在SAR匹配导航系统中,由于基准图和实时图尺寸较大,而且二者被获取时成像角度和分辨率等参数不同,使得传统景象匹配算法很难同时满足实时性和鲁棒性要求.我们通过对多种小波函数的对比分析,挑选出了性能最优的一种,并构造图像金字塔,利用由粗到细的迭代匹配策略实现了算法的实时性.设计了基准图和实时图的多尺度圆模板,并基于D-S证据理论进行融合,实现算法的鲁棒性.统计试验结果表明,算法在保证匹配时间的前提下可以容忍较大的几何畸变.%In the SAR matching navigation system, the large size of reference image and real-time image, together with the difference in imaging angle,resolution,etc,makes the traditional matching methods can not meet the real-time performance and robustness simultaneously. Comparing various wavelet functions, we select a better one to construct the image pyramid,and realize the real-time performance via a coarse-to-fine matching strategy. The multi-scale circular templates are designed, and fused by the Dempster-Shafer (D-S) evidence theory,achieving the robustness.Evaluation results indicate that the proposed method can accommodate big distortions on the premise of ensuring the matching time.
展开▼
