Advances in lidar point cloud processing

Abstract

LIDAR sensors and LIDAR systems utilized for precise surveying in various fields of application are operated from significantly distinct platforms ranging from static platforms during a single 3D scan acquisition in terrestrial or static laser scanning to a multitude of different platforms in kinematic laser scanning like mobile laser scanning, UAV-based laser scanning or airborne laser scanning. The related fields of application impose substantially different requirements with respect to accuracy, measurement rate, and data density. The results have to serve various data consumer communities and impose vastly dissimilar requirements on the LIDAR equipment, e.g., size, weight, cost and performance. Still, there are some general issues one has to address in data processing and delivery. In some cases, the emphasis lies specifically on rapid point cloud processing and delivery – although delivery time requirements may range from seconds up to weeks, depending on the application at hand. Processing requirements are demanding as in this paper we assume final point clouds to be clean – i.e. virtually noise free –, georeferenced, and consistent. We discuss general challenges in the data processing chain applicable to all types of LIDAR, regardless of the underlying technology, i.e. waveform LIDAR, discrete LIDAR, single-photon LIDAR, or Geiger-mode LIDAR. Applications include, e.g., rapid generation of data previews for the operator in kinematic LIDAR and the automated registration of all acquired point clouds in stop-and-go acquisition with static LIDAR.