Research Extract4D

Motivation

This project seeks to advance information that can be gained on surface dynamics by integrating multimodal 4D point clouds in time series-based change analysis for various data sources and types of surface activity.
This requires the generalization of a state-of-the-art 4D point cloud method, the extraction of 4D objects-by-change (4D-OBCs), which delineates individual surface activities in an observed scene in space and time.

Adaptive 4D-OBC Extraction for various Point Cloud Inputs

4D-OBC extraction will be extended to automatically adapt to different input point clouds (laser scanning vs. photogrammetry, variable temporal sampling and spatial scales, different types of surface dynamics), mainly by considering spatially and temporally variable uncertainties, and by accounting for irregular temporal sampling.

Based on this, different strategies of integrated change analysis are investigated, with fusion approaches at the data level of 3D time series and at the feature level of change information derived from each unimodal 4D point cloud dataset.

Advancing 4D Change Analysis for Environmental Research

The project makes use of virtual laser scanning of a variety of surface dynamics scenarios and investigates three use cases of real 4D point cloud data from different geographic settings. By this, the project optimizes automatic analysis and improves the results which can be obtained from 4D point clouds in topographic monitoring of complex natural scenes. This basic research renders methods of 4D change analysis transferrable to a variety of use cases in environmental research.

The observation of surface dynamics in natural landscapes provides valuable insights into Earth shaping processes, their complex interactions, and environmental drivers. 

The analysis of surface activities in 4D point cloud data has therefore become an integral part of topographic monitoring in environmental research. 

These data contain detailed 3D information of 

• the topography (i.e., the forms and features of the land surface)
• with time as an additional dimension

acquired by techniques like 

• permanent terrestrial laser scanning (TLS) or
• time-lapse 3D photogrammetry at (sub-)hourly intervals over months to years.

Recent methods of change analysis incorporate the time series information to identify and characterize surface activities with variable spatio-temporal properties. So far, these methods focus on data from unimodal acquisitions (e.g., TLS only) and assume a regular acquisition interval of near-continuous 3D observation systems. 

In practice, however, 4D point clouds are often adaptively sampled in time, both by design or due to external influences on regular acquisition schedules. 
Furthermore, supplementary acquisitions, such as multi-station TLS or UAV-based 3D sensing, are often performed to increase coverage and reduce uncertainty.