基于鲁棒曲线匹配的星表特征跟踪方法

Planetary Terrain Features Tracking Method Based on Robust Curve Matching

针对星际着陆的特征跟踪及障碍识别过程,提出了一种基于尺度不变的曲线描述及匹配方法。该方法利用曲线邻域的梯度信息形成特征描述符,无需知道各点曲率或曲线之间的几何构型,且对旋转、缩放、视角变化及光照影响都具有较强的鲁棒性。仿真实验表明,该算法不仅能够对不同尺度的陨石坑、岩石边缘等障碍或自然陆标进行识别,而且能够实现边缘曲线特征的跟踪,误匹配率低。

A curve description and matching algorithm which utilizes the scale eon.,tant is presented for feature tracking and hazard recognition process in planetary landing. The way to form the curve descriptor which is constructed with the curve neighborhood gradients information is introduced. The robustness and the efficiency of our graph matching process is shown in rotation, scale, moderate viewpoint changes and light effects, without knowirg the curve geometry or geometrical configuration between the curves. This algorithm can not only be used for edge curve feature tracking with low mismatching, but also effectively identify the craters, rocks and other hazardous or natural landmarks at different scales.