摘要
提出一种基于LBP算子与多尺度分析手段的高分辨率遥感影像道路提取方法。首先使用邻域均值作为阈值进行二值化,以改进一般LBP算子;再根据局部方差与尺度间的关系,找到适合道路LBP纹理值的最佳尺度;最后利用区域生长和形状指数提取最佳尺度下的影像道路信息,并通过尺度收缩关联到原始影像。实验结果表明本文方法可取得较好效果。
This paper presents a new approach for road extraction from high resolution imagery based on local binary pattern（LBP） operator and multi-scale analysis method.First,use local mean as threshold to binarization for improving the basic LBP.Next,choose the optimal scale to obtain road texture using LBP according to the relation between local variance and scales.Finally,extract the road from image at this scale by using region growing and shape index,then get the road form original image by scale shrinking.Experiment result indicates the new method is effective.
出处
《遥感信息》
CSCD
2010年第2期3-7,11,共6页
Remote Sensing Information
基金
湖南省自然科学基金(07JJ3060)
