节点文献

物探飞行模式下的机载LiDAR数据与影像配准研究

Airborne LiDAR Data and Image Registration in Geophysical Prospecting Mode

【作者】 张靖

【导师】 江万寿;

【作者基本信息】 武汉大学 , 摄影测量与遥感, 2011, 博士

【摘要】 航空地球物理勘查技术(简称航空物探)是一种重要的地球物理勘查技术,与传统的地面探矿方法相比,具有速度快、效率高、使用劳力少,能在短期内取得大面积区域的探测资料等优点,已成为矿产资源普查的重要技术手段,为国民经济发展做出了重大贡献。航空物探侧重于对各种地球物理场信息的探测,缺乏准确的地表几何信息和丰富的纹理信息,将航空遥感与物探集成能够在获得物探数据的同时,获取现势性的影像地图、精确的空间位置信息以及实时的姿态信息,为物探数据的重力地形改正、人文异常剔除、重/磁参数的定位等提供重要参考,对提高物探数据解译的精度和效率具有重要意义。由于航空物探飞行方式与航空遥感存在很大差异,物探飞行模式为航空遥感数据处理,特别是航空影像数据与LiDAR数据的配准带来了很大困难,本文在深入分析物探飞行模式的特点后,提出了一套航空影像与LiDAR点云配准的自动化方案:首先对LiDAR系统进行检校以获得系统误差改正参数,然后利用该参数对LiDAR点云进行系统误差纠正,最后对航空影像两两之间构建立体像对,并进行立体影像密集匹配,将匹配点云与LiDAR点云进行ICP配准,利用配准结果解算像机与激光扫描仪之间的偏心角和偏心分量。本文的主要工作和创新点总结如下:(1)推导了机载LiDAR定位严格方程本文详细阐述了LiDAR系统组成,以及各部分工作原理,介绍了LiDAR系统中所采用的坐标系以及坐标转换方法,并在此基础上推导了LiDAR严格定位方程,并对LiDAR系统中的各类系统误差进行了分析。(2)提出了一种基于虚拟连接点模型的LiDAR系统误差全自动检校方法综合分析了现有的LiDAR系统误差检校方法,实现了一种基于共面约束的LiDAR系统检校。由于同名平面的自动提取与匹配比较困难,本文提出了一种虚拟连接点模型,解决了离散激光点云之间的对应性问题,并在此基础上设计了一种基于虚拟连接点模型的LiDAR系统误差全自动检校方法。(3)实现了一种强化拓扑约束和最小二乘传播的立体影像匹配方法对现有的立体匹配算法进行了综合分析,并详细讨论了立体匹配中常用的各种约束条件,实现了一种融合多种约束条件的立体匹配算法,在匹配过程中强化了硬性的拓扑约束,用来获取可靠的匹配种子,并利用最小二乘匹配参数的传播来克服地形起伏的影响,提高匹配的可靠性。(4)设计了一种立体像对与LiDAR点云的配准方法采用ICP算法对影像匹配点云与LiDAR点云进行配准,针对匹配点云与LiDAR点云之间的差异,设计了一套点云预处理方案分别对匹配点云和LiDAR点云进行预处理,以提高ICP算法中对应点确定的精度。(5)设计了一种海量离散点云的四叉树索引方法由于在ICP算法迭代过程中需要反复确定最近点,涉及到大量的数据查找和定位操作,由于LiDAR点云是非结构化的离散点,在海量LiDAR点云中进行随机查找效率极低,本文设计了一种离散点云的四叉树索引方法,有效的提高了最近点查找效率。

【Abstract】 Aerial geophysical prospecting (AGP) is an important method in the field of geophysical prospecting. Comparing with the traditional methods, AGP is much faster and more efficient with less labor force. It has become an important means of mineral resources exploration, making a significant contribution to national economic development. AGP aims at acquiring the geophysical information other than the topographical information and texture information of the ground surface, it motivates the idea of integrating aerial remote sensing into AGP. In this way, we can also get the corresponding aerial images, accurate height information and real-time attitude parameters which can serve as a reference for gravity terrain correction, geophysical anomaly elimination and gravity/magnetic parameters locating, etc. The enhancement of the accuracy and efficiency indicates the promising application prospect of this integrated system.Due to the big difference between the flight mode in AGP and traditional aerial remote sensing, however, the data processing becomes a big problem, especially the registration of aerial images and LiDAR points. In this paper, an automatic registration plan is introduced to solve this problem, considering the characteristics of AGP. First, it uses the correction parameters calculated by LiDAR system calibration to rectify the LiDAR point cloud. Second, it performs dense matching in each stereo pair, and carries out the registration procedure using ICP algorithm with the resulting matching point cloud and the LiDAR point cloud. Finally, it gives out the eccentric angle and eccentric component between the camera and the laser scanner. The major work and innovation are summarized as follows:(1) Rigorous equation of airborne LiDAR positioning is derived.This paper describes the components of LiDAR system as well as their working principle. It also presents the coordinate systems and coordinates transformation methods used in LiDAR system. On this basis, rigorous equation of airborne LiDAR positioning is given, and the system errors are analyzed.(2) An automatic LiDAR system error calibration method based on virtual connection point model is proposed.Comprehensive analysis of the existing LiDAR system error calibration methods is given, to achieve a coplanar-constraint-based LiDAR system calibration. As the corresponding planes are difficult for extraction and matching automatically, this paper presents a virtual connection point model, solving the correspondence problem between the discrete laser point clouds. Based on this model, a design is proposed of automatic LiDAR system error calibration.(3) A topological constraints strengthened stereo matching method based on least squares propagation is proposed.This paper discusses the existing stereo matching algorithms and the various constraints commonly used in stereo matching, then gives out a multi-constraint stereo matching algorithm. This algorithm strengthens rigid topological constraints to get the reliable matching seeds that are used for least squares propagation to overcome the effects of topography, improving the reliability of stereo matching.(4) designed an effective method for registration of image matching points with LiDAR pointsICP algorithm is selected for registration of image matching points and LiDAR points. Taken into account the differences between the two point clouds, this paper presents a preprocessing program for image matching points and LiDAR points respectively, which greatly enhances the accuracy of the corresponding points under ICP framework.(5) designed a quadtree indexing method for massive discrete point cloudSince the ICP algorithm requires repeated iterative process to determine the nearest point, it involves large amounts of data search and location operations. As LiDAR point cloud is unstructured, the efficiency is very low of random search in massive LiDAR point cloud. The proposed quadtree indexing method effectively raises the efficiency to find the nearest point.(6) proposed a registration model of stereo pair and LiDAR point cloud based on camera eccentric angle and eccentric componentAs camera and laser scanner are tightly connected in Leica ALS50 system, sharing one set of POS data, the major deviation between images and LiDAR data can be seen as the eccentric angle and eccentric component between camera and laser scanner. This paper presents a registration model based on camera eccentric angle and eccentric component. This model reduces the number of unknowns, so that all the stereo pairs can be rectified using a unified system error correction.

  • 【网络出版投稿人】 武汉大学
  • 【网络出版年期】2012年 07期
节点文献中: 

本文链接的文献网络图示:

本文的引文网络