Ship Targets Association Based On Satellite Imaging And Electronic Information

Target association is an important guarantee for information fusion and target tracking,and it can judge whether the information from different sensors is derived from the same target.Target association is important to estimate the trend of the marine battlefield and to obtain the active power of the battlefield.This paper is based on the background of marine surveillance of the electromagnetic reconnaissance satellite and the imaging remote sensing satellite,making a deep research on the target association,and the main work is as follows:1.In view of the fact that the observed data from heterogeneous sensors can not be modeled,and the existing target correlation algorithms can not deal with the positioning error,false alarm and missing detections in the observed data,this paper presents a new method of the target association based on point pair local topolgy(PPLT).Firstly,the association of the ship position information from the satellite image information and electronic information is translated into the point pattern matching problem.The position information of ship target from the satellite image information and electronic information is regarded as the template set and the target set,respectively.And the point pair local topological feature,namely,point pair local topolgy(PPLT)is put forward to describe the two points set.Then PPLT is combined with probability relaxation labeling(PRL)and spetral matching(SM),respectively,puting forward two kinds of target association algorithms based on PPLT-PRL and PPLT-SM.Finally,the simulation results show that the PPLT-PRL algorithm and the PPLT-SM algorithm have good robustness to outliers,noise and rotation.Simulation experiments of ship formation show that the PPLT-PRL algorithm and the PPLT-SM algorithm have good robustness.The PPLT-PRL algorithm has good anti-false alarm ability,and the PPLT-SM algorithm has better anti-localization error ability,and verifies the practicability of the two algorithms.2.In view of the poor performance of the target correlation using only ship position information,and considering the relatively stable characteristics of the target attribute,the two are combined to complement each other,and a method based on point pair local topology(PPLT)and attribute information is put forward.Firstly,the feature of ship target is extracted,and the attribute feature extraction problem is transformed into target recognition problem.attribute features of the radiation source from satellite electronic information and attribute features of ship targets from the satellite image information are reduced by the rough set,respectively.And then the new feature combination is input into the IHDR classifier,the recognition result is obtained.While the identification results of the satellite electronic information are about the radiation source,using the configuration information of radars and ships is to get recognition results of ship targets.Then the probabilistic distribution function of PPLT is combined with the probability distribution function of target attribute using D-S evidence theory.Finally,applying probabilistic relaxing labeling(PRL)and spetral matching(SM)is to find the optimal solution,respectively.The ship target association algorithm based on PPLT-PRL and attribute feature,and the ship target association algorithm based on PPLT-SM and attribute feature are proposed.Experimental results in different scenarios show that the two algorithms have better anti-jamming and usability than the algorithm using only a single feature.And the ship target correlation algorithm based on PPLT-PRL and attribute features has better anti-interference,and the ship target correlation algorithm based on PPLT-SM and attribute features has better performance of anti-localization error.3.Considering many target association algorithm can not effectively deal with the non-rigid deformation and false alarm existing ship formations,a new method of the target association based on inhomogeneous gaussian mixture models(IGMM)is proposed.Similarly,the association of the satellite image information and electronic information is transformed into the problem of the point pattern matchng,and both target position information is equivalent to two point sets.The shape context of the two point set is extracted,and then inhomogeneous gaussian mixture models are constructed according to the similarity of SC.Finally using the EM algorithm iterates the transformation function of two po int sets and the false alarm rate.The experimental results show that the algorithm not only can solve the non-rigid deformation,but also is also robust to the position error,false alarm and missing datas.