合成孔徑雷達藉由雷達天線的移動，分析雷達與目標物的相對運動，處理雷達回波訊號的資料，藉此合成大的天線孔徑，以提高雷達的解析度。本論文首先介紹兩種合成孔徑雷達的成像演算法的流程，包含基本距離都卜勒演算法以及鳥鳴刻度演算法，利用機載雷達及衛載雷達的參數，模擬單點目標物的成像，並將模擬數據與成像標準做比較，機載雷達在斜視角7度，衛載雷達在斜視角6度可以得到合乎要求的峰值副瓣比。在雷達成像演算法中，都卜勒參數的估測是不可或缺的一環，都卜勒參數的誤差會導致雷達影像成像品質下降，為了降低雜訊對估測造成的影響，我們使用距離壓縮後的回波信號，搭配選擇性區間篩選出高能量的雷達資料，分別以多視互相關演算法及多視互乘頻率演算法進行點目標物的都卜勒參數估測，可得到較為準確的估測結果。根據參考文獻[1]，我們正確地判讀並萃取出RADARSAT-1的原始回波訊號及雷達發送訊號，利用RADARSAT-1的雷達訊號進行成像演算法及都卜勒參數估測的驗證，根據回波的頻譜失真程度，篩選可信賴的都卜勒基頻估測值，降低都卜勒基頻估測值誤差約二十赫茲，並且以回波的平均頻譜能量加權都卜勒歧義估測值，提高都卜勒歧義估測值的準確率。為了達成合成孔徑雷達的即時成像，必須使用硬體加速實現合成孔徑雷達成像演算法訊號以提升訊號處理的效率，由於雷達訊號的動態範圍非常大，我們使用客製化浮點數的形式設計雷達訊號處理的硬體架構，利用RADARSAT-1的雷達訊號決定分別以6位元及10位元表示客製化浮點數中整數位及小數位的位元數，並以10次迭代的座標旋轉數字計算來實現雷達成像演算法中的複數乘法，最終可達55dB的訊號量化雜訊比。;Synthetic Aperture Radar (SAR) synthesizes a large aperture to improve image resolution by moving the radar antenna and using signal processing techniques. This thesis introduces data flow of two SAR imaging algorithms, Range Doppler Algorithm (RDA) and Chirp Scaling Algorithm (CSA). A point target environment is simulated by using aircraft and satellite SAR parameters. The requirement of PSLR can be satisfied with squint angle smaller than 6 degrees for satellite SAR imaging and 7 degrees for aircraft SAR imaging. Doppler parameter estimation is essential in SAR imaging algo-rithm. The error of Doppler parameter estimation lowers the quality of SAR image. Using range-compressed radar signals with selective window technique helps to choose the region with large signal strength, which can improve the estimation accu-racy by the Multilook Beat Frequency (MLBF) algorithm. We use data of RADARSAT-1 to verify the SAR imaging algorithm and Doppler parameter estimation algorithm. The whole data of RADARSAT-1 are divided into several chunks to evaluate the relia-bility of estimated Doppler parameter by analyzing the symmetry of spectrum and the SNR level. The weighted combination of the Doppler ambiguity is used, which improves the correctness of detection. In order to implement real-time SAR imaging, the data quantization during the RDA flow is evaluated. The customized float-ing-point representation is employed with 10-bit mantissa for the FFT/IFFT processing. The complex multiplication can be realized by 10-stage CORDIC rotation to support 55dB output SNR.
