Research On The Algorithm Of Radar-network Registration

Data registration is an important part of information fusion in radar-network. The key to data registration is radar system errors estimation. The disadvantages of most existing algorithms for radar system errors estimation include relatively simple models, too many constraints and so on. The most important point is that the observability problem is seldom considered in the algorithms. According to some relative research results, however, it is proved that the influence of the observability cannot be ignored in system errors estimation for it may determine the accuracy of the results in certain circumstances.The thesis analyses the observability and focuses on the system errors estimation problem in low observability circumstances. The main contributions can be summarized as follows:(1) The influence of the observability on the precision of the system errors estimation is difficult to be described quantitatively. Aiming at the problem, a model for the radar-network errors estimation is established and a system errors estimation algorithm based on the detection of the observability is proposed. It can be concluded by theoretic analysis that the observability of the system errors estimation model established with the true track data and the observed data from radars separately is approximately the same. Based on the fact the observability of the system is evaluated by an indirect detection method. Simulation experiments show that the performance of the whole algorithm is good.(2) The imbalance of the distribution of the observation data and the increase of the condition number of the information matrix enhance the influence of the noise on system errors estimation in low observability circumstances which may result in the divergence of the estimation. To solve the problem, a system errors estimation algorithm is proposed which is based on low-pass filtering. The noise within the original observation set is removed by the preprocessing which enhances the precision of the system errors estimation. Experimental results show that the performance of the existing algorithm has been improved greatly.