基于区域协方差的惯性组合导航景象匹配算法

 针对景象匹配辅助惯性组合导航系统需要快速准确获取飞行器位置和航向偏差的要求,提出了一种基于区域协方差的实时图像匹配算法.算法采用区域协方差矩阵的距离作为图像匹配时的相似性度量.首先,对图像进行高斯平滑滤波,提取图像的多种特征计算区域协方差矩阵,利用金字塔多级匹配技术进行全局搜索,获得测试图在参考图中像素级匹配位置.然后,利用全局匹配的结果,在实测图上选取多个局部区域,分别进行局部搜索匹配,获取参考图和实测图中一一对应的局部匹配集.最后,利用RANSAC算法和最小二乘算法计算出两幅图像间的最优相似变换参量.仿真分析表明,该算法能满足景象匹配辅助惯性组合导航系统实时性、精确性和鲁棒性的要求.

Scene Matching Algorithm Based on Region Covariance for INS Integrated Navigation System

The scene matching aided inertial integrated navigation system needs to get both the aircraft position errors and the course deviation relative to the present flight path simultaneously.A real-time scene matching algorithm based on region covariance was proposed.The distance of region covariance matrixs as the similarity measuring for image matching was introduced.Firstly,the image was smoothed by a Gaussion filter,and features of the image were extracted.Global searching algorithm using multi-level pyramid matching was put forward.Then,multiple local regions in real image was selected,and local searching and matching was utilized.The local match set between real image and referance image was obtained.Finally,the least square algorithm ans RANSAC method were applied to get the distortion transformation parameters between real image and referance image.The results of simulation show that the proposed image matching algorithm can satisfy the accuracy demands of scene aided navigation system and provide the system location and course error-correcting information.