一种基于协方差矩阵的快速抗旋转景象匹配算法  

A Fast Anti-rotation Scene Matching Algorithm Based on Covariance Matrix

作　　者：符艳军 孙开锋[2] 刘军兰[1] FU Yanjun;SUN Kaifeng;LIU Junlan(College of Engineering,Xi’an International University,Xi’an 710077,China;Xi’an Precision Machinery Research Institute,Xi’an 710077,China)

机构地区：[1]西安外事学院工学院,西安710077 [2]西安精密机械研究所,西安710077

出　　处：《弹箭与制导学报》2022年第5期20-26,共7页Journal of Projectiles,Rockets,Missiles and Guidance

基　　金：陕西省自然科学基金(2020JM-637);西安外事学院高层次人才启动基金(XAIU2019011)资助。

摘　　要：针对无人机景象匹配导航系统对匹配算法的实时性和鲁棒性的要求,以协方差矩阵最小特征值为匹配算法的相似性测度,提出了一种改进的匹配算法。仿真结果表明,实时性方面,通过在离线方式下预先建立基于基准图的累加和矩阵,使协方差矩阵的在线计算量减少50%左右;抗旋转方面,对基准子图采用基于虚拟圆索引的快速圆形窗口定位算法,可以抵抗实测图5°~10°左右的旋转变化。该方法不仅在实时性、抗旋转方面有所改善,对实测图存在噪声干扰、小范围遮挡及小比例变化也有一定的鲁棒性。Aiming at the real-time and robust requirements of UAV scene matching navigation system,an improved matching algorithm is proposed based on the minimum eigenvalue of covariance matrix.In terms of real-time performance,by pre-establishing the summation matrix based on the reference image in offline mode,the online calculation of covariance matrix is reduced by about 50%in the matching process.In the aspect of anti-rotation,the fast circle window positioning algorithm based on virtual circle index is used for the reference sub image,which can resist the real-time image with 5°~10°rotation.The simulation results show that the proposed method not only improves the real-time performance and anti-rotation,but also has certain robustness to the noise interference,small range occlusion and small scale transform of the actual image.

关 键 词：景象匹配 协方差矩阵 最小特征值 实时性 抗旋转 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]