期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2007
卷号：XXXVI-3/W52
页码：156-161
出版社：Copernicus Publications
摘要：Authorities operating in the field of coastal management require reliable area-wide height information for their responsibilities regarding to the safety of the coastal area. In this context the lidar technique replaces more and more traditional methods, such as terrestrial surveying, and is now the most important source for the generation of digital terrain models (DTM) in this zone. However, coastal vegetation interferes with the laser beam, resulting in a height offset for the lidar points depending on different vegetation types occurring in this region and their phenology. Various filter algorithms were developed for lidar data in vegetated areas, which are able to minimize this offset. But in very dense vegetation and hilly terrain these algorithms often fail resulting in certain residuals. In a previous approach the height offset was estimated based on grid data. In this algorithm the offset was linked to suitable features in the remote sensing data. A segment based supervised classification was performed using these features to partition the lidar data into different accuracy intervals. A major problem of this method arises from the fact that the accuracy intervals do not correspond to distinct and easily separable clusters in the feature space. Considering a single vegetation type the height offset exhibits a rather continuous characteristic. In a new approach this issue is tackled by modelling the offset with respect to the features using continuous functions. Additionally, feature extraction and classification are performed on raw data, in order to maintain the significance of the features by avoiding transformation artefacts and to increase the accuracy of the classification. On the basis of test data a comparison between the two methods is conducted to emphasize the problems and their solutions
