You are using an outdated browser that does not fully support the intranda viewer.
As a result, some pages may not be displayed correctly.

We recommend you use one of the following browsers:

Full text

Mapping without the sun
Zhang, Jixian

Zhen Xiong, Yun Zhang
Department of Geodesy & Geomatics Engineering, University of New Brunswick,
15 Dineen Drive, PO Box 4400, Fredericton, NB, Canada E3B 5A3
v009v@unb.ca, yunzhang@unb.ca
Commission VI, WG VI/4
KEY WORDS: Vehicle Information Extraction, One Pass, Quickbird Imagery
ABSTRACT: Vehicle information is useful for transportation management, security surveillance and military applications. The
vehicle information includes vehicle’s ground position, moving speed, and moving direction. Vehicle detection and vehicle
information extraction is usually based on Radar, SAR, or video. The platforms are almost ground based or airborne based; but
seldom is space-borne based. Basically these techniques are usually used for security surveillance or military applications. Some of
them just provide the service of catching vehicle, without any information on the vehicle’s position and velocity. An automated
method of vehicle information extraction from Quickbird imagery is presented in this paper. Because there is a time interval between
Quickbird’s panchromatic and the multi-spectral imagery, once we get a vehicle’s image positions on both the MS and the Pan
imagery, we can calculate vehicle’s ground position and distance the vehicle moved during the time interval, so as to calculate
vehicle’s moving speed and moving direction. In this research, an image classification based vehicle detection method, a region
growing method to refine the image position, a multi-step image matching method to search the corresponding vehicle position, and
an algorithm to calculate vehicle’s ground position from its image position based on DEM are developed. From vehicle’s image
position to ground position, image position, sensor model, DEM, image matching, and region growing are involved in the process. In
this paper, the methodology is introduced first. Later an experiment and result are given. Finally the discussion and conclusion are
Moving target detection is a fast growing research field. Most
of techniques for moving target detection are based on Radar
[Liu Guoqing, et al., 2001; Nag, S., et al., 2003; Liu, C.-M. Jen,
C.-W., 1992], SAR [Dias, J.M.B., et al., 2003; Hong-Bo Sun, et
al., 2002; Pettersson, M.I., 2004; Soumekh, M.I., 2002], or
video [Munno, C.J., et al., 1993]. The platforms are almost
ground based [Castellano, G., et al., 1999; Nag, S., et al., 2003;
Munno, C.J., et al., 1993; Pettersson, M.I., 2004] or airborne
based [Liu Guoqing, et al., 2001; Hong-Bo Sun, et al., 2002;
Soumekh, M.I., 2002]. But seldom is space-borne based.
Basically these techniques are usually used for security
surveillance or military applications. Some of them just deliver
the service of catching moving target, without any information
on the target’s position and velocity. Based on different
equipment, different techniques were adopted for catching
moving targets, such as using generalized likelihood ratio as a
threshold to decide which target is moving [Liu Guoqing, et al.,
2001; Dias, J.M.B., et al., 2003; Pettersson, M.I., 2004]; using a
filter for digital moving target detection [Nag, S., et al., 2003];
applying a fractional Fourier transformation [Hong-Bo Sun, et
al., 2002]; or utilizing victo’s frequency domain spatiao-
temporal filtering and spatio-temperal constraint error of image
frame pairs to detect and track moving targets (e.g., personnel)
in natural scenes in spite of low image contrast, changes in the
target's infra-red image pattern, sensor noise, or background
clutter [Munno, C.J., et al., 1993].
In this paper we introduce a new technique to vehicle’s
positions and moving speed based on a single set of high spatial
resolution satellite imagery. Such single set of satellite imagery
consists of one panchromatic image and the corresponding
multi-spectral imagery, which are acquired by some current
high resolution satellites, such as SPOT, IKONOS, and
Quickbird. Although the time interval between panchromatic
image and multi-spectral image is very small, if the vehicle is
moving, the panchromatic image and multi-spectral image
should record this position change during satellite time interval.
So theoretically if the vehicle is moving, we can find the
vehicle is moving and even moving speed of the vehicle. We
developed an algorithm to calculate the position and moving
speed of moving vehicle which is acquired by both
panchromatic and multispectral images at a very close time.
To date, many satellites can acquire both the panchromatic
(PAN) and the multi-spectral (MS) images at the same time.
But because of technique arrangement, these satellites usually
catch the panchromatic and multi-spectral images not really at
same time. Usually there is a very small time interval between
the PAN and the MS imagery. Therefore, if a ground vehicle is
moving, theoretically this moving vehicle should be record in
different ground position. That is to say, if we can calculate
vehicle’s ground position from image coordinates, we should
obtain two different coordinates from the PAN and the MS
images respectively for the same moving vehicle. Then from
these two ground coordinates, we can calculate the moving
speed and moving direction of the vehicle. This is the elements
of our moving vehicle detection technique.
However, because the time interval between the PAN and the
MS imagery is very small, less than 1 second, the position of
moving vehicle changes within this small time interval is also
very small. Therefore if the error of position calculation is
greater than the value of its position change, we can never
detect moving vehicle correctly. Therefore, some methods must
be used to minimize the errors of vehicle position calculation.
Following is the detailed steps of moving vehicle information