重庆理工大学学报2026，Vol.40Issue(13)：11-20,10.DOI:10.3969/j.issn.1674-8425(z).2026.07.002
基于深度学习的精细化车型分类与速度检测研究
Deep learning based research on refined vehicle model classification and speed detection
摘要
Abstract
Traditional vehicle speed extraction methods suffer low accuracy.This paper employs object detection and tracking algorithms based on deep learning to realize real-time vehicle speed detection.According to national and industrial vehicle classification standards,the correlation of various classification methods was reviewed.With vehicle model image feature information,a refined vehicle classification system was developed.The YOLOv10 network framework was selected and the CA attention mechanism was introduced.Wise-IOU v3 was employed to optimize bounding box regression and add a P2 detection layer.CAW-YOLO vehicle detection model was proposed to enhance detection accuracy for similar vehicle models.Image distortion was corrected and a PTP speed detection model was developed by ByteTrack object tracking algorithm with perspective transformation.Results indicate that,CAW-YOLO model improves mAP0.5 by 3.8％compared to that of YOLOv10 and the relative error value for speed detection in the PTP model is under 9％.Four vehicle classification systems are assessed and more accurate speed data are obtained by a refined vehicle classification system.关键词
交通感知/目标检测/目标跟踪/车型分类/速度检测Key words
traffic perception/target detection/target tracking/vehicle classification/speed detection分类
信息技术与安全科学引用本文复制引用
徐慧智,李雁平..基于深度学习的精细化车型分类与速度检测研究[J].重庆理工大学学报,2026,40(13):11-20,10.基金项目
黑龙江省自然科学基金项目(PL2025E012) （PL2025E012）
