Coastal Dynamics 2013 Paper No. LIDAR DATA FOR CLASSIFICATION AND SCIENTIFIC MONITORING IN COASTAL AREAS

In this paper we propose a method for the classification and mapping of habitats in Wadden Sea areas from lidar data. For this task, we derive crucial features from the 3D point cloud based on the geometry and intensity of the backscatter. The characteristics of three different classes (water, mudflat, mussel bed) are learned in a training step of a supervised classification approach. For this purpose, we implement a method capable of modelling context: a Conditional Random Fields framework. From the labeled point cloud 2D objects are generated for the classes water and mussel bed. Thus, we determine water-land-boundaries as first step towards digital terrain modelling and contribute to the mapping of habitats, two important tasks of coastal monitoring. The results are evaluated on a test site of the German Wadden Sea in the North Sea.