高測頻下閃躲目標追蹤

Abstract

在一般的追蹤系統中,卡曼濾波器 (Kalman filter) 已經被廣泛地應用。 於白高斯雜訊 (white Gaussian noise) 的假設下,該濾波器方為最佳。 然而在一些應用上,雜訊並非如此。高量測頻率時,雜訊不能再被視為白雜 訊;並且,由於追蹤目標表面不規則的電磁波反射,使得雜訊非高斯 (glint noise)。本論文即探討在這些雜訊環境下的閃躲目標追蹤法則。 我們提出了一個簡單又可行的方法來估計彩色雜訊 (color noise) 之參 數。從所附的模擬結果,可以看出這個方法獲致了很好的效果。在對付 glint 雜訊的問題上,我們採行前置處理 (preprocessing) 的概念,以 (Order Statistic Filters) 取代 Hewer 所提的穩健性估計器。OSFs使 得這個概念的實行更為簡化並且也大幅增進了追蹤性能。 The Kalman filter is widely used in the tracking problem. The Kalman filter is widely used in the tracking problem. The optimality of the Kalman filter is claimed under the assumption of white Gaussian noise. In some applications, the assumption is not valid. The noise arising in a system with high measurement frequency is not white and the noise due to target glint is non-Gaussian. In this thesis, we consider the maneuvering target tracking under these environments. We propose a simple and practical method to estimate the parameters of color noise. It is shown the method achieves satisfactory results. For the treatment of glint noise, we adopt the concept of preprocessing. Order Statistic Filters (OSFs) are proposed in place of the existing robust estimator. The OSFs provide simpler implementation and do improve the tracking performance.