Object‐based classification of terrestrial laser scanning point clouds for landslide monitoring

Mayr, Andreas; Rutzinger, Martin; Bremer, Magnus; Elberink, Sander Oude; Stumpf, Felix; Geitner, Clemens

doi:10.1111/phor.12215

Cited by 51 publications

(32 citation statements)

References 48 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As the presented classification of change objects shows, spectral features can be used to support this task. Alternatively or complementarily, geometric features could be exploited to distinguish various natural objects in an automated time series analysis (Mayr et al, 2017). Figure 7.…”

Section: Land Cover Classmentioning

confidence: 99%

Unmanned Aerial Vehicle Laser Scanning for Erosion Monitoring in Alpine Grassland

Mayr

Bremer

Rutzinger

et al. 2019

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

Abstract. With this contribution we assess the potential of unmanned aerial vehicle (UAV) based laser scanning for monitoring shallow erosion in Alpine grassland. A 3D point cloud has been acquired by unmanned aerial vehicle laser scanning (ULS) at a test site in the subalpine/alpine elevation zone of the Dolomites (South Tyrol, Italy). To assess its accuracy, this point cloud is compared with (i) differential global navigation satellite system (GNSS) reference measurements and (ii) a terrestrial laser scanning (TLS) point cloud. The ULS point cloud and an airborne laser scanning (ALS) point cloud are rasterized into digital surface models (DSMs) and, as a proof-of-concept for erosion quantification, we calculate the elevation difference between the ULS DSM from 2018 and the ALS DSM from 2010. For contiguous spatial objects of elevation change, the volumetric difference is calculated and a land cover class (bare earth, grassland, trees), derived from the ULS reflectance and RGB colour, is assigned to each change object. In this test, the accuracy and density of the ALS point cloud is mainly limiting the detection of geomorphological changes. Nevertheless, the plausibility of the results is confirmed by geomorphological interpretation and documentation in the field. A total eroded volume of 672&thinsp;m3 is estimated for the test site (48&thinsp;ha). Such volumetric estimates of erosion over multiple years are a key information for improving sustainable soil management. Based on this proof-of-concept and the accuracy analysis, we conclude that repeated ULS campaigns are a well-suited tool for erosion monitoring in Alpine grassland.

show abstract

Section: Land Cover Classmentioning

confidence: 99%

Unmanned Aerial Vehicle Laser Scanning for Erosion Monitoring in Alpine Grassland

Mayr

Bremer

Rutzinger

et al. 2019

ISPRS Ann. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…The presented work builds upon an automated landslide classification pipeline for 3D point clouds as a basis for object monitoring (Mayr et al, 2017). From this as a starting point our contribution is now progressing towards the identification and characterisation of landslide changes within the entire time series of classified point clouds by 4D object-based analysis.…”

Section: Workflow and Methodsmentioning

confidence: 99%

“…These rules reclassify and correct for instance misclassified 'erosion' segments outside the two landslides. A detailed description of the point cloud classification methods is provided in Mayr et al (2017). …”

Section: Point Cloud Classificationmentioning

confidence: 99%

See 1 more Smart Citation

Multitemporal Analysis of Objects in 3d Point Clouds for Landslide Monitoring

Mayr¹,

Rutzinger²,

Geitner³

2018

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:To date multi-temporal 3D point clouds from close-range sensing are used for landslide and erosion monitoring in an operational manner. Morphological changes are typically derived by calculating distances between points from different acquisition epochs. The identification of the underlying processes resulting in surface changes, however, is often challenging, for example due to the complex surface structures and influences from seasonal vegetation dynamics. We present an approach for object-based 3D landslide monitoring based on topographic LiDAR point cloud time series separating specific surface change types automatically. The workflow removes vegetation and relates surface changes derived from a point cloud time series directly to (i) geomorphological object classes (landslide scarp, eroded area, deposit) and (ii) to individual, spatially contiguous objects (such as parts of the landslide scarp and clods of material moving in the landslide). We apply this approach to a time series of nine point cloud epochs from a slope affected by two shallow landslides. A parameter test addresses the influence of the registration error and the associated level of detection on the magnitude of derived object changes. The results of our case study are in accordance with field observations at the test site as well as conceptual landslide models, where retrogressive erosion of the scarp and downslope movement of the sliding mass are major principles of secondary landslide development. We conclude that the presented methods are well suited to extract information on geomorphological process dynamics from the complex point clouds and aggregate it at different levels of abstraction to assist landslide and erosion assessment.

show abstract

“…However, an ongoing important key question is the reliable estimation of error budgets in real world open lab setups in order to determine which changes are due to real mass movement and which ones do occur due to the measurement principle itself considering sensor characteristics, measurement setup, and environmental conditions [25,35,36]. Ongoing research focuses on optimizing measurement setups, minimizing registration errors, automating registration [37] and automating change detection for interpreting landslide induced surface changes [38,39].…”

Section: Introductionmentioning

confidence: 99%

Derivation of Three-Dimensional Displacement Vectors from Multi-Temporal Long-Range Terrestrial Laser Scanning at the Reissenschuh Landslide (Tyrol, Austria)

et al. 2018

Self Cite

View full text Add to dashboard Cite

Deep-seated gravitational slope deformations (DSGSDs) endanger settlements and infrastructure in mountain areas all over the world. To prevent disastrous events, their activity needs to be continuously monitored. In this paper, the movement of the Reissenschuh DSGSD in the Schmirn valley (Tyrol, Austria) is quantified based on point clouds acquired with a Riegl VZ®-6000 long-range laser scanner in 2016 and 2017. Geomorphological features (e.g., block edges, terrain ridges, scarps) travelling on top of the landslide are extracted from the acquired point clouds using morphometric attributes based on locally computed eigenvectors and -values. The corresponding representations of the extracted features in the multi-temporal data are exploited to derive 3D displacement vectors based on a workflow exploiting the iterative closest point (ICP) algorithm. The subsequent analysis reveals spatial patterns of landslide movement with mean displacements in the order of 0.62 ma − 1 , corresponding well with measurements at characteristic points using a differential global navigation satellite system (DGNSS). The results are also compared to those derived from a modified version of the well-known image correlation (IMCORR) method using shaded reliefs of the derived digital terrain models. The applied extended ICP algorithm outperforms the raster-based method particularly in areas with predominantly vertical movement.

show abstract

Object‐based classification of terrestrial laser scanning point clouds for landslide monitoring

Cited by 51 publications

References 48 publications

Unmanned Aerial Vehicle Laser Scanning for Erosion Monitoring in Alpine Grassland

Unmanned Aerial Vehicle Laser Scanning for Erosion Monitoring in Alpine Grassland

Multitemporal Analysis of Objects in 3d Point Clouds for Landslide Monitoring

Derivation of Three-Dimensional Displacement Vectors from Multi-Temporal Long-Range Terrestrial Laser Scanning at the Reissenschuh Landslide (Tyrol, Austria)

Contact Info

Product

Resources

About