摘要
物联网中不同模态的数据在时间、空间和尺度上存在错位,直接融合会导致维度灾难,进而掩盖异常数据,影响测量效果。为了提升数据质量,及时发现并处理异常,提出基于卷积神经网络的物联网多模态传感数据智能测量方法。采用SVM通过核函数将数据映射至高维空间并构造最优超平面,规避维度灾难,实现数据融合;利用DPC与KNN联合的聚类算法,通过设定局部密度与距离阈值优化异常检测,反映出数据的真实分布情况。并将聚类结果输入卷积神经网络,通过卷积层、池化层和全连接层提取高级特征,有效识别出数据集中的异常值,实现物联网多模态传感数据智能测量。实验结果表明,所提方法在数据智能测量中可以将查准率和查全率的调和平均数保持在85以上,说明可以准确测量物联网多模态传感数据。
Data of different modalities in the Internet of Things(IoT)are misaligned in time,space,and scale,and direct fusion can lead to dimensional disasters,masking abnormal data and affecting measurement effectiveness.To improve data quality,detect and process anomalies in a timely manner,an intelligent measurement method for multimodal sensing data in the IoT based on CNN was proposed.Using SVM to map data to high-dimensional space through kernel functions,the optimal hyperplane was constructed to avoid dimensional disasters and achieve data fusion.Subsequently,the clustering algorithm combining DPC and KNN was used to optimize anomaly detection by setting local density and distance thresholds,reflecting the true distribution of the data.Inputting the clustering results into the convolutional neural network to extract advanced features through convolutional layers,pooling layers,and fully connected layers,and effectively identify outliers in the dataset,intelligent measurement of multimodal sensing data in the IoT was achieved.The experimental results show that the proposed method can maintain a harmonic mean of precision and recall above 85 in intelligent data measurement,indicating that it can accurately measure multimodal sensing data of the Internet of Things.
作者
徐新
XU Xin(Nanjing Tech University Pujiang Institute,Nanjing 211200,China)
出处
《国外电子测量技术》
2025年第4期15-20,共6页
Foreign Electronic Measurement Technology
基金
国家自然科学基金(61702229)
江苏省自然科学基金(18KJD520001)
南京工业大学浦江学院科研重点培育课题(njpj2022-1-07)
南京工业大学浦江学院青年教师发展基金(PJYQ03)。
关键词
物联网
多模态传感数据
数据智能测量
支持向量机
卷积神经网络
internet of things
multi modal sensing data
intelligent measurement of data
support vector machine
convolutional neural network
