基于多模型协同优化的烟叶非法运输车辆检测与识别研究

打开文本图片集

中图分类号：U463.6 文献标识码：A 文章编号：1003-8639（2025）11-0061-0

ResearchonDetectionandRecognitionofIlegal Tobacco TransportVehiclesBased onMulti-ModelCollaborativeOptimization*

Fu Zeyu, Zhao Guihong, Jiang Feng,Dai Linqing,Han Xiaomei, Yang Peidong (Yunnan Tobacco Company Qujing Branch, Kunming 655ooo, China)

【Abstract】In response to the problems of insuficient detection accuracy in complex scenarios,poor real-time performanceinresourceconstrainedenvironments,andweakvehicletraceabilitycapabilities inthesupervisionofilegal tobacco transportation,this studyconstructs a"detectionrecognition deployment"multi-modelcollborativeoptimization system: improving RT-DETR (RT-DETR-R18-P2) to solve high-precision detection in complex scenarios,improving ResNet34 (ResNet34-SE) to achieve vehicle attribute traceability,and adapting YOLO series models (YOLOv5 /v& 3）to resourceconstrained deployment,achievingcomplementaryadvantages through "functionaldivision-targetcollborationdynamic scheduling".Verifiedby 358O vehicle images from the tobacco checkpoint in Qujing,Yunnan,the mAP50 of the collaborative system in complex scenes reached 92.8% ,the inference speed of edge deviceswas15O FPS,and the accuracy of attribute recognition was 93.8% . This formed an intelligent supervision scheme with "full scene coverage and fullresourceadaptation",providing technical supportforthe prevention and controlofillegal tobacco transportation.

【Key words】illegal transportation of tobacco leaves;multi-model collaboration;vehicle detection；RT

DETR; YOLO; SE attention mechanism 0 引言

烟叶运输监管需满足复杂场景高精度检测、边缘卡口实时响应和违法车辆精准溯源三重需求，传统单一模型难以兼顾。（剩余5583字）