Multi-sensor Target Tracking Data Fusion Technology

Multisensor target tracking is a paradigm of information fusion technique dealing with target tracking problem. It estimates target state through intelligent integration of multisensor data to gain better tracking performance than a single sensor does.Mulitisensor target tracking is the intersectional technique of multiple subjects. It has gained popularity over past decades with the advent of vigorous sponsorship in many areas. The successful applications of multisensor information fusion technique have demonstrated an attractive perspective. However, many relevant theories and techniques remain to be developed to meet the various needs in practice.Based on multisensor target tracking system,the research of this thesis focuses on target location technique, target trajectory smoothing, filter, prediction based on a parameterized model, target tracking with nonlinear measurement, outlier detection and distributed systematic error estimate based on real-time data fusion and so on. The main achievements are shown as the following:In multisensor target tracking system, the measurement available is usually uncertain. To address this problem, a sequentially fusion method of location and velocity measurement data is proposed. This method fuses available measurement one by one. As a recursive algorithm, its physical meaning is remarkable with high accuracy , and it is easy for practical use.Target trajectory smoothing, filter, prediction technique based on a parameterized model is widely used in aerospace measurement and control. Here, the author put the emphasis on target trajectory smoothing, filter, prediction technique based on a parameterized model in multisensor target tracking system. In this situation, we are confronted with not only the problem that truncation error is large while the target maneuvering and the trajectory changing notably, but also the problem that how to fuse all kinds of data effectively. In this case, the author puts forward a novel filter named as free node spline function polynomial least square filter based on spline function theory. Compared with a polynomial expression of trajectory, the spline expression of the trajectory can make the truncation error decrease greatly; Combined with the matched rule of target state,which means position parameter are difference of velocity parameter, the match polynomial least square filter and a filter that fuses state information and range rate information are both given. Because of the use of prior information, the number of parameter to be estimated is reduced.and the parameter estimation accuracy is enhanced.The extended Kalman filter(EKF) and converted measurement Kalman filter(CMKF) have been widely used in radar target tracking. However, the performance of these algorithms that are based on linear approximation degrades considerably in highly nonlinear situation. Whereas the EKF requires the evaluation of the Jacobian to obtain the observation matrix, the CMKF needs it to compute the measurement error covariance. Both of them employ linear approximation, andthus linear error is inevitible. When nonlinear degree is high, the linear approximation error must be considered. Chapter 4 discusses the target tracking problem with velocity measure. A filter algorithm that can overcome the nonlinear effects is developed. Based on the correct evaluation of the means and covariance of the measurement error in Cartesian coordinate system, the algorithm processes the radar measurements sequentially, and the linearization of measurement equation is no longer neccessary. The performance of this algorithm is excellent, and estimation accuracy and computation efficiency are both improved.In chapter 5, the method of combining the exact target state parameter fused by multi-sensor tracking data and the tendency of the whole tracking data to detect outliers is proposed. Distributed multi-sensor fusion method can resolve the problem of patchy detection efficiently and quickly. Distributed track fusion problem in multi-sensor tracking system wh...