[an error occurred while processing this directive] [an error occurred while processing this directive]

Institute of Geography > 3D Geospatial Data Processing Group >

Research

Methods & Algorithms

Virtual laser scanning is a tool to create simulated point cloud data as would be acquired by a LiDAR sensor. In our open source project HELIOS++ we develop the method of VLS further to provide e.g. training data for machine learning, survey planning, ground truth creation, education, etc.

Virtual Laser Scanning for Machine Learning

We investigate the close coupling of learning algorithms with virtual laser scanning and real point cloud data to use benefits of both 1) the realism of real − but sparse − training data and 2) the multitude of options of object-sensor interactions that can be generated with VLS.

4D Objects-By-Change (4D-OBC)

Time series-based 3D change analysis of natural surfaces enables to extract surface changes forming objects-by-change in their varying spatial and temporal extents from large and dense 4D geospatial point clouds.

M3C2-EP

M3C2-EP presents how error propagation for 3D change detection can be used to accurately quantify the minimum detectable change, the so-called Level of Detection. Measurement and processing uncertainties are included and the concept can be extended to more complex error models.

Multi-Directional Change

Methods of multi-directional change quantification provide an alternative view of movement where change processes operate in a direction that is not surface-normal, the underlying process(es) may not be known and movement(s) across the point cloud scene is not oriented along a single axis.