基于GADF-GCBAM-ResNet的DAS周界安防信号识别方法OA
DAS perimeter security signal recognition method based on GADF-GCBAM-ResNet
为提升分布式声波传感(DAS)系统采集的振动信号识别精度,提出了一种DAS周界安防信号识别方法.采用格拉姆角差场(GADF)将一维时序振动信号转换为二维图像,经残差网络(ResNet)中的残差块提取特征后,采用分组卷积注意力模块(GCBAM)对其进行分组调整特征权重,以提升ResNet对核心特征及边缘细节的捕获效能.搭建了DAS周界安防实验系统,采集攀爬、拍打、踩踏、跑步及无入侵5种振动信号进行测试.实验结果表明,所提方法的识别准确率达99.3%,相较于卷积神经网络(CNN)、CNN-长短期记忆网络(LSTM)及ResNet等传统模型性能更优.
To improve the recognition accuracy of vibration signals acquired by a distributed acoustic sensing(DAS)system,a DAS perimeter security signal recognition method is proposed.The Gramian angular difference field(GADF)is employed to convert the one-dimensional time-series vibration signals into two-dimensional images.After feature extraction via residual blocks in a residual network(ResNet),a grouped convolution block attention module(GCBAM)is applied to adjust feature weights in groups,thereby enhancing ResNet's capability to capture core features and edge details.A DAS perimeter security experimental system is built,and five types of vibration signals—climbing,tapping,trampling,running,and no intrusion—are collected for testing.Experimental results demonstrate that the proposed method achieves a recognition accuracy of 99.3%,compared with traditional models such as convolutional neural network(CNN),CNN-long short-term memory network(LSTM),and ResNet,the proposed method achieves superior performance.
尚秋峰;王若楠
华北电力大学 电子与通信工程系,河北 保定 071003||华北电力大学 河北省电力物联网技术重点实验室,河北 保定 071003||华北电力大学 保定市光纤传感与光通信技术重点实验室,河北 保定 071003华北电力大学 电子与通信工程系,河北 保定 071003||华北电力大学 河北省电力物联网技术重点实验室,河北 保定 071003||华北电力大学 保定市光纤传感与光通信技术重点实验室,河北 保定 071003
信息技术与安全科学
周界安防分布式声波传感残差网络卷积注意力模块格拉姆角差场
perimeter securitydistributed acoustic sensingresidual networkconvolutional block attention moduleGramian angular difference field
《光通信技术》 2026 (4)
60-64,5
河北省科技计划项目(SZX2020034)资助.
评论