高取樣率雷達系數對戰術運動目標的追蹤技巧

Abstract

When a maneuvering target is tracked by a radar system at high measurement frequency, the measurement noise is usually highly correlated and is often modeled as a first-order Markov process. By generating the artificial measurement and reformulating the measurement equation, some efficient decorrelation algorithms can be derived corresponding to several maneuvering target tracking methods: the state augmentation method, the maneuver detection method, the input estimation method, the Gholson's multiple model method, and the interacting multiple model method. If the measurement noise is complicated and is modeled as the sum of a high-order autoregressive process and a white process, then, by adding some noise-correlation variables to the target state, the decorrelation process can also be completed. The tracking performances with and without decorrelation are carefully evaluated by theoretic analysis or computer simulation results. It can be found that significant improvement, particularly in velocity and acceleration estimations, can be provided by the decorrelation process. If some of the parameters including the noise-correlation parameters are undnown, these parameters should be estimated before the decorrelation process starts to work. If the state augmentation method is employed for tracking the target, the autocorrelations of the innovation can be obtained as functions of the unknown parameters. Then, using this relationship and taking the time-average autocorrelations to approximate the statistical autocorrelations, the unknown parameters can be estimated adaptively. A modified structure denoted as 'multiple level estimator' may provide good performance in estimating the parameters with efficient computation. If the maneuver detection method or the input estimation method is employed for tracking, a simple (approximate) parameter estimation technique is available. 高取樣率雷達系統追蹤一戰術運動目標時，其測量雜訊常呈相當的關聯性而可模式 化為一階的馬可夫過程。對應於幾種常見的戰術運動目標追蹤技巧：狀態擴充法， 戰術運動偵測法，輸入估計法，高爾森多重模式法及交連式多重模式法，可利用產 生人工測量值的過程，重新推導測量方程式，而得到幾種有用的消除雜訊關性的方 法。如困測量雜訊相當複雜，而被模式化為一高階自動遞迴過程和一白色過程之和 ，則將雜訊關聯性變數併入目標狀態考量，消除雜訊關聯性的作用也可以完成。在 本論文裡，消除雜訊關聯性前後的追蹤效困均以理論分析或計算機模擬作了詳細的 評估與比較。我們可發現，經過消除雜訊關聯性的處理，追蹤效困得到重大的改善 ，特別在對目標速度及加速度的估計上。 如困包括雜訊關聯性參數的某些參數為未知，則在執行消除雜訊關聯性處理之前， 這些未佑參數宜先估計出來。若我們用狀態擴充法追蹤目標，則推導出測量值創新 處理的自相關函數為未佑參數的函數，並取自相關函數的時間平均值代替統計值， 可將這些參數適應性地估計出來。本論文裡還提出一個取名為〞多重層級估計法〞 的修正式結構，可以有效地計算參數而達到良好估計效果。此外，如果以戢術運動 偵測法或輸入估計法來追蹤目標，參數估計可以用一簡化的近似法求得。