LIDAR Automatic Feature Detection and Object Classification

激光雷达（光探测和测距）是一种相对较新的技术，用于快速准确地扫描地球表面。由此产生的几何模型可以应用于许多领域的关键应用，包括城市规划、管理和可视化；安全;更新和维护地籍（财产线）数据；景观变化检测；虚拟现实;以及行动任务规划。在大多数相关应用中，地理空间特征（在此简称为特征）或诸如建筑物、植被、道路和水道等物体的描绘和识别是主要的兴趣。由于传统的人工提取方法成本高、耗时长，而且大量的激光雷达数据来自不同的来源，因此对自动激光雷达特征提取的需求越来越大[Cary 2009]。然而，必须指出的是，在工业环境中发现的激光雷达数据集的规模（可能有数十亿个点）带来了巨大的性能挑战，甚至超出了最优雅的已发表研究技术的范围。

LiDAR (LIght Detection And Ranging) is a relatively new technology for scanning the earth's surface quickly and accurately. The resulting geometry models can be applied to crucial applications in many fields, including urban planning, management, and visualization; security; updating and maintaining cadastral (property line) data; landscape change detection; virtual reality; and operations mission planning. In most relevant applications, the delineation and identification of geospatial features (herein referred to simply as features), or objects such as buildings, vegetation, roadways and waterways, are of primary interest. Since traditional manual extraction methods are costly and time consuming, and large quantities of LiDAR data are being gathered from different sources, there is a growing demand for automatic LiDAR feature extraction [Cary 2009]. However, it must be noted that the scale of LiDAR datasets (potentially billions of points) found in industrial settings introduces formidable performance challenges, beyond the scope of even the most elegant published research techniques.