TRIPLET IMAGE MATCHING FOR AIRBORNE DIGITAL SENSOR ADS40
Lin Zhu, Tadashi Sasagawa, Kikuo Tachibana, Yumiko Fukuzawa
GIS Institute, PASCO Corporation 1-1-2 Higashiyama, Meguro-ku, Tokyo 153-0043, Japan
(lin_zhu, tadashi_sasagawa, Kikuo_tachibana, yumiko_fukuzawa)@pasco.co.jp
TS WG HI/2
KEY WORDS: Airborne Digital Sensor, Triplet Image matching, Coarse-to-fine, Cross-correlation
ABSTRACT:
3D reconstruction from multiple images is one of the most attractive and challenging research fields in photogrammetry and
computer vision. One of the practical applications of 3D reconstruction techniques is automated generation of Digital Surface
Models (DSMs) from satellite or airborne images. However, it is not easy to obtain a reliable DSM through image matching mainly
due to the complexity of urban environments and lack of adequate information for accurate matching. A recently developed
commercial Airborne Digital Sensor ADS40 can provide plentiful information and detailed description of complex urban features.
Thus, there is a great demand for development of corresponding new strategies to exploit the abundant information and special
characteristics of the new sensor. This study introduces a hierarchical triplet image matching strategy for ADS40.
1. INTRODUCTION
DSMs/DTMs are required for topographic mapping, spatial
data visualisation, ortho-image production and many other
engineering and environmental applications. Automated DSM
generation through correlation-based stereo is still a
challenging field due to the complexity of urban environments.
The efforts to acquire reliable DSMs often suffer with a lot of
problems such as occlusions, height discontinuities, repetitive
patterns and ambiguous textures.
An effective way to increase the performance of image
matching is to increase the number of stereoscopic
measurements, which relies on a large amount of data in order
to achieve a high redundancy (Nonin, P., 2003). Handling such
large amount of data can be cost effective if the whole process
is highly automated, however this can hardly be realized using
traditional aerial photography.
Recently, a number of airborne digital CCD line sensors have
been developed, including ADS40, TLS, DPA and HRSC.
These digital push-broom sensors can provide plentiful
information and detailed description of complex urban features
and greatly enhance the efficiency of a photogrammetric
processing. Thus, there is a great demand for development of
corresponding strategies to exploit the abundant information
and special characteristics of CCD line sensors, and
subsequently improve the matching performance.
In this study, a hierarchical coarse-to-fine matching strategy is
proposed for airborne digital sensor ADS40. An algorithm for
feature-based matching over multiple views leads to substantial
reduction of problems caused by occlusions through the
intersection of more than two image rays. Adequate utilization
of ADS40 data, which has a wide radiometric range and a good
signal to noise ratio, improves the ability of reducing the
ambiguity at matching level. Also, the precise exterior
orientation acquired from GPS/INS system enables to perform
a geometric constrained matching and speed up the searching
process.
2. AIRBORNE DIGITAL SENSOR ADS40
The Airborne Digital Sensor ADS40 is the first commercial
airborne push-broom camera developed by Leica Geosystems.
It incorporates latest GPS and INS technology for sensor
orientation and new developments in sensor technology, and
allows precise photogrammetric processing and multispectral
image analysis and interpretation.
DT SS
MN x SONS
denen SNSS
ATRIA eS RN
ANNE ANA
Figure 1. Three-line scanner
The ADS40 sensor exploits a three-line-scanner principle.
Long linear CCD arrays on the focal plane capture
panchromatic images looking forward, nadir and backward
from the aircraft (Figure 1). Besides, four CCD lines for
multispectral bands (RGB and NIR) can provide high-
resolution data both for achieving the redundancy for
photogrammetric processing and multispectral image analysis.
Figure 2 shows different viewing angles of panchromatic and
multispectral images. A specially designed Position and
Orientation System (POS) from Applanix ** Corporation
collects data about the aircraft’s trajectory to ensure 2
successful ground processing, and meets the goals of high
image quality. Each panchromatic channel consists of two lines,
each with 12000 pixels, staggered by 0.5 pixel, usually one
staggered line for each panchromatic channel is used.
International Arch
TCI CALORE 16H
N
1
Backward view
All the panchrom
radiometric resolut
The characteristics
A more detailed «
sensor can be fou
(2000) and Tempel
The ADS40 sens
flown strip, i.e. |
several strips quasi
in both geometry
superior to that ii
facilitates the subs
rectification, auto
DEM generation ar
Focal length
Pixel size
Field of view aci
Pixels per CCD
Pixels per CCD
Dynamic range
Sterco angle (for
Stereo angle (na
Stereo angle (na
Stereo angle (RC
Table 1.
3.
The ADS40 sensoi
geometric and radi
recorded in three p:
from different view
geometric reconst
importance to the re
Although it is possi
(panchromatic, RGI
using only panchror
À hierarchical coa
triplet matching due
Matching is perfor: