N帧连续景象匹配一致性决策算法

A Consistent Decision Algorithm for N Frames Continuous Scene Matching

下视景象匹配是各类巡航飞行器实现精确导航的重要手段,它是通过事先装载在飞行器的地面基准图像与实时获取的地面景象图进行相关匹配,完成飞行器的飞行误差修正和精确定位.由于基准图和实时图在获取手段、时间上等方面的差异,在相关匹配中,配准点往往落在相关阵的次峰上,导致匹配失败.该文提出了多帧连续景象匹配,利用单次匹配后形成的相关阵中的主次峰信息与导航信息进行一致性决策的图论算法.由于引入了相关阵中的次峰信息,并结合惯导位移进行一致性判断,使得匹配算法能有效剔除误匹配点,从而提高了算法的匹配概率和鲁棒性.

Looking down scene matching is a crucial technique for many cruise aircrafts to correct their navigation errors. By correlation matching between loaded reference image and real time image in the same area, the navigation errors of cruise aircrafts is corrected and precise positioning is implemented accordingly. Since the difference in acquisition sensor and time for reference and real time image, the true matched points are sometimes not on the primary peaks but on the secondary peaks of correlation matrix, which resulted in the matching was to fail. In this paper, a consistent decision algorithm for multi frames continuous scene matching is proposed. The core of the algorithm is to make use of the information of the primary and secondary peaks of correlation matrix and the displacement of inertial guidance to delete false matched points. Since the information of multi peaks and inertial guidance are introduced, the robust of algorithm and probability of matching are improved effectively.