摘要
针对机载LiDAR建筑物点云提取过程中与树木紧邻的建筑物难以提取,已有先滤波后提取算法效率低等问题,提出一种结合区域生长与主成分分析的机载LiDAR建筑物点云提取算法。该算法首先对粗差剔除后的机载LiDAR离散点云构建TIN三角网,依据建筑物边缘点所在三角形的特征提取建筑物边缘点;然后将邻域特征优化后的建筑物边缘点作为种子点进行区域生长得到建筑物点云;最后采用主成分分析对提取结果进行检核,剔除非建筑物点云,在此基础上基于连通性对建筑物点云进行单体化分割,剔除小面积区域,得到最终的建筑物激光脚点数据。实验选取国际摄影测量与遥感协会提供的三组典型区域的LiDAR点云数据进行建筑物提取,并与传统形态学和区域生长两种建筑物点云提取算法进行比较,结果表明本文算法可以实现建筑物点云的高精度提取,且对地形及不同类型屋顶的建筑物具有良好的自适应性,验证了算法的可靠性。
For the airborne LiDAR buildings point clouds extraction process,it is difficult to extract buildings points adjacent to the trees,and extraction efficiency of the algorithm after filtering is inefficiency.Aiming at those problems,this paper propose an airborns LiDAR buildings points extraction method based on region growing and principal component analysis.Firstly,the proposed altithom uses airborne LiDAR discrete points which are eliminated gross error to construct the TIN triangulation.After that extracting buildings boundary points according to the feature of the boundary triangle.Secondly,the buildings point clouds would be obtained by region growing with seed points which were from buildings boundary points optimized by neighborhood spatial feature.Finally,the extraction results are checked by principal component analysis algorithm,meanwhile the non-buildings point clouds would be filtered out.Point clouds regions are segmented based on building connectivity and small regions are removed,the final buildings laser foot data is obtained.Three typical LiDAR point clouds data,which are provided by the International Society for Photogrammetry and Remote Sensing,are selected for buildings extraction experiments.By comparing with the traditional morphology and region growing point clouds extraction methods,the test results show that the proposed algorithm can achieve high precision buildings point clouds extraction and have good adaptability to terrain and buildings of different roof types.The reliability of the algorithm is verified.
作者
王竞雪
洪绍轩
WANG Jing-xue;HONG Shao-xuan(School of Geomatics,Liaoning Technology University,Fuxin,Liaoning 123000,China;Faculty of Geosciences and Environmental Engineering,Southwest Jiaotong University,Chengdu,Sichuan 610031,China)
出处
《信号处理》
CSCD
北大核心
2018年第9期1094-1104,共11页
Journal of Signal Processing
基金
国家自然科学基金青年基金项目(41101452)
辽宁省教育厅科学研究一般项目(LJYL010)
关键词
机载LIDAR
建筑物提取
区域生长
主成分分析
aerial LiDAR
buildings extraction
region growing
principal component analysis
