基于非下采样Contourlet变换的无人机景象匹配算法

卫星导航作为无人机重要的导航方式常因受到干扰而不可用,针对此问题,提出一种同样可辅助惯性设备进行导航的景象匹配算法。首先采用非下采样Contourlet变换(NSCT)对基准图和实测图进行多层级分解,以获得不同分辨和方向的稀疏表示图像,然后以基于稳健统计量的Hausdorff距离(LTS-HD)作为相似性测度进行由粗到精的分级匹配。仿真结果表明,基于NSCT方法的图像匹配概率相对于小波分解提高了14%;景象匹配辅助惯导后,可有效抑制位置误差发散,误差被控制在30 m以内,可满足无人机中等精度导航要求。

Scene matching algorithm for unmanned aerial vehicle based on Nonsubsampled Contourlet transform

As an important navigation method for unmanned aerial vehicle(UAV),the satellite navigation is always unavailable because of the interference.To solve this problem,a scene matching algorithm which can improve the ability of inertial navigation is proposed. Firstly, the target image and reference image are decomposed into multi-scale and multi-direction sparse representation image with nonsubsampled contourlet transform(NSCT),then the least trimmed squares Hausdorff distance(LTS-HD) is used as similarity measure for roughness-to-precision hierarchical matching. The simulation results show that the matching probability using NSCT is 14% higher than using wavelet decomposition and the divergence of position error has been brought under effective control when using the scene matching aided inertial navigation system.The position error has been controlled within 30 meters and can meet the needs of medium accuracy for UAV navigation.