Clustering of estimated spatial locations in networked sensors
Ivan P. Pokrajac, Predrag Okiljević, Miodrag S. Vračar · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2013
Multisensor data fusion combines data from multiple sensors to overcome interferences that may not be possible from a single sensor or source alone. In military application data fusion can be used to integrate the individual sensor data into common operational picture of the battlefield. However, there is still possibility to improve quality of the individual sensor. Improving of accuracy in estimation of spatial location is investigated in this paper. Some novel methods and algorithms for estimation of spatial location are compared such as Discrete Probability Density (DPD) method, fusion of multiple bearing lines and mean-square distance algorithm. These methods for estimation of spatial location use two-step positioning technique (indirect technique) based on estimation of a specified parameter such as angle of arrival (AOA). In the network where is possible to provide multiple spatial locations from the spatially close sources, clustering of estimated spatial location is very important. The estimated spatial locations that correspond to a source are spatially close to one another will have a larger likelihood than those estimated spatial locations that are not correspond to the source. In this paper methods and algorithms for estimation of spatial location are compared where it is multiple spatial locations, for the same sources spatially close. Clustering has been performed based on estimated spatial locations and appropriates the covariance matrix.