场景识别中基于改进YOLOv8的路况识别与分类  

Road condition recognition and classification based on improved YOLOv8 in scene recognition

作　　者：刘晓 孙皓月[1] 张碧宁 王俊博 LIU Xiao;SUN Haoyue;ZHANG Bining;WANG Junbo(HeBei University of Architecture,Zhangjiakou,Hebei 075024,China)

机构地区：[1]河北建筑工程学院,河北张家口075024

出　　处：《计算机应用文摘》2024年第14期143-145,共3页Chinese Journal of Computer Application

基　　金：基于深度学习网络模型的交通网络识别与分类(XY2024039)。

摘　　要：针对路况场景中交通量密集、目标小、背景复杂等因素导致的目标检测精度低、易漏检误检问题,提出了一种基于改进的YOLOv8路况识别分类方法。首先,在骨干网络中引入Ghost模块以加快检测速度;其次,在颈部网络中融合注意力机制CBAM,增强对目标的关注,从而解决漏检和误检问题。实验结果显示,该算法的FPS提高了3.9个百分点,mAP提高了7.3%。相较于同类方法,该方法在速度和精确度之间取得了良好的平衡,并实现了显著的改进。A road condition recognition and classification method based on improved YOLOv8 is proposed to address the issues of low target detection accuracy and susceptibility to missed detections and false detections caused by factors such as dense traffic volume,small targets,and complex backgrounds in road conditions scenarios.Firstly,introduce the Ghost module into the backbone network to accelerate detection speed.Secondly,integrating the attention mechanism CBAM in the neck network enhances the focus on the target,thereby solving the problems of missed and false detections.The experimental results show that the FPS of the algorithm has increased by 3.9 percentage points,and the mAP has increased by 7.3%.Compared to similar methods,this method achieves a good balance between speed and accuracy,and has achieved significant improvements.

关 键 词：YOLOv8 GhostNet 注意力机制 路况识别 

分 类 号：TP389[自动化与计算机技术—计算机系统结构]