Automatic DEM Extraction from an IKONOS Stereo pair over an Urban Area
Taejung Kim
Satellite Technology Research Center
Korea Advanced Institute of Science and Technology
373-1 Kusung-Dong, Yusung-Gu, Taejon, Korea 305-701
Email: tjkim@satrec.kaist.ac.kr
Yong-Jo Im
Satellite Technology Research Center
Korea Advanced Institute of Science and Technology
373-1 Kusung-Dong, Yusung-Gu, Taejon, Korea 305-701
Email: yjim@satrec.kaist.ac.kr
Abstract:
Automatic extraction of Digital Elevation Models (DEMs) from urban scenes is one of the most active research topics over many years. This paper addresses the experiments carried out to generate DEMs over urban areas from IKONOS images using automated stereo matching techniques. An IKONOS image pair over Taejon city was used for tests. Using ground control points obtained from differential GPS, camera model was set up . The stereo matching algorithm used in this experiment was originally developed for SPOT images at 10m resolution. It was also successfully applied to KOMPSAT- 1 EOC at 6.5m resolution. Although the algorithm used were not designed to overcome the problem of height discontinuities, occlusions and shadows, our intention is to analysis the quality of output DEM using the current technique and to check to what extent a DEM over an urban area can be feasibly extracted. As a result, a DEM over urban areas was produced. Although small houses appear as small “cloudy” patches, large apartments and
industrial buildings are visually identifiable. The boundaries of such structures do not produce sharp edges in the output DEM but appear as blurred clouds. However, when an 1:1000 digital map was overlaid to the DEM from IKONOS, the height information around building boundaries could be enhanced. This may suggest that DEMs from IKONOS image can be used for map revision. The DEM was compared with 18 ground control points and showed 6.5m RMS errors. The experiment and analysis in this paper show that 1m resolution satellite images can be used for urban DEM generation together with the existing digital maps and that, however, algorithm improvement is required to handle occlusion, shadows, and height discontinuities .
1. Introducion
Automatic extraction of Digital Elevation Models (DEMs) from urban scenes is one of the most active research topics over many years. Surface characteristics of urban areas, such as shadows, occlusions or height discontinuities, hinder any automated schemes to work in reliable, robust or effective manners. DEM generation over urban areas using ster eo matching, lidar processing, or SAR interferometers all faces difficulties in handling such surface characteristics.
This paper addresses the experiments carried out to generate DEMs over urban areas using automated stereo matching techniques. In particular, this paper will focus on the feasibility analysis of automated DEM generation from the 1m resolution spaceborne IKONOS images. We will generate a DEM over an urban area from an IKONOS stereo pair using stereo matching algorithm developed originally for the SPOT images [Park et al., 2000]. Although this algorithm were not designed to overcome the problem of height discontinuities, occlusions and shadows, our intention is to analysis the quality of output DEM using the current technique and to check to what extent a DEM over an urban area can be feasibly extracted.
