【作者】 吕小微；

【导师】 赵亦工；

【作者基本信息】 西安电子科技大学 ， 模式识别与智能系统， 2011， 硕士

【摘要】 合成孔径雷达是自上世纪五十年代逐步发展起来的一种雷达成像技术,其高分辨率的成像能力,全天时、全天候的工作模式,多波段、多极化的发展趋势,促使其在民用、军用领域得到了广泛应用,极大方便了人们的生产生活,有利于国防建设,具有广阔的发展前景。目前,由于人们对信息的需求程度越来越高,SAR需要采集的回波数据日趋庞大,而现今有限的下传链路物理带宽严重影响了数据面向地面成像装备高效、及时的传输。介于此,SAR原始数据压缩技术应运而生。本文从数据压缩的基本原理入手,在这一部分重点介绍用于熵压缩的量化理论以及用于评价压缩能力的性能指标。随后,讨论了SAR成像的基本原理,并分析验证了SAR原始数据的统计特性。基于SAR数据的统计特性,完成了三种典型的SAR原始数据压缩算法的原理阐述和仿真,并逐一进行了对比实验和分析。现有SAR原始数据压缩算法主要分为标量量化、矢量量化和基于变换域的量化。本文选取的三种典型算法,即基于时域和标量量化的块自适应量化算法(BAQ)和幅度-相位算法(AP),以及基于变换域的FFT-BAQ算法。其中,BAQ算法是最基本的压缩算法,其分别针对数据的实、虚部做量化,运算简单,易于硬件实现。AP算法可以认为是BAQ算法的延伸,是特别针对保护数据的相位信息的压缩算法。在同比压缩下,实验结果表明,AP算法的相位保留程度大大优于BAQ算法,非常有利于后期成像处理。FFT-BAQ算法实际上是两个BAQ量化器和一个变换操作的组合,其利用数据经某种变换后表现出的能量更为集中的特点,对数据的高能量区执行多比特量化,低能量区执行少比特量化或零量化,通过变比特编码从而实现更好的保留数据原貌的功能。在同压缩比下,FFT-BAQ算法的信噪比明显高于BAQ算法,且成像结果更为清晰。更多还原

【Abstract】 Due to its strong ability of high resolution in imaging, all 24 hours a day and any weather in working mode, and multi-band and multi-polar in developing trend, Synthetic Aperture Radar (SAR) has been extensively applied to kinds of fields, both in public and by military. As one kind of Radar imaging technique developed eventually since the fifty’s in last century, SAR facilities the life of human-beings and the national defense greatly in recent years, and definitely own a wider development prospect. In order to meet the growing needs of information, the amount of echo data increases rapidly, however, which is not tolerated by the physical bandwidth of download link. Therefore, SAR raw data compression technique emerges with the tide in ages.This paper commences with the basis principle of data compression, focuses on the quantization theory used in entropy compression. Subsequently, we discuss the SAR imaging simply and analyze the raw data statistic character. Based on which, we choose and achieve three classical SAR raw data compression algorithms, all compared and analyzed in raw data domain and image domain.There are three ways existing in SAR raw data compression field, scalar quantization, vector quantization and based-transform domain quantization. In this paper, we pick up three algorithms to represent our experiment, namely by Block Adaptive Quantization (BAQ), Amplitude-Phase (AP) and FFT-BAQ. The first two are performed in time domain and based on scalar quantization and the last one is performed in frequency domain. Among them, BAQ is the earliest and most primary algorithm, which aims to the real part and imagery part of 2-D complex raw data and be easy to apply on hardware. AP is very similar to BAQ in some extent except it aims at keeping the phase information. The experiment result presents that AP leading the ability of protecting phase information which will be beneficial for imaging later. FFT-BAQ consists of two BAQ and one operation of transform. This method makes advantage of data reconstructed by energy concentration after some transformation, so the lower frequency part can be quantized with multi-bit and the higher frequency part can be quantized with low-bit, namely the bit-rate variable compression. In the situation of the same compression ratio, FFT-BAQ can retain raw data better than BAQ and present high performance.更多还原