摘要
机载激光雷达扫描技术可以以点云形式快速获取地形表面高精度三维信息。基于激光雷达扫描数据及建筑物本身的拓扑信息就可以对建筑物进行精确的重建,而重建中最关键的技术是对点云数据进行分类,进而进行地物识别。对大规模三维点云数据进行快速分类,提出一种采用区域分割结合基于最小二乘平差的多项式拟合方法,将大量离散的三维点云分割后进行多项式拟合,并将二维数据分类转化为一维数据分类。在分类的基础上,将建筑物几何规则作为约束条件提取了房屋边缘。实验分析表明,该方法既能去除多余噪声,又能有效保留特征点,分类的总误差率低于3%。
Airborne laser scan technique is able to acquire the three dimensional geographic information of areas and objects on the ground quickly with the form of point clouds.Based on LiDAR data and topographic information of objects,the model of building can be determined.We classify the point clouds into terrain and off-terrain points based on region segment polynomial approximation and the least squares adjustment,and then extract edge of building with the method of RANSAC and topographic knowledge.The experiment result shows that this method can delete noisy points and preserve feature points efficiently which the total error rate is less than 3%.
出处
《吉林大学学报（地球科学版）》
EI
CAS
CSCD
北大核心
2010年第5期1205-1210,共6页
Journal of Jilin University：Earth Science Edition
基金
国家'863'项目(2006AA12Z102)
教育部新世纪优秀人才支持计划(NCET-07-0353)
国家自然科学基金项目(60873147)
高等学校博士学科点专项基金项目(20060183042)
关键词
地物识别
最小二乘逼近
点云数据
边缘提取
objects recognition
least squares approximations
point clouds
building edge extraction
