【作者】 任三孩；

【导师】 常文革；

【作者基本信息】 国防科学技术大学 ， 信息与通信工程， 2011， 博士

【摘要】 将合成孔径雷达(Synthetic Aperture Radar,SAR)技术应用于精确制导武器,不但可以有效提高其全天时、全天候的工作能力,而且增加了探测隐蔽和伪装目标的能力。紧密结合我军当前精确制导技术的发展需求,本文围绕SAR回波模拟技术、SAR实时图预处理技术、SAR基准图制备技术和SAR景象匹配技术四个方面开展了研究。本文主要工作及创新点概括如下:1.研究了高动态飞行条件下复杂地形SAR回波模拟方法。针对传统电磁建模方法无法同时模拟地物散射特性差异和地面高程起伏的问题,提出一种结合数字高程模型(Digital Elevation Model,DEM)和正射图的电磁建模方法,提高了三维场景电磁建模的逼真度。详细分析了雷达平台运动误差对瞬时斜距和回波信号的影响,推导了雷达平台存在视线误差时回波信号的二维频域表达式。针对弹载平台的高机动飞行条件,提出一种时频域处理相结合的SAR回波快速模拟方法,并分析了该方法的有效性。2.研究了SAR实时图中相干斑噪声和图像几何畸变的处理方法。主要包括两个方面:(1)针对传统相干斑抑制方法耗时过长的问题,设计了一种在抑制效果和运算速度上折中的相干斑抑制方法;(2)针对弹载SAR多普勒中心频率与地面高程起伏无关的特点,提出将SAR图像方位向几何校正与运动补偿一起完成,然后校正距离向畸变,大幅度降低了几何校正的处理难度。为进一步提高几何校正方法的实时性,对所提方法进行了有效的改进,并根据不同地形区的SAR实时图进行了实验,得到了几何校正方法选取的准则。3.研究了复杂地形SAR基准图制备技术。具体工作为:(1)针对SAR适配区选择制约因素多、参数设计难和适配性预测函数构建复杂等问题,提出一种SAR适配区分层选择方法。该方法通过适配性特征参数的逐层筛选降低了预测函数的复杂度,为选取高性能的SAR适配区提供了参考。(2)针对复杂地形SAR图像特征点提取难度大的问题,提出一种基于DEM和图像叠掩特征的SAR基准图自动制备方法。该方法通过改进的特征点提取方法和两级特征点匹配方法,提高了基准图的制备精度。详细分析了平台运动误差和DEM高程误差对基准图制备精度的影响,为基准图制备方法的实际应用提供了指导。4.研究了SAR景象匹配技术。针对弹载SAR实时图与基准图尺寸大,而且二者的成像角度、成像平台等可能存在差异等特点,提出将图像灰度和尺度不变特征变换(Scale Invariant Feature Transform,SIFT)特征点结合起来,采用由粗到精的方式完成实时图与基准图的匹配。该方法通过一种基于小波变换和变尺度圆模板相融合的匹配方法,解决了灰度相关法耗时长及对图像旋转和尺度变化敏感的问题,提高了图像匹配的实时性和鲁棒性;通过一种预处理方法和特征点外点筛选方法,解决了SIFT算法运用于SAR图像时出现的问题,提高了特征点的匹配精度,实现了亚像素级的图像匹配。本文基于大量的仿真数据和实测数据,开展了充分的实验验证,展示了本文研究成果良好的实际应用价值。更多还原

【Abstract】 The precision guided weapon, which makes use of synthetic aperture radar (SAR) technology, not only improves the all-weather and all-day capabilities, but also increases the capabilities of detecting concealed and camouflage targets. Focused on the exigent requirements of precision guidance technology, four subjects are researched in this dissertation: SAR raw signal simulation, SAR real-time image preprocessing, SAR reference image preparation and SAR scene matching.The main achievements and contributions of the dissertation are as follows:1. The technology of SAR raw signal simulation under conditions of flight trajectory deviation and complex terrain is studied. A method of backscatter modeling, which combines the digital elevation model (DEM) and orthography, is introduced since the traditional modeling methods have problems in simulating the difference in backscatter properties and terrain undulation simultaneously. The method models the three-dimensional scene with high fidelity. The influence of flight trajectory deviation on slant range and raw signal is analyzed, and the two-dimensional expression in frequency domain of raw signal with motio errors is derived. A fast raw signal simulation method which combines the time domain and the frequency domain is proposed based on the large flight trajectory diviation, and the validity of the method is also discussed.2. The preprocessing method for speckle noise suppression and geometric distortion correction in SAR real-time image is studied, which includes two main aspects: (1) In order to decrease the computation time of traditional methods for speckle noise suppression, a compromise method between suppression performance and calculation speed is introduced. (2) A novel geometric correction method, which corrects the geometric distortion in azimuth direction in the process of motion compensation, and then corrects the geometric distortion in range direction in succession, is proposed based on the fact that the Doppler center has no relationship with the terrain undulation for missile-borne SAR. The method decreases the computation complexity greatly. In order to refine the method, several improvements are made on it. Then, some experiments are accomplished based on the refined methods, and some rules of geometric correction method selection are acquired.3. The technology of SAR reference image preparation under complex terrain is studied, which includes: (1) In order to resolve the problems of SAR matching area selection, such as the constrain conditions are excessive, and the construction of feature criterions and predicting functions is difficult, a hierarchical way of selecting optimal scene matching area is presented. The method solves the difficulties in constructing predicting function by feature selecting hierarchically, and provides a reference for choosing better SAR matching areas. (2) Based on the fact that it is difficult to extract precise control points (CP) in SAR images under complex terrain, an automatic SAR reference image preparation method based on DEM and SAR layover is developed. The method improves the geometric accuracy by an improved CP detection method and a two-stage CP matching method. The influence of motion errors and DEM errors is also analyzed, which is important for application.4. The technology of SAR scene matching is studied. A coarse-to-fine image matching method, which combines the gray and scale invariant feature transform (SIFT) point feature, is proposed according to the characters of SAR scene matching, such as the sizes of real-time image and reference image are both large, the imaging angles and the imaging platforms of them are different. In order to decrease the computation time and the sensitivity to rotation and scaling change of the cross-correlation matching method, a matching method based on wavelet transform and multi-scale circular template fusion is introduced to satisfy the requirements of the real-time and robust. In order to overcome the problems when SIFT is used in SAR image, a preprocessing and an outlier filtering method is presented to improve the accuracy of SIFT features, based on which the sub-pixel matching accuracy is reveived.A large amount of experiments based on simulative and real datas are accomplished to show the significance of the study.更多还原