现代电子技术2024，Vol.47Issue(20)：141-147,7.DOI:10.16652/j.issn.1004-373x.2024.20.022
改进YOLOv8算法的交通标志小目标检测
Traffic sign small target detection based on improved YOLOv8 algorithm
摘要
Abstract
In allusion to the problem of inaccurate small target recognition in traditional network models for traffic sign detection,an improved Ghost-YOLOv8 traffic sign detection model is proposed.GhostConv is used to replace all Conv,and a novel GhostC2f module is used to replace all C2f,so as to make the model lightweight.The upsampling method is substituted with CARAFE to better retain the intricate details in the image.The GAM attention mechanism module is introduced in the Neck section to enhance the semantic information and location information in the features.In order address the issue of losing semantic information caused by scale inconsistency during the detection of small targets,a small target detection layer is incorporated to enhance the fusion of deep and shallow semantic information.The experimental results show that the model can enhance the Recall,mAP@0.5,and mAP@0.5:0.95 measures in the Chinese traffic sign identification dataset TT100K by 6.8%,4.0%,and 3.6%respectively.It decreases the number of model parameters and the model size by 1.069×106 and 1.9 MB respectively.It is shown that the proposed model can reduce the number and size of parameters and can better detect small targets that cannot be detected by the original model under the premise of constant accuracy.In comparison with the comparison algorithm,it has better performance and is suitable for edge computing devices,and has practical application value.关键词
YOLOv8/交通标志/小目标检测/GhostNet/CARAFE/GhostC2f/GAM注意力机制Key words
YOLOv8/traffic sign/small target detection/GhostNet/CARAFE/GhostC2f/GAM attention mechanism分类
信息技术与安全科学引用本文复制引用
赵会鹏,曹景胜,潘迪敬,李刚..改进YOLOv8算法的交通标志小目标检测[J].现代电子技术,2024,47(20):141-147,7.基金项目
国家自然科学基金项目(51675257) （51675257）
国家自然科学基金青年基金项目(51305190) （51305190）
辽宁省教育厅基本科研项目(面上项目)(LJKMZ20220976) （面上项目）
辽宁省自然科学基金指导计划项目(20180550020) （20180550020）
