摘要
针对目前高分辨率遥感影像的道路自动提取算法研究中的不足,该文提出了一种基于并行角度纹理特征的半自动道路提取算法:用户输入完成道路中心线上的起始点、道路方向、道路宽度等初始化工作,利用并行角度纹理特征获取道路前进方向,用抛物线参数方程构建道路轨迹模型来预测道路轨迹点,使用角度纹理特征值构建的紧质度系数和抛物线的曲率变化来约束道路轨迹点,验证失败则转入手工跟踪;往复执行以提取道路中心线。试验证明,本算法是一种稳健的道路半自动提取算法。
Aiming at the insufficiency of studying on the automatic extraction of roads from high-reso- lution RS imagery, the paper proposed a semi-automatic road tracker based on Parallel Angular Texture Signature （PATS） ： users performed the primitive input of the start point on the central line, the direction and the width of the road, employed PATS to get the moving direction of current road, and used parabola to model the road trajectory and predict the road track point, then verified the new added road point by the compactness of Angular Texture Signature polygon and the curvature change of the parabolic. Meanwhile, if the verification was failed, the man-made tracking was started. The processing was executed repetitively for the extraction of the road. Experimental result showed the feasibility of the semi-automatic method.
出处
《测绘科学》
CSCD
北大核心
2015年第5期55-59,共5页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41371405)
关键词
半自动道路提取
高分辨率遥感影像
并行角度纹理特征
模板匹配
Snakes算法
剖面匹配
semi-automatic road extraction
high-resolution remotely sensed imagery
parallel angu- lar texture signature
template matching
Snakes algorithm
profile matching
作者简介
作者简介：林祥国（1981-），男，副研究员．博士，博士后，现从事激光雷达点云数据处理和信息提取、高分辨率遥感影像目标识别的研究。Email：linxiangguo@gmail．com
