DLR-Logo -> http://www.dlr.de
DLR Portal Home | Imprint | Privacy Policy | Contact | Deutsch
Fontsize: [-] Text [+]

Determination and verification of robust spectral features for an automated classification of sealed urban surfaces.

Heiden, Uta and Segl, Karl and Roessner, Sigrid and Kaufmann, Hermann (2005) Determination and verification of robust spectral features for an automated classification of sealed urban surfaces. In: Proceedings of the 4th EARSeL Workshop on Imaging Spectroscopy in Warsaw, pp. 151-159. 4th EARSeL Workshop on Imaging Spectroscopy, 2005-04-27 - 2005-04-29, Warsaw, Poland.

Full text not available from this repository.


Urban surface materials have an immense influence on ecological conditions of urban areas. In this context, frequent material-oriented identification of urban surfaces can support municipal authorities in planning processes. This requires hyperspectral remote sensing data that are characterized by high spectral and spatial resolution. For full exploitation of the information content, detailed spectral knowledge of urban surfaces is required and automated methods are needed for effective processing of those data. In this study, a methodology for the determination and validation of materials-specific spectral features has been developed that can be used for automated identification of urban surface materials. It is demonstrated for representative urban surfaces in the cities of Dresden and Potsdam, Germany. First, an urban spectral library consisting of field and image spectra was developed that has been analyzed in regard to typical spectral features and their variations. Based on a visual analysis, robust features that are highly independent of spectral variations caused by e.g. illumination and degradation have been selected. The quality of these features has been validated in regard to an improved separability of urban surface materials analyzing a confusion matrix within a classification process. For this purpose, the features were described numerically using adapted feature functions. The results of the classification show a very good separability especially for predominantly bright urban surfaces, but also for darker materials with weak reflectance features. Even differentiation between spectrally similar urban surfaces was feasible. The derived robust spectral features reduce the need for test-site-specific training information. This way, the developed methodology represents an important step towards fully-automated identification of urban surfaces.

Item URL in elib:https://elib.dlr.de/22594/
Document Type:Conference or Workshop Item (Speech)
Title:Determination and verification of robust spectral features for an automated classification of sealed urban surfaces.
AuthorsInstitution or Email of AuthorsAuthor's ORCID iDORCID Put Code
Journal or Publication Title:Proceedings of the 4th EARSeL Workshop on Imaging Spectroscopy in Warsaw
Refereed publication:No
Open Access:No
Gold Open Access:No
In ISI Web of Science:No
Page Range:pp. 151-159
EditorsEmailEditor's ORCID iDORCID Put Code
Keywords:imaging spectroscopy, spectral separability, urban surface materials, numerical features
Event Title:4th EARSeL Workshop on Imaging Spectroscopy
Event Location:Warsaw, Poland
Event Type:international Conference
Event Start Date:27 April 2005
Event End Date:29 April 2005
HGF - Research field:Aeronautics, Space and Transport (old)
HGF - Program:Space (old)
HGF - Program Themes:W EO - Erdbeobachtung
DLR - Research area:Space
DLR - Program:W EO - Erdbeobachtung
DLR - Research theme (Project):W - Vorhaben Spektrometrische Verfahren und Konzepte der Fernerkundung (old)
Location: Oberpfaffenhofen
Institutes and Institutions:German Remote Sensing Data Center > Environment and Security > Environment and Geoinformation
Deposited By: Heiden, Dr.rer.nat. Uta
Deposited On:20 Mar 2006
Last Modified:24 Apr 2024 19:04

Repository Staff Only: item control page

Help & Contact
electronic library is running on EPrints 3.3.12
Website and database design: Copyright © German Aerospace Center (DLR). All rights reserved.