Overview

A key activity during the post-disaster response and reconstruction phases is the Structural Damage Assessment (SDA). Recent developments in photogrammetry and computer vision allow for very dense models in 3D points from aerial multi-perspective images to be overlapped. Both 3D point clouds and MDIs are rich sources of damage mapping information. Unmanned aerial vehicles (UAVs) usually produce multi-perspect aerial imagery. However, in crisis situations, these data can be scarce. An interesting alternative are aerial video footage after a catastrophe, but video frames are not good enough with low resolution or redundancy.

There is still a lack of exploreability of 3D modelling video frames, particularly for post-disaster applications. In this research, the quality of aerial 3D videomodels was evaluated from the perspective of the geometric/absolute and SDA-application and compared with aerial still-imaging models for two different fields of study. Particular influence on 3D model quality was identified on the video blur motions, the resolution and the frame redundancy.

The analysis has shown that video information generally produces more noisy and imprecise 3D point clouds; the external and absolute precision is still similar to imagery. The sensor closeness to the ground clearly impaired the low resolution video, while the major cause of noise was frame redundancy. However, these quality parameters were also linked to higher points and to improved damage representativeness in most cases. Video data have thus been shown to be adequate for the production of rich 3D information models for damages.