Research And Implementation Of Multi-Source Data Fusion In IoT Terminal

With the rapid development of the Internet of Things,many applications require accurate and reliable sensor data as support.However,due to external interference and the low accuracy of the sensor itself,the collected data may have the problem of inaccurate measurement.In Io T applications,a large number of sensor nodes are usually deployed for data collection of various parameters,and the acquired perception data has a strong correlation.It can be combined with the correlation of perception data in time,space,and physical attributes to reduce measurement errors.This thesis deeply studies the correlation characteristics between multi-source perception data,and proposes a multi-source data fusion method for Io T terminals.Mainly research the following aspects:1.Analyze the characteristics of the multi-source perception data of the Internet of Things,and study the relevance of homogeneous and heterogeneous data in time,space and attributes.Based on these analyses,a multi-source data fusion scheme for Io T terminals was proposed.2.Combining the spatial-temporal correlation characteristics of isomorphic perception data,a method of isomorphic data fusion based on spatial-temporal correlation is proposed.The method first uses the dynamic time warping distance to calculate the spatial distance of each terminal data at each time point to quantify the temporal and spatial correlation between the data,and then sets a threshold to exclude low correlation data,and finally weights are allocated according to the spatial-temporal correlation to complete the weighted data fusion.3.Combining the attribute correlation characteristics of heterogeneous perception data,a heterogeneous data fusion method based on attribute correlation is proposed.When the temporal and spatial correlation of the terminal's perceived data value is weak,the physical attribute correlation between heterogeneous data can be used for fusion.This method uses the BP neural network with less resource consumption to perform heterogeneous data fusion.First,the gray correlation method is used to analyze the heterogeneous data,and then the heterogeneous data set with attribute correlation is used as input,and the heterogeneous data fusion model training is performed.The trained model is put into the Io T terminal to run to achieve heterogeneous data fusion.In summary,this thesis combines the correlation characteristics of the perception data to study the multi-source data fusion method and build a test platform to test and verify the method on the Io T terminal.The results show that the proposed method can effectively reduce the data measurement error.