"空-车"LiDAR点云数据一体化的高质量自动集成方法研究

The objective of the undergoing project supported by NSFC is automatic three-dimensional reconstruction, in a large area, with good details and high geometric accuracy. Based on this, the new objective of this proposal is to obtain good details and realistic view of the façade sides of buildings or objects in automatic three-dimensional reconstruction. With the development of Light Detection And Ranging (LiDAR) technology, the family of such applications has grown to include airborne LiDAR, mobile LiDAR, terrestrial LiDAR, and indoor LiDAR. The performance, as well as application scope, of different platforms is variable, although they generally complement one another. Along with the progression of LiDAR technology, the integration of multi-platform LiDAR is an evident trend. Airborne LiDAR is quickly able to obtain three-dimensional information about objects over a large area. Despite abundant top-surface information, the corresponding facade information is lacking. Mobile LiDAR can quickly obtain detailed facade information, with high geometric accuracy, but obtaining top-surface information is difficult. Because airborne LiDAR and mobile LiDAR have their own advantages and disadvantages, the integration of these two technologies will collect a full set of information about object surfaces, both on the top and facade sides. This proposal is to research the mode to integrate airborne LiDAR data and mobile LiDAR data. A technical framework with high reliability, high geometric accuracy and high automation level will be presented, which includes four steps. First, airborne LiDAR data is used to extract road networks and building boundaries. Mobile LiDAR data is then used to extract building boundaries. The trajectory data of mobile LiDAR system will also be used. Thirdly, airborne and mobile LiDAR data is initially matched by using the road network (extracted from airborne LiDAR data) and the mobile trajectory data. Finally, based on the initial matching results, airborne and mobile LiDAR data is registrated, by using two sets of building boundaries extracted from two platform LiDAR data, respectively. The expected data product of this proposal is the integrated LiDAR data with a full set of information about object surfaces, covering the top and facade sides. This product will be very useful for three-dimensional reconstruction with good details, especially in façade sides. In addition, the methods and technologies of this study will be helpful for multi-platform joint of LiDAR technology and for three-dimensional modeling of smart city.

在青年科学基金项目—“大区域自动精细三维重建”研究的基础上，提出更高的目标—提升大区域自动三维重建中侧面微观细节恢复能力和全方位真实感保真能力。拟创新探索航空平台与车载平台（“空-车”）LiDAR技术的复合使用模式，立足于“空-车”平台“俯视-侧视”的角度互补优势，以“空-车”LiDAR点云数据一体化为目标，探讨特征引导下“空-车”LiDAR点云数据集成的研究框架，突破研究中的海量离散点云数据特征感知与提取（提取难）、异质平台点云数据自动匹配（匹配难）、不规则点云数据集成效果定量评价（评价难）等难点与关键点，提出一套新型“空-车”LiDAR数据高质量（高可靠性、高精度、高自动化）集成方法，生成“俯视-侧视”多角度联合的全方位三维精细点云数据，为精细三维重建提供新数据新思路，为星空车地多平台LiDAR技术的综合使用开展技术前瞻探索，为智慧城市中的空间信息基础设施建设等典型应用提供新途径。

围绕“空-车”LiDAR点云数据一体化集成目标，以集成多平台LiDAR点云数据的“特征提取—点云匹配—三维建模—三维评价—三维应用”为框架，提出了面向多平台点云数据集成利用的系列关键技术和算法，解决了点云技术研究中海量离散点云数据特征感知与提取难、异质平台点云数据自动匹配难、不规则点云数据集成效果定量评价难等问题。.（1）特征提取：多平台点云信息提取技术。围绕特征引导下“空-车-地”LiDAR点云数据集成的目标，充分挖掘地物特征信息提取在不同平台点云数据中的优势，优化了地物特征（建筑物、道路网、树木等）信息提取的质量。.（2）点云匹配：“空-车-地”多平台点云集成技术。针对不同平台点云数据匹配存在的困难，在上述地物特征提取基础上，充分利用地物轮廓、角点、可移动引导点、对称轴等参照物，克服了不同平台点云数据的异质性、离散型等难题。.（3）三维建模：集成多平台点云数据的三维建模。综合利用多平台点云数据，提出了基于多尺度格网的建筑物屋顶模型重建、基于结构单元的建筑物立面模型重建、基于连通性的立交桥模型重建、基于聚类生长的电力线提取等算法，实现了多种地物的高质量、自动化的精细三维重建。.（4）三维评价：综合视觉和几何质量的三维模型评价。针对三维建模的结果，提出了一套三维模型质量评价体系，克服了传统评价中主观因素对评定结果的影响，综合视觉和几何质量进行三维模型评价。.（5）三维应用：多平台点云集成与建模成果应用。在三维建模及其质量评价的基础上，开展了城市建筑和树木三维变化检测方法、三维场景路径导航技术、天然气管网爆燃三维分析方法研究，从三个角度挖掘点云数据的应用潜能。.在项目支持下，获得成果：发表SCI/SSCI/A&HCI检索论文31篇（第一标注27篇，第二标注4篇；第一作者18篇，通讯作者9篇）。授权发明专利4项，登记软件著作权2项。担任IEEE GRSL副主编。获国家科技进步奖二等奖（2014，5/10）、国家级教学成果奖二等奖（2014，10/15）、全国青年地理科技奖（2015）、国家优青（2016）、教育部长江学者青年学者（2016）、首批全国高校黄大年式教师团队成员（2017）。