stanbul 2004 
  
dard do not 
realization. 
be directly 
om natural 
constructing 
ged objects 
ommercially 
selection of 
of their 
‚stem that, 
© recording 
letection of 
| time. 
ckground is 
ime on the 
nnels or in 
y means of 
nultispectral 
| not only in 
in natural 
ing. It is a 
d surveying 
on- of the 
nay also be 
F the IACS 
ss London. 
nt, an earth 
    
ALTERNATIVE APPROACHES FOR UTILIZING LIDAR DATA AS A SOURCE OF 
CONTROL INFORMATION FOR PHOTOGRAMMETRIC MODELS 
A. F. Habib?, M. S. Ghanma*, C. J. Kim?, E. Mitishita? 
* Department of Geomatics Engineering, University of Calgary 
2500, University Drive NW, Calgary AB T2N 1N4 Canada — (habib, mghanma)(@àgeomatics.ucalgary.ca, Cjkim@ucalgary.ca, 
® Departamento de Geomática, Universidade Federal Do Paraná, Caixa Postal 19.001, 81.531-970 Curitiba, Paraná, Brasil 
mitishita@ufpr.br 
WG 1/5 Platform and Sensor Integration 
KEY WORDS: Laser Scanning, LIDAR Intensity, Photogrammetry, Linear Features, Absolute Orientation. 
ABSTRACT: 
Laser scanning (LIDAR) is a recent technology that is receiving an increasing interest from professionals dealing with mapping 
applications. The interest in LIDAR is attributed to the rich geometric surface information provided by the data in the form of dense 
non-selective points. On the other hand, photogrammetric processing of stereo-images provides an accurate surface model 
represented by few points as well as a wealth of semantic information about the photographed objects. Considering the nature of 
photogrammetric and LIDAR data, it is clear that the two systems provide complementary information. However, the 
complementary information can only be fully utilized after successful alignment/absolute orientation of the photogrammetric and 
LIDAR models relative to a common reference frame. This paper deals with two alternative approaches for utilizing linear features 
derived from LIDAR data as control information for aligning the photogrammetric model relative to the LIDAR reference frame. 
The first approach incorporates LIDAR lines as control information directly in a photogrammetric triangulation. The second 
approach starts by generating a photogrammetric model through a photogrammetric triangulation using an arbitrary datum (no 
control information). LIDAR features are then used as control information for the absolute orientation of the photogrammetric 
model. A mathematical model is derived to utilize the LIDAR features for the absolute orientation of the photogrammetric model. 
These features can be extracted from range data using various methods. For example, planar patches can be extracted from 3- 
dimensional LIDAR data through segmentation techniques. Then, neighbouring planar patches can be intersected to generate linear 
features corresponding to object space discontinuities. LIDAR data pre-processing for the purpose of feature extraction is not a 
trivial task. An alternative and simpler approach is to use recorded intensities by laser scanners to directly identify and extract linear 
features from the LIDAR data. The paper presents a quantitative analysis of the performance of the different approaches for 
extracting linear features from the LIDAR data. The analysis is based on the quality of fit of the final alignment between the LIDAR 
and photogrammetric models. 
1. INTRODUCTION 
Light Detection and Ranging (LIDAR) is a modern technology, 
which has received wide acceptance and popularity due to its 
usefulness in mapping applications. Since the introduction of 
LIDAR in the mapping industry, its applications in GIS and 
other areas have multiplied. On the other side, photogrammetry 
is a well established mapping and surface reconstruction 
technique. However, the continuous development of LIDAR 
systems in the aspects of reduced hardware size and increased 
resolution and density, made it an increasingly favoured option 
in a variety of applications especially where rapid and accurate 
data collection on physical surface is required (Schenk and 
Csathó, 2002). 
Photogrammetric data is characterized by high redundancy 
through observing desired features in multiple images, making 
it more suited for mapping heavily populated areas. Richness in 
semantic information and dense positional data along object 
space break lines add to its advantages. Nonetheless, 
photogrammetry has its own drawbacks; where there is almost 
no positional information along homogeneous surfaces and 
vertical accuracy is worse than the planimetric accuracy. A 
major obstacle in the way of automation in photogrammetry is 
the complicated and sometimes unreliable matching procedure 
especially when dealing with large scale imagery over urban 
areas. 
LIDAR, on the other hand, is a direct acquisition of positional 
information. Also it produces dense information along 
homogeneous surfaces, making it preferable in mapping Polar 
Regions. Still, LIDAR possesses few undesirable features that 
make it incapable of being a standalone reliable technology. 
LIDAR data has no redundancy and almost has no positional 
information along object space break-lines. Also, the 
planimetric accuracy is worse than the vertical, in addition to 
LIDAR data lacking semantic information. 
Both, photogrammetry and LIDAR, have unique characteristics 
that make them preferable in specific applications. One can 
observe that a negative aspect in one technology is contrasted 
by an opposite strength in the other. Hence, integrating the two 
systems would prove beneficial resulting in more understanding 
of information associated with physical surfaces (Baltsavias, 
1999). However, the complementary information can only be 
fully utilized after successful alignment/absolute orientation of 
the photogrammetric and LIDAR models relative to a common 
reference frame. (Postolov et al., 1999). 
The majority of registration methodologies rely on point 
primitives for solving the registration problem between two 
datasets. Such methodologies are not applicable to LIDAR