LOKI - Airborne Assessment of Critical Infrastructures (Luftgestützte Observation Kritischer Infrastrukturen)
Automatic damage assessment based on UAV-derived 3D point clouds can provide fast information on the damage situation after an earthquake. However, the assessment of multiple damage grades is challenging due to the variety in damage patterns and limited transferability of existing methods to other geographic regions or data sources. We present a novel approach to automatically assess multi-class building damage from real-world multi-temporal point clouds using a machine learning model trained on virtual laser scanning (VLS) data. We (1) identify object-specific change features, (2) separate changed and unchanged building parts, (3) train a random forest machine learning model with VLS data based on object-specific change features, and (4) use the classifier to assess building damage in real-world point clouds from photogrammetry-based dense image matching (DIM). We evaluate classifiers trained on different input data with respect to their capacity to classify three damage grades (heavy, extreme, destruction) in pre- and post-event DIM point clouds of a real earthquake event. Our approach is transferable with respect to multi-source input point clouds used for training (VLS) and application (DIM) of the model. We further achieve geographic transferability of the model by training it on simulated data of geometric change which characterises relevant damage grades across different geographic regions. The model yields high multi-target classification accuracies (overall accuracy: 92.0% - 95.1%). Its performance improves only slightly when using real-world region-specific training data (< 3% higher overall accuracies) and when using real-world region-specific training data (< 2% higher overall accuracies). We consider our approach relevant for applications where timely information on the damage situation is required and sufficient real-world training data is not available.
Die schnelle Klassifizierung von beschädigten und zerstörten Infrastrukturen sowie die Beurteilung der Standsicherheit ist eine Grundvoraussetzung für effektive Rettungsmaßnahmen im Katastrophenfall. Für eine zügige Bewertung von Schäden sind visuelle Prüfungen unerlässlich. Zur einheitlichen und objektiven Beurteilung von Erdbebenschäden werden Schadenskataloge für globale und materialspezifische Schadensmerkmale entwickelt. In diesem Beitrag werden, anhand von fünf Schadensgraden, Schadenskataloge für tragende und nichttragende Bauteile aus Stahlbeton und Mauerwerk sowie globale Merkmale vorgeschlagen. Die enthaltenen Beschreibungen, schematischen Abbildungen und Beispielfotos machen eine objektive Kategorisierung der verschiedenen Schadensgrade möglich. Bei geringen Schadensgraden sind maximale Rissbreiten verschiedener Rissarten entscheidend zur eindeutigen Abgrenzung der Schadensklassen. Für hohe Schadensgrade ist das globale Tragverhalten des Gebäudes maßgebend. Die Anwendung der Schadenskataloge wird erläutert und eingehend diskutiert.
Reliable predictions and estimations of earthquake damage and seismic behaviour of buildings play an important role regarding seismic risk analysis. In this contribution, numerical damage criteria for five different damage grades are proposed that establish a correlation between observed damage patterns and the seismic response of the building in non-linear analyses. These criteria contain thresholds for material-specific as well as global characteristics to register all types of damage for reinforced concrete buildings. Via frequency and spatial distribution, these damage criteria with their defined limit values are assigned to the five damage grades. Interstorey-drift or spectral displacement values are derived from the relevant positions of the damage grades on the pushover curves. These are the basis for the explained procedure of developing fragility functions. The whole process is exemplarily illustrated for a four-storey reinforced concrete frame building. The positions of the damage grades on the pushover curve and the resulting fragility functions are presented. The results are discussed and validated with data from the literature.
For seismic risk analysis, reliable predictions and estimations of earthquake damage and seismic behaviour of buildings are essential. A common method is the use of fragility curves. In this paper, fragility functions are developed based on various numerical damage criteria for five defined damage grades, from slight to destruction. The proposed new multiscale approach establishes a correlation between observed damage patterns due to foreign earthquakes and the seismic response of the building using thresholds for material-specific and global characteristics. This approach takes into account various possible damage patterns on different scales more comprehensively than the well-known approach of displacement criteria. Moreover, the approach is universal and adaptable for building classes as well as region-specific material and system characteristics. Several damage criteria with their defined limit values are assigned to the five proposed damage grades, whereby quantity and distribution of the exceeded criteria are relevant, since the first occurrence does not always lead to damage. With the new approach, damages that are not evident in the pushover curve in terms of strength degradation can be detected and taken into account for the damage thresholds. The derived displacement values associated with the damage levels are the basis for developing fragility functions. The results—damage criteria, pushover curves with damage grades, capacity curves as well as fragility functions and parameters—are presented for a four-storey reinforced concrete frame building. These results are discussed and validated with data from the literature. Comparisons to existing fragility functions in the literature show that our developed fragility functions are mostly located in the middle range, graphically as well as for the curve parameters. This specific example was chosen to present our multiscale approach, but for general building classes, numerous simulations with varying characteristics are essential and result in a higher standard deviation of the final fragility curves.
The analysis and interpretation of 3D topographic change requires methods that achieve low uncertainties in change quantification. Many recent geoscientific studies that perform point cloud-based topographic change analysis have used the Multiscale Model to Model Cloud Comparison (M3C2) algorithm to consider the associated uncertainty. Change measured with the M3C2 approach, however, is difficult to interpret where (1) change occurs in directions different to the direction of change computation or (2) the quantified magnitudes of change are exceeded by the associated uncertainty due to a rough surface morphology. We present a correspondence-driven plane-based M3C2 approach that is tailored to quantifying small-magnitude ( 0.1 m) 3D topographic change of rough surfaces by reducing the uncertainty of quantified change. The approach (1) extracts planar surfaces in point clouds of successive epochs, (2) identifies corresponding planar surfaces between two point clouds using a binary random forest classification, and (3) calculates M3C2 distances and the associated uncertainty between the corresponding planar surfaces. This correspondence-driven plane-based M3C2 does not require recognition or reconstruction of geometrically complex objects but instead quantifies change between less complex, homologous planar surfaces. The approach further allows to relate change directly to a moving object. We apply our approach to a bi-weekly time series of terrestrial laser scanning point clouds acquired at a rock glacier in the Austrian Alps. The approach enables a sevenfold reduction in the uncertainty associated with topographic change compared to standard M3C2. Significant change is therefore detected in 72.62% to 76.41% of the area of change analysis, whereas standard M3C2 detects significant change in only 16.21% (2-week timespan) to 59.96% (10-week timespan) of the same area. The correspondence-driven plane-based M3C2 complements 3D change analysis in applications that aim to quantify small-magnitude topographic change in photogrammetric or laser scanning point clouds with low uncertainties in natural scenes which are characterised by overall rough surface morphology and by individual rigid objects with planar surfaces (e.g., rock glaciers, landslides, debris covered glaciers).
Topographic laser scanning is a remote sensing method to create detailed 3D point cloud representations of the Earth's surface. Since data acquisition is expensive, simulations can complement real data given certain premises are met: (i) models of 3D scene and scanner are available and (ii) modelling of the beam-scene interaction is simplified to a computationally feasible while physically realistic level. A number of laser scanning simulators for different purposes exist, which we enrich by presenting HELIOS++. HELIOS++ is an open-source simulation framework for terrestrial static, mobile, UAV-based and airborne laser scanning implemented in C++. The HELIOS++ concept provides a flexible solution for the trade-off between physical accuracy (realism) and computational complexity (runtime, memory footprint), as well as ease of use and of configuration. Features of HELIOS++ include the availability of Python bindings (pyhelios) for controlling simulations, and a range of model types for 3D scene representation. Such model types include meshes, digital terrain models, point clouds and partially transmissive voxels, which are especially useful in laser scanning simulations of vegetation. In a scene, object models of different types can be combined, so that representations spanning multiple spatial scales in different resolutions and levels of detail are possible. HELIOS++ follows a modular design, where the core components of platform, scene, and scanner can be individually interchanged, and easily configured. HELIOS++ further allows the simulation of beam divergence using a subsampling strategy, and is able to create full-waveform outputs as a basis for detailed analysis. We show how HELIOS++ positions among other VLS software in terms of input model support and simulation of beam divergence in a literature survey. We also perform a direct comparison of simulations with DART, where we employ a scene from the Radiative Transfer Model Intercomparison (RAMI). This example shows that HELIOS++ takes about 10 times longer than DART for parsing and preparing the 3D scene, but performs about 314,000 times faster in the beam simulation, achieving 200,000 rays/s. Comparing HELIOS++ to its predecessor, HELIOS, revealed reduced runtimes by up to 99%. Virtually scanned point clouds may be used for a broad range of applications as shown in literature. We could identify four main categories of use cases prevailing at present, which benefit from simulated LiDAR point clouds: data acquisition planning, method evaluation, method training and sensing experimentation. We conclude that a general-purpose LiDAR simulator can be employed for many different scientific applications, as long as it is ensured that the simulation adequately represents reality, which is specific to the given research question.
p>In the event of an earthquake, damaged and destroyed buildings are of central importance. Using a combination of automatic approaches and human crowdsourced visual interpretation based on unmanned aerial vehicle (UAV) derived data for the classification of earthquake damage offers a fast and objective assessment of the damage situation. Earthquake engineering knowledge is transferred to these innovative methods by developing and implementing a damage catalogue. This damage catalogue includes typical damage patterns for five damage grades ranging from crack widths to failure modes and focuses on the two common building materials - reinforced concrete and masonry. This paper presents the structure of such damage catalogue, defines crack widths and gives examples for particular damage grades. Moreover, the application of the damage catalogue in automatic and crowdsourcing approaches for a classification into five damage grades is explained.</p