lass是什么意思云平台

转载

mob64ca13f87273 2024-12-18 10:28:33

文章标签 lass是什么意思云平台激光雷达数据 sed 文章分类 云平台云计算

lass是什么意思云平台_数据

本文为美国俄勒冈州立大学（作者：Erzhuo Che）的博士论文，共186页。

激光雷达（LIght Detection And Ranging，简称Lidar）是一种利用脉冲激光形式的光束进行遥感的技术，它能够高效、准确地获取场景的三维数据。根据安装平台的不同，激光雷达数据采集可分为机载激光扫描（ALS或机载激光雷达）、地面激光扫描（TLS或地面激光雷达）和移动激光扫描（MLS或移动激光雷达）。激光雷达技术已广泛应用于地形测绘、水深测绘、公用事业测绘、工程测量、农业、林业、地质、建筑、工业设施、文化遗产、资产管理、建筑等领域。然而，在大数据量的情况下，有效地处理激光雷达产生的密集数据集仍然是一个挑战。此外，由于扫描模式、范围、视角和其他因素，地面和移动激光雷达数据的点密度在整个场景中可能会有很大的变化，这就给开发健壮的处理方法带来了不同的挑战，而ALS点云往往分布更均匀。为了克服TLS和MLS数据处理中的困难，本研究将点云组织成二维网格结构，称为扫描模式网格，它代表了扫描仪收集数据的方式。

本文共分四章，探讨了利用扫描模式网格处理点云数据的可能性和性能改进。本文通过扫描线密度分析，提出了一种有效的TLS数据地面滤波方法。地面滤波是激光雷达数据处理中常用的一种方法，它将点云数据分为地面点和非地面点两类。该方法首先根据对每个扫描线内点密度的分析，分离出候选地面、密度特征和未识别点。其次，使用扫描模式的区域生长使用密度特征作为边界对候选地面进行聚类，并进一步细化地面点。这两个阶段分别处理和分析每次扫描中的TLS数据，利用扫描模式网格提高效率。接下来的两章开发了一种新的点云分割方法，该方法在分析期间将多个扫描的扫描模式网格连接起来。点云分割将具有与几何、色度辐射和/或其他信息相似属性的点分组，以帮助提取和解释点云。提出的分割方法只需要基本的几何信息，主要包括两个步骤。首先提出了一种新的特征提取方法：正态变异分析（Norvana）消除了一些噪声点，提取了边缘点，而不需要对每个点进行一般的（且容易出错的）正态估计。其次，区域生长利用边缘点作为边界将光滑曲面上的点进行分组，从而得到分割结果。

与可以直接从结构化格式（例如，ASTM E57）构造的TLS数据不同，移动激光数据通常以无组织方式（例如，ASPRS-LAS）存储。最后一章提出了一个有效的移动激光雷达数据处理框架，其中包括一种重建扫描仪轨迹的方法，以便根据扫描仪的采集顺序和转数将无组织点云构造成扫描模式网格。然后，将Norvana的边缘检测、正态估计、特征提取和分割等概念扩展到适用于移动激光数据的处理，并命名为Monorvana。此外，该框架还引入了一种有效的基于扫描模式网格的数据可视化方案。所有这些方法都实现了并行处理以获得更高的计算性能。通过对不同空间尺度、分辨率和场景类型的扫描仪采集的多个地面和移动激光雷达数据集进行测试，定性和定量地证明了该方法的有效性、效率、鲁棒性和通用性。这项研究的主要贡献是一个通用的点云处理框架，它可以有效地支持各种应用程序的各种细化、处理和分析。

Lidar (LIght Detection And Ranging) is aremote sensing technology using light in the form of a pulsed laser, whichenables efficient, accurate, 3-D data acquisition of a scene. Depending on themounting platform, lidar data acquisition can be categorized into AirborneLaser Scanning (ALS, or airborne lidar), Terrestrial Laser Scanning (TLS, orterrestrial lidar), and Mobile Laser Scanning (MLS, or mobile lidar). The lidartechnique has been widely used for a plethora of applications includingtopographic mapping, bathymetric mapping, utility mapping, engineeringsurveying, agriculture, forestry, geology, architecture, industrial facilities,cultural heritage, asset management, construction, and so forth. However,efficiently processing the dense datasets produced by lidar still remainschallenging given the large data volume. In addition, because of the scanpattern, range, view angle, and other factors, the point density forterrestrial and mobile lidar data can vary dramatically across the scene, whichraises different challenges in developing robust processing methods compared withan ALS point cloud, which tends to be more evenly distributed. To overcome thechallenges in processing TLS and MLS data, in this research, the point cloud isstructured into a 2-D grid structure called the scan pattern grid, whichrepresents the way that a scanner collects data. This dissertation, comprisingfour manuscripts, explores the possibilities and performance improvements ofexploiting this scan pattern grid to process point cloud data. This firstmanuscript presents an efficient ground filtering method for TLS data via aScanline Density Analysis. Ground filtering is a common procedure in lidar dataprocessing, which separates the point cloud data into two classes: groundpoints and non-ground points. The proposed process first separates the groundcandidates, density features, and unidentified points based on an analysis ofpoint density within each scanline. Second, a region growth using the scanpattern clusters the ground candidates using the density features as boundariesand further refines the ground points. Both stages process and analyze the TLSdata in each scan separately, exploiting the scan pattern grid for efficiency.The next two manuscripts develop a novel point cloud segmentation with anapproach that links the scan pattern grids from multiple scans during theanalysis. Point cloud segmentation groups points with similar attributes withrespect to geometric, colormetric radiometric, and/or other information to helpwith object extraction and interpreting the point cloud. The proposedsegmentation method only requires the basic geometric information and consistsof two main steps. First, a novel feature extraction approach, NORmal VAriationANAlysis (Norvana), eliminates some noise points and extracts edge pointswithout requiring a general (and error prone) normal estimation at each point.Second, region growing groups the points on a smooth surface using the edgepoints as boundaries to obtain the segmentation result.

Unlike TLS data that can be directlystructured from a structured format (e.g., ASTM E57), Mobile lidar data isusually stored in an unorganized manner (e.g., ASPRS LAS). The final manuscriptpresents an efficient mobile lidar data processing framework including anapproach to reconstruct the scanner trajectory such that an unorganized pointcloud can be structured into the scan pattern grid based on the order ofacquisition and revolutions of the scanner. Then the concept of Norvana foredge detection, normal estimation, feature extraction, and segmentation, isextended to be suitable for processing mobile lidar data and is namedMo-norvana. Additionally, the proposed framework also introduces an efficientdata visualization scheme exploiting the scan pattern grid. All of the proposedmethods implement parallel processing to obtain a higher computationalperformance. The effectiveness, efficiency, robustness, and versatility aredemonstrated both qualitatively and quantitatively by testing multipleterrestrial and mobile lidar datasets collected by different scanners withdifferent spatial scales, resolutions, and scene types. The key contribution ofthis research is a generalized point cloud processing framework that canefficiently support a wide range of refinements, processes, and analysis for avariety of applications.