一個新的空用雷達激動目標追蹤演算法

Abstract

激動性目標追蹤法(Maneuvering Target Tracking)的目的在於精確地掌 握一個加速度突變目標物的軌跡。現今對於激動性目標物追蹤的解法約略 可分類為估差植基法(Residual Based methods)及多模組法(Multiple Models methods )。多模組法能提供較精確的軌跡估測追蹤，但和估差植 基法相較得付出高計算複雜度的代價。在本論文中，我們介紹一個新的激 動性目標追蹤演算法。這個方法將嘗試借鏡上述方法的優點來達到低計算 複雜度並具有相當精確度的要求。交連式多模組法(Interactive Multiple Model)和輸入估測法(Input Estimation)是兩個上述追蹤法分 類的代表性演算法。在本論文中，我們採用這兩個演算法的特性。透過實 驗模擬，我們發現新的演算法的確能達到像交連式多模組法的高精確度， 而分析我們演算法的計算複雜度僅僅只需透過一個卡爾曼濾波器(Kalman filter)來建置。 Maneuvering tracking is to track a target which is suddenly accelerated by the piolt. There have been two approaches proposed for maneuvering target tracking: residual-based approach and multiple-model approaches. It has been known that the latter approach provide a better tracking performance, but has a much higher computational complexity, as compared with the former one. In this thesis, we present a new maneuvering tracking algorithm which can extract the benefits of both approaches so that the tracking accuracy can be compatible with that of multiple model while complexity with that of residual- based approach. Interactive multiple model method and input estimation method are two outstanding maneuver tracking algorithms belonging to residual-based and multiple model approaches, respectively. In this thesis, we have adopted some features inherenting from the two algorithms. Our simulation has shown that the new algorithm can have a tracking accuracy which is compatible with interactive multiple model. Analyzing the complexity, we find that the algorithm is as simple as a single Kalman filter.