一种结合无迹卡尔曼滤波的矿区道路追踪算法

A mine area road tracking algorithm combined with unscented Kalman filtering

针对传统遥感影像道路中心线提取方法容易受到场景中噪声干扰这一问题,该文研究了一种局部窗口检测与无迹卡尔曼滤波全局追踪相结合的算法,并将其应用于高分辨率遥感影像矿区道路的提取。局部窗口检测分为内外窗口,外窗口利用边缘分布直方图计算局部道路方向,内窗口通过求解系列方差搜索道路中心点。在全局追踪阶段,综合考虑道路的先验信息和观测信息,运用无迹卡尔曼滤波实现道路中心点迭代追踪。该文选用高分辨率露天矿区道路遥感影像进行实验。实验结果表明,对于高分辨率遥感影像矿区不同曲率道路,该方法能够准确提取道路的中心线,对车辆压盖及遮挡等干扰因素具有稳健性。

Aiming at the problem that the traditional remote sensing image road centerline extraction method is susceptible to noise interference in the scene,an algorithm combining local window detection and unscented Kalman filter global tracking was proposed and applied to high-resolution remote sensing images extraction of mining roads in this paper.The local window detection was divided into inner and outer windows.The edge distribution histogram was used to calculate the local road direction in the outer window,and the road center point was searched by solving the series variance in the inner window.In the global tracking stage,the prior information and observation information of the road were comprehensively considered,and the unscented Kalman filter was used to realize the iterative tracking of the road center point.The high-resolution open-pit mining road remote sensing images were used for experiments,and the results showed that the centerline and boundary of roads with different curvatures in high-resolution remote sensing images could be accurately extracted,which was robust to interference factors such as vehicle cover and occlusion.