調頻連續波雷達系統之多目標物偵測與辨識

Abstract

主動式安全在車用雷達系統的設計上已成為近年來的趨勢。為了掌握目標物的動態，目標物偵測的設計成了系統中最重要的一環。在傳統的設計上主要採用恆虛警率(constant false alarm rate)的偵測器(detector)搭配角度估計器(angle estimator)用以得到精確的目標物資訊。然而，在多目標物的環境及雜波(clutter)的影響下，偵測器會發生為了滿足恆虛警率而犧牲掉偵測效能的問題。除了偵測器之外，角度估計也會因為信號在頻譜上的重疊(overlapping)而導致估計錯誤。在本論文中，吾人首先利用頻譜上鄰近頻段之間的相關性設計了可調整參數的動態偵測程序。該程序在各種環境之下皆能維持穩定的偵測效能。為了再進一步處理雜波造成的影響，吾人利用先前可靠的追蹤資訊設計一個程序來避免偵測遺漏。在角度估計上，吾人根據陣列信號間的特徵設計了辨識機制(identification)，用以解決信號重疊造成的影響。經由模擬驗證，吾人所提出的程序相較於傳統的目標物偵測系統改善了其在多目標物環境下的偵測品質，並能根據環境做動態的參數調整，以達到主動式安全的需求。

Active safety has caught much attention in automotive radar system recently. To acquire target’s movement, the detector plays an important role in the system. Conventional detectors adopt constant false alarm rate (CFAR) with an angle estimator to obtain precise target information. However, in the multi-target scenario, performance of the detector degrades due to the constraint of CFAR. In addition, signal overlapping in the spectrum leads to angle estimation error. In this thesis, we first design an adaptive detector to maintain the detection performance in time-varying environment. To deal with clutter effect, we propose a miss avoidance procedure by using the reliable prior tracking data. In addition, identification is proposed to mitigate the overlapping effect according to the characteristics of array signal. From simulation results, the proposed procedure achieves the goal of active safety by improving the detection quality and adjusting parameters dynamically according to the environment.