首页|期刊导航|信息与控制|基于YOLOv9的露天矿电铲的轻量化实时斗齿状态检测方法

基于YOLOv9的露天矿电铲的轻量化实时斗齿状态检测方法OA

Lightweight Real-time Detection Method for Bucket Teeth Status of Electric Shovels in Open-pit Mines Based on YOLOv9

中文摘要英文摘要

电铲作为露天矿开采的核心装备,其斗齿在高强度作业过程中极易发生齿套脱落、断裂等故障.故障部件一旦混入矿石流,将引发破碎设备卡堵、损毁及皮带撕裂等严重事故.针对上述问题,以 YOLOv9-s(you only look once v9-s)模型为基础,提出了轻量化高精度检测 FDP-YOLO 模型,以实现动态采矿环境中各种斗齿故障状态的自动实时识别.通过引入 FasterNet 主干网络,对颈部网络中的 AConv模块进行改进,引入 DualConv(dual convolution)构建 ADualConv,并且采用 PConv(partial convolution)重新轻量化 P-RepNCSPELAN4,保证模型特征提取能力的同时,减少了参数数量和浮点运算量.结果表明,FDP-YOLO模型权重减小 36.2%,参数量降低 36.2%,浮点运算数减少 48.1%,其平均精度均值 mAP@50 达到 99%.该模型能够在保证检测精度的前提下,大幅提升检测速度,有效满足露天矿电铲斗齿状态的实时精准检测需求,为露天矿智能化开采提供了可靠的技术支持,具有显著的工程应用价值和推广潜力.

As the core equipment in open-pit mining production,electric shovels are prone to problems such as gear sleeve detachment and breakage during operation.When the fallen parts mix into the ore,it can easily cause accidents such as crushing equipment jamming,damage,or belt tearing.To solve these problems,we propose a lightweight and high-precision model based on the improved YOLOv9-s model.By introducing the FasterNet backbone network,improving the AConv module in the neck network with dual convolution(DualConv),and re-lightweighting P-RepNCSPELAN4 with partial convolution(PConv),the proposed model's feature extraction ability is ensured while reducing the number of parameters and floating-point operations.The results show that FDP-YOLO model reduces the weight by 36.2%,the number of parameters by 36.2%,the floating-point operation GFLOPs by 48.1%,and mAP@50 reaches 99%.The status detection method can satisfy the intelligent application need of open-pit mine electric shovel tooth detection.

刘晓松;荆洪迪;李歆;潘福成;于庆磊;张兴帆

中国科学院沈阳自动化研究所,辽宁 沈阳 110016中国科学院沈阳自动化研究所,辽宁 沈阳 110016中国科学院沈阳自动化研究所,辽宁 沈阳 110016中国科学院沈阳自动化研究所,辽宁 沈阳 110016东北大学资源与土木工程学院,辽宁 沈阳 110819中国科学院沈阳自动化研究所,辽宁 沈阳 110016

矿业与冶金

斗齿目标检测轻量化FasterNetDualConvPConv

bucket toothobject detectionlightweightFasterNetDualConvPConv

《信息与控制》 2026 (4)

652-664,13

国家自然科学基金项目(U21A20106)

10.13976/j.cnki.xk.2025.0162

评论