An Approach for Filtering Lidar Data in Coastal Vegetated Areas Using Intensity Information and Multiple Echoes

Accurate digital terrain models (DTM) are one of the most important requirements for many applications in coastal management and safety, such as the calculation of the volume of dunes and dikes for the purpose of coastal protection. Airborne LIDAR sensors provide dense height information of large areas in an efficient manner, therefore such data are appropriate to derive suitable DTM. Besides reasons of efficiency and economy, the accuracy and especially the reliability of the data are essential factors for the applicability in safety related domains. In case of moderate surface roughness in non-vegetated areas LIDAR DTM usually provide a standard deviation in height of less than 15cm. However, the accuracy and reliability of the LIDAR DTM points suffer if the laser beam interacts with vegetation. Several filter algorithms were developed in order to eliminate the vegetation points in LIDAR data sets. Usually, they apply geometric criteria, for instance the slope in a defined neighbourhood, to solve this task. However, in areas of very dense vegetation and rough terrain, where only a few laser pulses are able to penetrate the canopy, such processing often fails resulting in an upward height shift of the derived DTM. In this paper additional features are proposed, which correspond to the reflectance characteristics of the backscattering objects, to support the filtering proccess. The introduced new algorithm uses intensity information and the distribution of multiple echoes for adaptive determination of the weights during an iterative surface fitting. Based on several control areas located in different types of coastal shrubberies the potential of this method is demonstrated. The results show that the integration of the new features decreases the differences between the LIDAR based surface and the control measurements by a few centimetres.