一种改进顶帽变换与LBP高程纹理的城区建筑物提取算法

发布时间：2018-01-29 02:52

本文关键词： LiDAR 数字航空影像 LBP 形态学建筑物提取　出处：《测绘学报》2017年09期 　论文类型：期刊论文

【摘要】：利用LiDAR数据的建筑物提取存在植被点与建筑物点难以区分的问题,利用航空影像进行城区建筑物提取则无法有效剔除阴影区域植被。本文融合LiDAR和航空影像两种数据源,提出了改进顶帽变换及局部二进制模式(LBP)高程纹理分析的建筑物提取算法。首先将LiDAR数据进行规则格网化,通过改进顶帽变换提取地面数据点,然后根据航空影像计算归一化差值植被指数(NDVI)值进行植被粗提取,计算LBP高程纹理,精细区分植被点与建筑物点,最后利用形态学操作填充建筑物孔洞,以检测出的建筑物点为种子点进行区域生长,得到完整的建筑物点集合。试验基于ISPRS提供的Vaihingen数据集中复杂多植被城区场景,试验结果表明,本文算法能够有效区分植被与建筑物,实现建筑物准确提取。
[Abstract]:It is difficult to distinguish vegetation points from building points in building extraction using LiDAR data. The use of aerial images to extract urban buildings can not effectively eliminate the shadow area vegetation. This paper combines LiDAR and aerial image two kinds of data sources. A building extraction algorithm based on the improved top hat transform and local binary pattern LBP height texture analysis is proposed. Firstly, the LiDAR data are netted by rules. The ground data points are extracted by the improved top cap transform, and then the normalized difference vegetation index (NDVI) value is calculated according to the aerial image, and the LBP elevation texture is calculated. Finally, the morphological operation was used to fill the holes in the building, and the detected building point was used as the seed point for regional growth. A complete set of building points is obtained. The experimental results show that the complex multi-vegetation urban area scenario is based on the Vaihingen data set provided by ISPRS. This algorithm can effectively distinguish vegetation from buildings and achieve accurate extraction of buildings.
【作者单位】：空军工程大学信息与导航学院;
【基金】：陕西省自然科学基金(2015JM6346)~~
【分类号】：P23
【正文快照】： 近年来,基于机载传感器数据的城区目标自动检测已经成为研究的热点,机载LiDAR(lightdetection and ranging)系统不易受天气、时段的影响,在地形测绘、环境监测及武器精确制导的军民事领域均得到了广泛的应用。建筑物是城市区域的重要目标,近几年,基于LiDAR系统的建筑物自动检

【相似文献】