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：2002
卷号：XXXIV Part 2
页码：311-316
出版社：Copernicus Publications
摘要：This paper discusses two aspects for filtering Digital Surface Models namely the derivation of parameters for filtering based on a segmentation of laser scanner data and the detection of breaklines. A problem associated with the successful filtering of laser data is the determination of appropriate thresholds used to classify the single height values. Heterogeneous areas, like areas including forest, buildings and undulated or rough terrain, need a specific setting of parameters to control the filtering mechanism. In order to do so, investigations on the use of segmentation approaches, like those being presently implemented within the Definiens eCognition software have been tested using data sets containing height information together with simple texture measures or additional intensity data of the reflected laser beam. The obtained results are presented and limitations are discussed. Depending on the morphology of the terrain it sometimes would be very helpful, if break- or other form-lines of the terrain could be extracted from the existing data. This aspect is addressed within the second part of this paper showing some preliminary work for the detection of break-lines from laser altimetry data . The investigations have been performed on data from a German north sea island, which has very specific topographic properties with regard to surface roughness and absolute heights. For the extraction of break- lines from this laser altimetry data some edge detection filters like the Deriche- and the LOG filter have been tested. Although there are many ways of optimising and automating the approaches for break-. and form-line detection, a wast amount of manual work remains to first define the suitable thresholds and then to clean-up the results i.e. to remove small detected edges (noise).
关键词：Segmentation; Filtering of laser scanner data; Break-line detection
