面向成像差异的高精度强适应SAR景象匹配算法研究

SAR scene matching can effectively improve the all-day and all-weather operating ability of the precision guidance weapons, which makes it have a very attractive application prospect in the precision navigation field. However, since the reference image and the real-time image are obtained by different SAR sensors which maybe have different imaging band, imaging angle and imaging conditions, therefore there exist great differences between two images in many aspects, which affects SAR scene matching performance seriously, and restricts the development of SAR imaging guidance weapons. Aiming at improving SAR scene matching performance,this research focuses on the relative theory and technology problems involved in SAR scene matching. The major research contents are included in the following. Firstly, a novel reference image generation algorithm will be proposed to generate reference images with high suitabiliy, which can improve the feature similarity between reference image and real-time image. Secondly, based on the mission characteristic of scene matching, the multi-scale, multi-type, adaptive common invariant feature extraction algorithm will be studied, and the multiple sub-region and multiple level matching algorithm based on spatial topology theory will be proposed, which can realize the high-precision multi-source SAR image matching. Finally, based on requirements for practical application, adaptability evaluation parameters will be designed for scene matching algorithms, the association study of scene matching algorithms with different typical target regions will be conducted through theoretic analysis and experimental verification, which can improve the adaptability of matching algorithms to complicated battlefield environments, and provide theoretic support for selecting the optimal matching algorithm in different operating environments. Theoretic breakthrough and technique innovation are expected for this research, which can promote the application of SAR imaging guidance techniques in precision guidance weapons.

SAR景象匹配可以有效提高精确制导武器全天时全天候作战能力，但匹配所用的基准图和实时图来自不同成像平台，存在较大成像差异，导致景象匹配性能下降，制约了SAR成像制导武器的发展。本项目针对此问题，以提高成像差异下的SAR景象匹配性能为目的，对其中涉及的基础理论和关键技术进行研究。具体内容包括：研究基于特征一致和图像反演的基准图制备方法，解决高适配性基准图生成问题，提高不同成像条件下基准图与实时图的特征相似性；研究多尺度、多类型、自适应的共性不变特征提取方法，并基于空间拓扑理论进行多子区分层次匹配方法研究，实现高精度的异源SAR图像匹配；从实际应用需求出发，研究匹配算法适应性评估指标体系，进行匹配算法与典型目标区域的关联性分析，提高匹配算法对复杂战场环境的适应性。本项目预期在上述方面取得理论突破和技术创新，对推进SAR成像制导技术在精导武器中的应用起到积极的作用。

SAR景象匹配可以有效提高精确制导武器全天时全天候作战能力，但匹配所用的基准图和实时图存在较大成像差异，导致景象匹配性能下降。本项目针对此问题，以提高成像差异下的SAR景象匹配性能为目的，对其中涉及的基础理论和关键技术进行研究。主要研究内容如下：.1）SAR基准图制备算法。针对基准图与实时图的成像差异问题，本项目提出了一种面向成像差异的景象匹配区层次选取法。针对基于单一特征指标选取景象匹配区不全面的缺陷，本项目提出了一种基于多属性综合评价的匹配区选取算法和一种基于主成分分析的景象匹配区选取算法。上述算法选取出的景象匹配区具有较强的稳定性、较大的信息量和较高的适配性，能够有效降低基准图与实时图差异对景象匹配性能的影响。.2）异源SAR景象匹配算法。针对景象匹配对算法匹配精度和速度要求较高的特点，本项目提出了一种基于轮廓压缩特征的多尺度景象匹配算法和一种多分辨率多子区SAR景象匹配算法。针对包含封闭均匀区域的SAR图像，提出了一种基于区域特征的由粗至精的SAR图像匹配算法。针对包含形状轮廓明显地物的SAR图像，提出了一种基于线特征的由粗至精的SAR图像匹配方法。针对连续多帧的SAR图像，提出了一种基于惯导信息融合的SAR连续景象匹配算法。上述算法能有效克服待匹配图像之间的灰度差异、旋转差异和尺度差异，对复杂噪声环境具有较好的适应性，达到了项目要求的技术指标，具有较好的应用前景。.3）匹配算法适应性评估。针对不同场景类型的匹配区内匹配算法的选择问题，研究了景象匹配算法的适应性评估方法。构建了评估指标体系，从实时图与基准图的差异性特征、景象匹配实验和典型场景成像原理三个方面进行匹配算法对典型场景的适应性评估，实现了匹配算法与场景类型之间的关联性分析。.4）综合实验验证。基于获取的大量实飞数据，构建了实验评估环境，验证了本项目所提算法的有效性和先进性。