The sintering trolley is a key component of the sintering machine.How to automatically re-fuel and lubricate the wheels of the sintering trolley to ensure its stable operation is one of the main factors to ensure the operation rate of the sintering machine and the sinter output.The difficulty of au-tomatic refueling and lubrication is to obtain the position information of the wheels in real time.Therefore,a trolley wheel detection model based on MobileNetV2-SSD for the real-time positioning of trolley wheels was proposed.Considering the harsh working environment of the sintering site,it is im-possible to equip a conventional computer to run the model.In order to make the model run on a mo-bile terminal or embedded device with low computational power,the detection model proposed in this paper selects the lightweight MobileNetV2 on the backbone network,and the overall detection archi-tecture uses SSD,the detection speed is further improved while taking into account the detection ac-curacy.Finally,the precision,recall and mAP value of the model have reached more than 90％on the actual data set collected,and the detection rate of the GTX1060 or above graphics card has reached more than 40 fps,meeting the industrial real-time detection standards.The experimental re-sults prove the effectiveness of the model.
NETL
NSTL
万方数据
维普
