The Research On Transboundary Disturbance Based On Fiber Sensing Stream Data

In recent years,fiber bragg grating array sensing technology has been gradually applied to security system,but the grating array is accompanied by noise,large capacity and difficult to obtain signal information,which makes the monitoring of security system greatly troubled.Traditional supervised clustering algorithms,such as the traditional density-based spatial clustering of applications with noise(DBSCAN),Kmeans,Fuzzy C-means(FCM)rely on similarity to find outliers,nevertheless,this clustering process requires human intervention,which takes lots of time and efforts,and cannot even achieve the expected timeliness in large amounts of data.Therefore,traditional clustering algorithms are increasingly unable to meet the design requirements of transboundary disturbance systems.In order to solve these problems,this paper proposes a grating array hybrid denoising,feature extraction and variance selection strategy based on signal eigenvalues,and defines an adaptive parameter DBSCAN(AP-DBSCAN)method,the stream computing model is constructed,and the improved algorithm is combined with the stream computing model.A new unsupervised streaming-AP-DBSCAN(SAP-DBSCAN)clustering algorithm is proposed,which sloves the problems of data stream collection,real-time data analysis and identification of different types of human disturbances.The main work of this thesis includes:(1)Aiming at the problem of interference caused by multi-noise and high transmission frequency grating arrays.Firstly,a hybrid denoising scheme of grating array is proposed,and the features of grating array are extracted simultaneously.Then,screening important features based on variance of multidimensional features of signals.Finally,the accuracy of the proposed pre-processing strategy is verified by the experiments.(2)Aiming at the problem that the traditional DBSCAN algorithm can't adapt parameters,firstly,the unsupervised clustering algorithm is studied.By improving the sensitivity of the input parameters of the DBSCAN algorithm,excessive human intervention is avoided.According to the pre-processed data eigenvalues,the distance matrix is constructed to find the maximum value in the optimal column vector to obtain the neighborhood radius Eps.Then,according to the determined Eps,the parameter estimation method of Poisson distribution for the data in the Eps neighborhood of each point is obtained by the minimum neighborhood number MinPt.Finally,the APDBSCAN algorithm based on the unsupervised adaptive input parameter is proposed.In order to make the proposed algorithm adapt to the grating array stream data to achieve real-time and accuracy,a stream computing model was established,a streaming unsupervised algorithm SAP-DBSCAN is proposed.The experimental results under real stream data confirm that the algorithm has better accuracy and timeliness.(3)The proposed unsupervised SAP-DBSCAN clustering algorithm is applied to the transboundary disturbance system,and the real-time cross-border disturbance intrusion detection was performed based on grating array.The time-effectiveness of the algorithm is verified by experiments.