采用循环相位相关的LPT景像匹配算法  

An Algorithm for Scene Matching Using Circulation Phase Correlation and LPT

作　　者：雷蕾[1] 李向阳[1] 齐浩程[2] 

机构地区：[1]中国飞行试验研究院,西安710089 [2]西安应用光学研究所,西安710065

出　　处：《电光与控制》2010年第1期93-96,共4页Electronics Optics & Control

摘　　要：景像匹配辅助惯性组合导航系统中的景像匹配算法必须同时具有克服旋转和小尺度变换的能力。通过分析发现,对数极坐标变换(LPT)能将它们分别转换为子变换图之间的循环、线性平移关系。鉴于经典的相位相关法检测的是图像间的线性平移,提出了一种改进相位相关法来检测循环平移,最终利用LPT变换和新相关法实现了景像匹配算法。仿真分析表明,当存在任意的旋转、小范围的尺度变化和噪声畸变时,本文算法匹配精度达到93%以上,时间在1.3s以内,基本满足了景像匹配辅助导航系统对图像匹配算法准确性和实时性的要求。Image matching algorithm for scene matching aided navigation system must have ability to overcome scale and rotation at the same time. The Log-Polar Transform （LPT） can respectively transform them into linear or circulation translation relations between the sub-transformation charts. Because the phase correlation techinque only determines the linear-translation movement, an improved phase correlation method was presented,which can detect the circulation translation. The new correlation method and LPT were used for scene matching. The experiment showed that the algorithm present here has a matching precision of 93% within time of 1.3 s, which can meet the needs of accuracy and real-time performance in scene matching aided navigation system.

关 键 词：景像匹配 循环相位相关 对数极坐标变换 组合导航 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP301.6[自动化与计算机技术—计算机系统结构]