TY - JOUR

T1 - Water-land-classification in coastal areas with full waveform lidar data

AU - Schmidt, Alena

AU - Rottensteiner, Franz

AU - Sörgel, Uwe

PY - 2013/5/1

Y1 - 2013/5/1

N2 - In this paper, we investigate full waveform lidar data acquired over the German Wadden Sea areas in the south eastern part of the North Sea. We focus especially on classification of the 3D point clouds with the aim to determine water-land-boundaries. This is a first step towards digital terrain model generation in order to analyse the terrain topography in coastal areas and, by comparing different epochs, its dynamics. For the classification of the lidar points, we learn typical class features in a training step and combine local descriptors with context information in a conditional random fields (CRF) framework, a probabilistic supervised classification approach capable of modelling contextual knowledge. We compare the results with those obtained by a fuzzy logic based approach developed specifically for the water-land- classification in Wadden Sea areas. With the latter approach we achieve a correctness rate of more than 82% for water detection. By integrating context, the results can be significantly improved by approximately 10%. Moreover, we investigate the waveform features of the data which reveals unexpected nonlinear effects concerning the decomposition of the waveforms.

AB - In this paper, we investigate full waveform lidar data acquired over the German Wadden Sea areas in the south eastern part of the North Sea. We focus especially on classification of the 3D point clouds with the aim to determine water-land-boundaries. This is a first step towards digital terrain model generation in order to analyse the terrain topography in coastal areas and, by comparing different epochs, its dynamics. For the classification of the lidar points, we learn typical class features in a training step and combine local descriptors with context information in a conditional random fields (CRF) framework, a probabilistic supervised classification approach capable of modelling contextual knowledge. We compare the results with those obtained by a fuzzy logic based approach developed specifically for the water-land- classification in Wadden Sea areas. With the latter approach we achieve a correctness rate of more than 82% for water detection. By integrating context, the results can be significantly improved by approximately 10%. Moreover, we investigate the waveform features of the data which reveals unexpected nonlinear effects concerning the decomposition of the waveforms.

KW - Classification

KW - Coast

KW - Conditional random fields

KW - Lidar

KW - Water

UR - http://www.scopus.com/inward/record.url?scp=84877278865&partnerID=8YFLogxK

U2 - 10.1127/1432-8364/2013/0159

DO - 10.1127/1432-8364/2013/0159

M3 - Article

AN - SCOPUS:84877278865

VL - 2013

SP - 71

EP - 81

JO - Photogrammetrie, Fernerkundung, Geoinformation

JF - Photogrammetrie, Fernerkundung, Geoinformation

SN - 1432-8364

IS - 2

ER -