new technical support for the detection of small target pavement diseases.
作者
龙成
LONG Cheng(Xinyu City Highway Enterprise Development Center Branch Appropriate Branch Center,Xinyu 336600,China)
出处
《青海交通科技》
2025年第3期142-146,共5页
Qinghai Transportation Science and Technology
关键词
小目标检测
路面病害检测
注意力机制
模型轻量化
YOLOv8
small target detection
road surface disease detection
attention mechanism
model lightweighting
YOLOv8
