基于广义SIFT特征驱动的LiDAR点云严密配准/平差模型研究

项目名称： 基于广义SIFT特征驱动的LiDAR点云严密配准/平差模型研究

项目编号： No.41271444

项目类型： 面上项目

立项/批准年度： 2013

项目学科： 天文学、地球科学

项目作者： 王永波

作者单位： 中国矿业大学

项目金额： 75万元

中文摘要： 针对基于同名点状特征匹配的LiDAR点云序列配准方法存在误差累积、难以实现数据无缝拼接的问题，本研究立足于尺度空间理论，探索点云的多尺度表达及其基于三维描述的地理实体表面SIFT特征直接提取方法，以点状、线状、面状SIFT特征作为配准基元，以同名SIFT特征间的距离与角度偏差作为配准误差的量化指标来建立相应的误差度量标准，构建LiDAR点云配准的严密模型；以配准之后同名特征间的偏差为零作为测站点定权的基本约束条件，研究配准误差约束下基于距离准则的测站点定权方法，并据此构建基于广义SIFT特征驱动的LiDAR点云严密平差模型，计算各测站坐标变换矩阵的改正数，进而实现对采样点坐标的改正。通过本研究，实现基于LiDAR点云的地理实体及其环境信息的全方位、高质量表达，达到构建整体一致的地理实体三维表面模型的目的，项目研究成果对推动LiDAR在空间三维数据采集中的深层次应用有着重要的理论与现实意义。

中文关键词： 三维激光扫描；三维空间相似变换；配准；四元数；Plücker坐标

英文摘要： Considering the error accumulation of classical point-like-feature based registration methods which makes it hard to seamless register point clouds from two adjacent stations, this research presents the multi-scale representation approach of terrestrial LiDAR point cloud and the approach to extract SIFT features from LiDAR point cloud. Similarity measure is constructed which is based on the difference between the same features from two adjacent stations, and the rigorous registration model is presented to put LiDAR point clouds from adjacent stations together. Generalized SIFT point-like features, linear features and planar features are selected as registration primitives, an approach to calculate the weight of each station is presented which considers the effect of both the distance between adjacent stations and the registration errors between same SIFT features from two different stations. Based on above work, a generalized SIFT feature constrained adjustment model for LiDAR point clouds is given to revise the transformation matrix of each station. And so the exact coordinates of each sample point can be calculated accurately. By this research, comprehensive and high quality representation of geo-spatial entities and their environments can be achieved with ease, topological consistent 3d surface model of geo

英文关键词： LiDAR；3D Similarity Transformation；Registration；Quaternions；Plücker Coordinates