Edward M. Mikhail
MULTI-SOURCE FEATURE EXTRACTION AND VISUALIZATION
IN URBAN ENVIRONMENTS
Edward M. Mikhail
Head, Geomatics Engineering
1284 Civil Engineering Building
Purdue University
West Lafayette, IN 47907-1284, USA
mikhail @ecn.purdue.edu
KEY WORDS: Sensor Model, Invariance, Hyperspectral, Building Extraction, Road Grid, Visualization
ABSTRACT
Basic research is being performed by a team composed of specialists in photogrammetry, spatial image analysis, remote
sensing, computer vision, and visualization, for the purpose of efficiently extracting urban features from multi-image
sources and construction and visualization of the resulting database. The team members work cooperatively such that
the effort is an integrated research. Topics discussed include: sensor modeling for data registration, photogrammetric
invariance, DEM supported classification of hyperspectral imagery, DEM and thematic data supported building
extraction, DEM supported road-grid extraction, and visualization in support of photogrammetry and exploitation
research.
1 INTRODUCTION
Extraction of information from imagery has been the domain of photogrammetry, remote sensing, and image
understanding/computer vision for many years. To be sure, the types of imagery used and the theories and techniques
applied have varied somewhat from one of these three disciplines to another. Nevertheless, the primary objective of all
is to obtain correctly labeled features which are geometrically and positionally accurate enough to be useful for a
variety of applications. The practice in the past has been for researchers and practitioners in each of these three areas to
work essentially independently from others. Of course, each area was aware of the activities of the others, and
attempted to adapt and use methodologies developed by the others to the extent possible by their understanding of such
methodologies. The increased prevalence of imagery in digital form, and the introduction of new sources of data,
brings to focus the inadequacy of such independent pursuit of a similar goal. It has become quite apparent that
combined integrated team research by experts in these fields is likely to yield significantly more than what can be
expected from the sum of the individual efforts. Nowhere can this be more apparent than in the extraction and
visualization of labeled spatial features in urban environments. This task has been, and continues to be, the most
demanding in time and effort. In order to meet this challenge, and to put in place a team to address this problem in an
integrated fashion, the US Army Research Office, under the Multidisciplinary University Research Initiative, MURI,
awarded a 5-year project to Purdue University as the lead institution. The MURI team members and their speciality
MURI - Multidisciplinary University | d Vision [
: Imagery, mun :
Research Initiative Rapid and Affordable
Generation of Terrain and Detailed Urban Feature Support Info.
Natjonal 3D Models
tical Terrain
gius
Data
* Purdue University - * MARCONI Integrated
CE/Geomatics Systems, Inc.
— Spatial Image — Wisualization/Simulation
Analysis/Photogrammetry — . US Army TEC
Geospatial
Data Base >
Tr. A
TT.
* Purdue University - — Collaborating Lab
ECE/Remote Sensing ENN
= Mu an Hyper-Spectral ~ Technology Transfer
» USC - Institute of
ong NS
Systems
— Image Understanding
++
Basic and Perspective, Bn Maps/
Applied R& D Animations Charts
Figure 1. MURI Team Members Figure 2. Vision
592 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.
areas
South
projec
RE
and fi
featur
on Col
2 SE
Since
are "1
mean:
types
subse
à proc
cultur
2.1 M
Frame
photo.
three
often
and r:
photo:
reasor
partici
invari
for ar
lx,
Althoi
video
can be
groun
systen
transf
relativ
are ex