利用方向纹理特征从影像上搜索道路中心线

摘要

针对遥感影像道路中心线提取问题，提出利用方向纹理特征辅助道路中心线匹配搜索的思想，以人工给出的道路种子点和初始搜索方向为基础，运用方向纹理特征直接从遥感影像上匹配搜索道路的中心线。算法通过预测道路的宽度、弯曲等情况来自动调整纹理窗口的大小，因而能够很好地搜索到道路中心线，此外算法针对树木遮挡、车辆压盖设计了相应的处理策略。实验表明：该方法对遥感影像上不同类型道路均有很好的提取效果，且提取效率高于现有的道路提取算法。%Aiming at intelligent extraction of the road centerline from remote sensing images an idea of searching the road centerline with angular texture signature is proposed in this paper. Based on the artificially provided seed points of the road and the initial search direction the algorithm directly search and match the road centerline with the angular texture signature from remote sensing images. Since the method is able to automatically adjust the tex-ture window size by prediction of the width bending of the road etc it performs well in searching the centerline. Furthermore the method can overcome the influence caused by the presence of vehicles and shadow of trees. The experimental results show that the method can achieve a perfect extraction effect on the different types of the roads on remote sensing images and has better searching efficiency than the existing road extraction algorithms.