基于面向对象算法的道路信息提取研究

摘要

Road,as one of the most important facilities.The extraction on of road is of great significance to the spatial database update of ground.The object-oriented thought is introduced to the road extraction on images in this paper.In accordance with the" heterogeneity" feature of local and adjacent regions,makes multi-scale segmentation,generates a set of "homogeneous" pixels,and then adopts the nearest neighbor classification.It aims at exploring the optimal method of road information extraction which is based on object-oriented algorithm.%道路作为重要的基础设施,其信息的快速提取对于地面空间数据库的更新具有重要的理论与现实意义.本文将面向对象的思想引入影像道路分析提取中,按照局部区域与相邻区域的“异质”特征对高分辨率影像进行多尺度分割,产生“同质”像素集,得到最优尺度参数;然后通过探究最优待征组合及最邻近分类提取,面向对象道路提取用户精度可以达到96.5％.通过多次实验对比分析,旨在探索基于面向对象算法道路信息提取的最佳方法.