视觉导航中陀螺仪辅助特征点匹配算法

为了解决视觉导航中无摄像平台条件下特征点匹配不准确的问题,提出了一种利用导航系统中陀螺仪输出辅助进行特征点匹配的方法.构建了陀螺仪角速度输出与特征点坐标变化之间的关系,分别采用相关系数匹配和双向匹配进行特征点匹配和匹配检验,通过插值计算得到亚像素精度的匹配结果;分析了该方法的各种误差来源,并对陀螺仪输出误差和摄像机焦距误差的影响大小进行了分析.实验表明:相对于SIFT(Scale Invariant Feature Transform)方法,此处提出的匹配方法准确性有了较大提高,并且在原系统基础上不需要增加额外设备,具有较好的实用性.

Gyroscope aided feature matching method in vision navigation

In order to solve the problem of feature points mismatching when there is no camera platform in vision navigation, a gyroscope aided points matching method was proposed. The relationship between the output of gyroscope and the coordinate changes of features was built. The nearest cross-correlation method and bidirectional matching algorithm was adopted to perform the point matching and matching result verification respectively. The sub-pixel matching result was obtained by interpolation. The error sources of the matching method were analyzed. The effect of the gyroscope output and the focus length calculation errors among them was calculated. The experiment of real scene proved that, the proposed method has much improvement in accuracy over scale invariant feature transform (SIFT), and it also has a better practicability for it does not need any other equipment except the existing ones.