The Sliding Window Type Target Maneuver Detection Algorithm Based On The Input Estimation

Maneuvering targets tracking has been widely recognized as a hotspot in current domestic and foreign research because of its extensive applications, both in military and civilian areas.How can we realize fast precision target tracking using a single model? It is critical that we should detect target maneuvering timely and effectively. Therefore,it is of great importance, both in theoretical research and engineering applications, to develop target maneuvering detection algorithm.Among these algorithms in current target maneuvering detection,Adjustable White Noise Model, Variable Dimension and Input Estimation are regarded as the classical ones. All of them have their own advantages and disadvantages in the implementation of target maneuvering detection according to different application environments. Input Estimation algorithm can estimate the size of target maneuvering directly from residuals without any prior assumptions about target maneuvering. So its unique benefit is widely approved in target maneuvering detection areas. Once a target maneuvering is detected,a compensation which lies in the estimation of target maneuvering is added to the filter immediately. It has been shown that Input Estimation algorithm can obtain good tracking performance comparatively. Unfortunately,Input estimation algorithm has its inherent defects, too. This paper will mainly aim at analyzing the shortage of current Input Estimation algorithm and developing some new ideas,so that we can improve its detection deficiency and acquire better tracking performance.Frist of all, this paper reviews the main target manevering detection algorithms briefly. Then comparisons of detection performance is given accoring to simulations of various tracks. On the basis of the analysis above, benefits and limitations of Input Estimation are summarized and suggestions of development are given.Then, Bogler algorithm is discussed and an improved measure is given by develop some of its derivation process. Simulations confirm that the improved algorithm has better tracking performance.After that, a new estimate algorithm is developed by using residuals between predicted value and actual observation of position,which realize the decoupling of designing of detector and estimator. As a result, the algorithm reduces computational load and improve the efficiency of the calculation. Simulation results show that the modified algorithm acheives the desired effect.Finally, this paper provide a comprehensive consideration of the selection of statistics, sliding window length and probability of false alarm,which have a significant impact to detection delay. According to the characteristics of these factors,analysis is developed either in qualitative or quantitative aspects. Relationships between these factors are discussed at last and some optimization plans are given.