利用合成算法从LiDAR数据提取屋顶面

发布时间：2018-04-26 17:29

本文选题：LiDAR + 屋顶面　；参考：《武汉大学学报(信息科学版)》2014年10期

【摘要】：区域增长法和随机抽样一致性(RANSAC)算法是从LiDAR数据提取屋顶面时常用的两类方法,但这两种方法都存在某些缺陷,使它们的应用受到了一定限制。针对LiDAR数据中建筑物脚点的特点,提出了一种融合以上两种方法优点于一体的合成算法。1根据脚点的法向量和粗糙度特征进行屋顶面粗提取;2在屋顶面粗提取结果的基础上,利用基于先验知识的局部采样策略和区域增长方式对传统随机抽样一致性算法进行扩展,实现屋顶面自动提取;3采用投票法解决屋顶面竞争问题,提高屋顶面的提取精度。实验结果表明,本文设计的合成算法能够有效地提取建筑物屋顶面。
[Abstract]:Regional growth method and random sampling conformance (RANSAC) algorithm are two kinds of methods commonly used to extract roof surface from LiDAR data. However, there are some defects in these two methods, which make their applications limited. In view of the characteristics of the foothold in the building of the LiDAR data, a combination of the advantages of the above two methods is proposed. The algorithm.1 extracts the roof surface according to the normal vector of the foot and the roughness feature. 2 on the basis of the rough extraction results of the roof surface, using the local sampling strategy and the regional growth mode based on the prior knowledge to expand the traditional random sampling consistency algorithm, the roof surface auto extraction is realized; 3 the roof competition is solved by voting method. The experimental results show that the synthetic algorithm designed in this paper can effectively extract the roof surface of buildings.

【作者单位】：武汉大学遥感信息工程学院;
【基金】：国家自然科学基金资助项目(61378078) 国家科技支撑计划资助项目(2012BAH34B02) 中央高校基本科研业务费专项基金资助项目(2012213020203,2012213020209)~~
【分类号】：TP391.41;TU746.3

【参考文献】