基于改进KAZE特征的合成孔径雷达匹配算法

自主导航系统是高超声速飞行器自主控制性能的关键因素.合成孔径雷达(SAR)系统作为高超声速飞行器的关键导航设备,其获取的SAR图像受高超飞行器影响存在严重噪声干扰.在分析KAZE特征检测算法的基础上,针对KAZE算法计算复杂度高的问题,引入Fast Explicit Diffusion算法,设计了快速非线性尺度空间构建方法,并在此基础上提出了改进KAZE特征检测算法和基于改进KAZE算法的惯导/SAR组合导航景象匹配算法.通过仿真验证表明,提出的改进KAZE特征算法在保留KAZE算法噪声鲁棒性的同时,有效降低了KAZE算法的复杂度,其计算效率与SIFT算法相当,位置误差估计精度优于SIFT算法.

SAR Image Matching Algorithm Based on Improved-KAZE

Abstract： Autonomous navigation system is the key technology of hypersonic vehicle(HV) autonomous control. As the key navigation system of the hypersonic vehicle, the synthetic aperture radar(SAR) images are affected by serious speckle noise. The SINS detecting features in nonlinear scale space are highly time consuming due to the computational burden of creating the nonlinear scale space. Based on the analysis of the characteristics of KAZE, the fast explicit diffusion was introduced to accerlarate feature detection in nonlinear scale space. Furthermore, an improved KAZE algorithm and SINS/SAR integrated scene matching algorithm was proposed. The simulation results indicate that the improved KAZE algorithm has better accuracy and robustness than SIFT, and can effectively improve the computational speed.
Key words： synthetic aperture radar; nonlinear scale space; image matching; integrated navigation; hypersonic vehicle