摘要
电动车入户充电存在安全隐患,容易发生火灾,针对高层住户电动车入户充电的问题,本文提出了基于TensorRT加速推理的电梯内电动车检测系统,通过在电梯内部署电动车检测设备来有效遏制高层住户电动车入户的问题。该系统使用YOLOX网络训练目标检测模型,数据类型分为电动摩托车、电动自行车与自行车三类,通过模型迁移将训练好的目标检测模型部署到JetsonNano设备上,该设备通过Jetson-GPIO来做到对电梯的控制。实验结果表明,基于TensorRT加速推理的电梯内电动车检测系统,在性能与识别准确率上均优于传统方法,该设备能够有效遏制高层住户电动车入户问题。
Electric two-wheelers charging at home has potential safety hazards and is prone to fires. In order to solve the problem of charging electric two-wheelers at home by high-rise residents, this paper proposes an electric two-wheelers detection system in elevators based on TensorRT accelerated inference. By deploying electric two-wheelers detection equipment in the elevator, this situation can be effectively contained. The system uses YOLOX to train the target detection model to distinguish electric motorcycles, electric bicycles and bicycles, and then deploys the trained target detection model to JetsonNano devices through model of migration. The device uses Jetson-GPIO to control the elevator. Experimental results show that the electric two-wheelers target detection system in elevators based on TensorRT accelerated inference is superior to traditional methods in performance and accuracy. The equipment can effectively curb the problem of electric two-wheelers entering the homes of high-rise residents.
出处
《计算机科学与应用》
2022年第4期847-857,共11页
Computer Science and Application