基于边缘计算的在途危险品姿态识别方法  

Edge computing based attitude recognition of dangerous goods in transit

作　　者：张伟 邓士杰 于贵波 ZHANG Wei;DENG Shijie;YU Guibo(Shijiazhuang Campus,Army Engineering University of PLA,Shijiazhuang 050003,China)

机构地区：[1]陆军工程大学石家庄校区,河北石家庄050003

出　　处：《现代电子技术》2025年第7期155-162,共8页Modern Electronics Technique

摘　　要：针对化学制品、火工品等危险品的在途安全状态监控问题,文中提出一种基于边缘计算的在途危险品姿态识别方法,旨在实时识别危险品在运输过程中的行为姿态。该方法采用边缘计算设备作为数据处理平台,首先通过姿态传感器实时抓取危险品在运输过程的三轴运动数据;然后综合滑动窗口技术与特征提取完成对数据流的处理,获得在途危险品行为姿态样本数据;最后,使用基于深度学习的CNN⁃BiLSTM⁃Attention分类模型完成对危险品行为姿态的识别。实验结果证明,该方法得益于边缘计算与深度学习的联合优势,能够准确可靠地识别在途危险品的行为姿态,具有一定的实际应用价值。Aiming at the monitoring of the safety status of the in⁃transit dangerous goods,for example,chemicals and pyrotechnic products,an in⁃transit dangerous goods attitude recognition method based on edge calculating is proposed to recognize the behavioral attitude of these goods in real time.In the method,the edge computing devices are adopted as the data processing platform.Firstly,the three⁃axis motion data of dangerous goods in the process of transportation is captured by the attitude sensor in real time;then,the sliding window technology and feature extraction are integrated to complete the processing of the data flow,and the behavioral attitude sample data of dangerous goods in transit are obtained;finally,the classification model based on the deep learning CNN⁃BiLSTM⁃Attention is used to complete the recognition of behavioral attitude of dangerous goods.The experimental results show that the method can accurately and reliably recognize the behavioral postures of in⁃transit dangerous goods thanks to the joint advantages of edge computing and deep learning,so it has a certain practical application value.

关 键 词：边缘计算 在途危险品 姿态传感器 滑动窗口技术 特征提取 深度学习 

分 类 号：TN911.7-34[电子电信—通信与信息系统] TP393[电子电信—信息与通信工程]