;ES 
CUE 
Kerry McIntosh 
  
IMPROVEMENT OF AUTOMATIC DSM GENERATION OVER URBAN AREAS USING 
AIRBORNE LASER SCANNER DATA 
Kerry MCINTOSH", Amnon KRUPNIK', Toni SCHENK" 
"Department of Civil Engineering, Technion — Israel Institute of Technology, Haifa, Israel 
tkerry, krupnik} @tx.technion.ac.il 
"Department of Civil and Environmental Engineering and Geodetic Science, The Ohio State University, 
Columbus, Ohio, USA 
schenk.2(@osu.edu 
  
Working Group III/2 
KEY WORDS: Data fusion, Laser, Photogrammetry, Surface Reconstruction. 
ABSTRACT 
Airborne laser scanning has provided an efficient and accurate method of data acquisition for the determination of 
visible surface topography. The laser data can be quickly processed to provide coordinates of points on the visible 
surface with high spatial frequency and precision. This technology has benefits as compared to photogrammetric 
techniques, however there are limiting factors due to the laser data having no thematic information. The limitations 
may be overcome by utilizing aspects of both laser altimetry and photogrammetry in the surface determination process. 
Research has been undertaken to accurately determine the visible surface in urban areas using airborne laser scanner 
data and digital aerial imagery. In this project, edges detected in the aerial images are used to refine the digital surface 
model (DSM) produced using data from an airborne laser scanner. The three dimensional edge information allows 
improvement of the laser DSM by providing accurate locations of surface discontinuities. Therefore the laser data and 
the edge information are merged to obtain the benefits of each data set, facilitating the generation of an accurate terrain 
model. 
The approach has been tested over an urban site covering Ocean City, Maryland, USA. Laser data and aerial imagery, 
acquired on the same day, are available. Experiments were performed to test and refine the algorithm. This paper 
presents the data fusion approach, describes the data set and details the results of experimentation undertaken to 
investigate the validity of the approach. 
1 INTRODUCTION 
Digital surface models (DSMs) of urban areas are becoming widely used in an increasing number of applications, such 
as digital orthophoto production, three dimensional (3D) city modeling and 3D building reconstruction. Methods for 
generating surface models include using laser scanner data and using digital photogrammetric methods. 
Airborne laser scanning is an efficient and accurate method of obtaining visible surface information with high spatial 
frequency. However these data provide only coordinates and have no thematic information. Terrain information, such 
as formlines or breaklines, is not available from the laser scanner data. The location of surface discontinuities may be 
estimated by methods such as planar segmentation (Schenk ef a/., 1999), but accurate positions are not available 
(Vosselman, 1999). The spatial distribution of the laser data will affect the level of inaccuracy that is introduced due to 
the lack of terrain information. A high spatial distribution of data may well provide enough information to define the 
surface to the required accuracy, however if the spatial resolution of the data is low, extra information will be required 
to accurately define the surface. 
Digital photogrammetry also provides a method of automatic DSM generation, however this process has limitations in 
accuracy due to image matching problems, which are caused by factors such as lack of image texture and 
foreshortening. An advantage of photogrammetric methods is that accurate locations of surface discontinuities can be 
obtained from aerial imagery. It is proposed to merge the accurate terrain information from the imagery with the laser 
scanner data, providing a more accurate data set than either of the two separate data sources. There are two phases to 
the proposed approach, surface matching and data fusion. 
In the first phase, surfaces are created from each data source and accurately registered to the same coordinate system. 
Theoretically, the two surfaces should be on the same coordinate system, however experience shows that systematic 
  
International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000. 563