Research On Online Information Tracing Method Based On IoTs

With the development of technology and the popularization of Internet of Things devices,it provides opportunities for large-scale data perception,monitoring environmental,and locating the space and time of events have become an urgent problem to be solved.Information traceability,that is,a technique or method for determining the spatial and temporal position information of an event by analyzing the spatial and temporal distribution of data.Mastering the spatial-temporal distribution characteristics of data has a direct impact on the information traceability results.Therefore,this thesis implements information traceability through field reconstruction technology;At the same time,in order to accurately estimate the characteristics of events in the environment and improve the credibility of the traceability results,a detection and early warning mechanism is needed.This thesis proposes a method of event detection and information tracing based on field reconstruction.Realizing field reconstruction based on spatial-temporal correlation of learning the data,establishing a confidebce interval for event detection based on the output result of field reconstruction;and conduct a grid search in space according to the detection result to find the location of the event;restore the event time-series data to location the occurrence time.The main research contents of this article are as follows:1.A spatio-temporal field reconstruction method based on Gaussian process is proposed.The basic principles of the Gaussian process are studied,the basic properties of common kernel functions are analyzed through simulation experiments,the method of model selection is summarized,and the hyperparameters of the Gaussian process in spatio-temporal field reconstruction are difficult to estimate.Due to Spectral Mixture kernel can realize automatic pattern discovery and extrapolation,as well as learning of complex changing pattern,and avoid the problem of hyperparameter estimation.Therefore,experiments on simulated spatial-temporal data using this kernel verify the performance of this method in spatio-temporal field reconstruction.2.An event detection method based on field reconstruction is proposed.The output of the Gaussian process has probabilistic significance.In this thesis,a 95% confidence interval is constructed based on the predicted value and standard deviation of the Gaussian process in the reconstruction of the spatio-temporal field.At the same time,the spatial field reconstruction is performed on the data at the current time.Finally,according to the measurement results of the confidence interval on the spatial field,event detection is implemented.3.This thesis uses field reconstruction technology to realize the backtracking of the space and time position of the event source.The Gaussian process can not only restore the original spatial appearance of the data,but also make estimates and predictions for the data at various spatio-temporal points.The spatial position of the event source in the environment is estimated through the space field;at the same time,the time series data of the estimated position is restored to realize the identification of the event occurrence time.4.Finally,this thesis uses the temperature data of Intel Berkeley Lab to test the method proposed in this thesis.Gaussian process was used to realize temperature field prediction,event detection and information tracing.The purpose of this thesis is to trace back the spatio-temporal position of the event.