Research On Data Fusion Methods In Airborne Multi-sensors For Target Identification

Data fusion is one of the core technologies of airborne multi-sensor systems,especially in target identification applications with some special constrains and requirements.In this dissertation,we pursued the high-effective target identification in an environment of the airborne platform,through the multi-sensors data fusion technologies.We focus on the key techniques such as the information registration and the decision with data fusion and provide some creative and innovative contributions as follows:(1)Accurate and reliable information registration.At first,we investigated the flexibility of airborne,the complexity of motion model and the heterogeneity of sensors that can cause the observed data with time-difference problems.So we built the model via time registration deviation estimation and analyzed the time registration deviation based on the typical target motion model and the complex target motion model.Secondly,according to the difference of observed coordinate from airborne sensors which didn't match with their target information space locations,we studied space registration method based on the polar coordinates and built a common model for space registration.And also we made the accurate estimation for the registration deviation according to the Kalman filtering method.We testified the algorithm performance through simulation experiment and proved its usability to airborne platform.The results show that the time registration method with complicated and unknown motion model can estimate the motion information and its error at any time point.The spatial registration error converges to the true value.(2)Aimed at the target dense distribution and the airborne platform sensors information association issues caused by complex electromagnetic environment,we proposed a method of synergetic information association based on identification feedback.We solved the problems both to traditional method error and highly missed related rate.Simulation experiments show that compared with the traditional method based on dynamic information,the accuracy of the proposed method is improved by 2%~5% under the condition of dense target distribution or cross-movement.(3)Aiming at the problem of fuzzy uncertainties and the conflict of multi-sensor information in airborne platforms that may result in low identification accuracy or even misidentification,a hierarchical multi-period fusion decision method based on conflict intensity measurement is proposed according to the different identification requirements.Firstly,a conflict intensity judgment method based on the confidence interval metrics is constructed to measure the conflict of evidence information from different sensors.Secondly,based on the measurement results of conflict intensity,different weight factors of evidence are set up to reinforce the system contribution of high-directivity consistency evidence.Finally,a hierarchical multi-period fusion method is proposed to resolve the weight correction evidence,in which the high-confidence evidences are firstly merged,the low-confidence evidences are followed,and the role of historical information is reinforced.Simulation experiments show that the optimal fusion level can be obtained according to different identification requirements to achieve the balance of accuracy and timeliness.At the same time,the method can overcome the interruption of strong conflict evident,and restrains much faster.(4)For the circumstances that the recent relevant fusion techniques are lack of classic situation application testify,we built a half-physical simulation testing environment based on the background requirements of national research projects and the research process.It combines classic combat scenes,operational objects and sensor placement and so on.Under the situation from above,the simulation contains multi-sensor fusion technique and key performance criteria.