Research And Application Of Information Fusion Algorithm

With the increasingly widespread application of wireless sensor networks and the generationand application of the Internet of Things, information fusion technology has been graduallyintegrated into the social life, which brings unprecedented convenience to people’s lives. With themore and more complicated of sensor data sources, how to obtain information comprehensively,rapidly and accurately has becomes a research focus. Therefore, the research of information fusionmethod has significant meaning.In this thesis, the monitoring data of multi-sensor is basic research object. By using the timeand spatial redundant information of monitoring data, a data-level information fusion model basedon time-dimension and a feature-level information fusion model based on spatial-dimension arebuilt, and the corresponding information fusion methods is proposed and applied to forest firemonitoring system. The research contents are as follows:In order to make a better use of scene context information in the monitoring data, aninformation fusion model based on sliding window is proposed. The basic ideal of this model isestimating the change rule of attribute by using context information of continuous data sequence inwindow; using the estimate result to descript the change rule of the attribute, and associating andfusing multi-sensors data by using the redundant information between different windows of oneattribute as well as that between the different attributes of the same window.For data-level information fusion, a Data-Level Information Fusion Algorithm based onTime-Dimension (DIFAT) is proposed. As the monitored object has its own change rules, afteranalyzing the time domain and frequency domain of data sequence in window, a model used todescript the change of monitored object is established. According to the model, optimizationmethod gives the optimal estimation of rule of the monitored object. Matlab simulation tool is usedto verify the validity of the algorithm, and the results show that the algorithm can effectivelyremove part of the noise, reduce uncertainty of perceptual data, and obtain reliable change rules ofattributes.For feature-level information fusion, considering the multi-sensor data fusion of both oneattribute and different attributes, a Feature-Level Information Fusion Algorithm based onSpatial-Dimension (FIFAS) is proposed. The FIFAS utilizes methods, such as data association and neural network, to determine the mapping relations between the monitored attributes. Theserelations are used to fuse multi-sensor data and determine the states of monitored attributes. Matlabsimulation tool is used to verify the validity of the algorithm, the results show that the algorithm candescript the real state of attribute very well, and make a reasonable estimate of the originalfrequency distribution of the attribute data.To further verify the practicality of the proposed algorithm, a forest fire monitoring system isdesigned. DIFAT and FIFAS algorithm are used to give forest fire warning level.