GPS视觉导航系统中多级实时匹配算法研究

The Study of GPS Vision Navigation System of Multi-stage Real-time Matching Algorithm

为了解决GPS导航中存在的精度低、不稳定和移动通信定位代价高的问题,引入了图像的外极限约束和多级实时匹配算法。该算法根据特征点对图像进行分级处理,首先匹配特征比较明显的点,再利用已匹配点的数据通过外极限约束确定后面几级像素点的视差范围,使得算法能够覆盖绝大部分点的真实视差,对于少部分落在搜索范围之外点的匹配,主要靠中值滤波去除。实验表明,多级实时匹配算法定位精确性和匹配速度优于传统的区域匹配算法。

There exists the problems of low accuracy, instability and costly mobile communication positioning in GPS navigation. The vision orientation system as an auxiliary navigation method can realize the function of high precision positioning within a large range. This paper introduces a matching constraint and layer-by-layer search structure which is suitable for GPS vision navigation system, and puts forward a feature aided multi-stage matching algorithm. The algorithm firstly makes the match of the more obvious characteristics points, then the matched points information are used to determine the subsequent points general disparity range through the geometric constraints, thus improving the matching speed and accuracy, it Makes the algorithm is able to cover most of the real parallax, the matching points that falls outside the searching scope, mainly remove by median filter. Using different types of outdoor natural topographic map matching experiments were conducted. Experimental results show that the multi-stage matching algorithm introduced in this paper is better than the basic area matching algorithm in matching accuracy and speed.