摘要
针对高分辨率影像中道路提取效果不佳这一问题,该文提出了一种道路段分割方法。在经典能量函数中引入无符号距离函数表示的形状先验,用以排除非道路干扰物;引入星形形状先验能量项用以解决细小干扰物、道路内部孔洞两个问题,此时可得到一段无干扰物、无孔洞的道路;基于动态外推的思路分割出一条完整的道路段。实验表明该方法可以有效获取影像中不同形态的道路段,同时为影像分割提供有益思路。
A road segmentation method was proposed aiming at the difficulty in extracting roads from high resolution images. Firstly, the shape prior described by the unsigned distance function was introduced into the classic energy function to eliminate non-road noises the star shape prior to solve noises and holes inside roads was introduced, a road without noises and holes would be obtaineda complete road could be gained based on dynamic growth. Experiments proved that the method was effective to obtain different types of roads and provided useful thoughts in image segmentation.
出处
《测绘科学》
CSCD
北大核心
2015年第6期102-105,124,共5页
Science of Surveying and Mapping
基金
国家自然科学基金项目(41271420)
江苏省测绘地理信息科研项目(JSCHKY201421)
关键词
道路分割
无符号距离函数
星形形状
动态外推
road segmentation
unsigned distance function
star shape
dynamic growth
