> points
the
naged
eters
ers
rs
formulation
quation(2).
ject space
observation
can derive
ined initial
the linear
ion (2) is
nent.
ytical Self
alytical self
id is based
arity and
ion in this
g of two or
is the main
arity is a
ally only if
| to effect a
neters are
ved in the
/as omitted
surements,
ecifications
nat reason,
| the quality
limensional
necessary
ystem was
:condly, we
nd the lens
iterested in
objects. In
simple test,
W.
limensional
amcorders,
hey have a
ics. On the
e the tools,
| is shorter
processor,
e hardware
and software of the system used to obtain the imagery of
moving objects.
Geometric and Radiometric Characteristics of the
Camcorders: In order to obtain basic characteristics
of camcorders, VCR, and frame grabber, we took images
under various conditions. There are many precise
methods to find lens distortion. We opted for an on-the-
job calibration, using a plane board with 5 x 5 circular
marks as control points. To coincide the optical axis of
the camera with the center target of the board, a cross
hair drawn on a glass was placed in front of the target
board. Using a theodolite, the central target and the cross
hair on the glass plate were aligned. Then the camcorder
was adjusted such that its optical axis coincided with this
line.
Image of Moving Objects: As we know from
aerial photography, the relative motion between the
sensor and the object causes image blur. When it comes
to imaging of moving objects, image motion depends on
the speed of the moving object, the pixel size in direction
of the motion, object distance, camera focal length, and
shutter speed. Theoretically, in order to get a clean
image, the ratio of motion of object during exposure to
the width of a TV line should not exceed 0.5. To get
practical recording of this phenomenon, we took images
with specified shutter speeds, of the target board falling
from different heights.
Hard/Software |Maker, Specifications
Model
Video Camera | |Panasonic, 1/3“ CCD,
AG-455P S-VHS, S-Video,
1/2" tape
II [Samsung , 1/13" CCD,
H-33 VHS, S-Video,
Hi-8 Tape
VCR Panasonic, S-VHS,
AG-1970 S-Video
Frame Grabber |CreativeLab, Resolution:
VideoBlaster SE 640*480(VGA)
Video Tape Sony, S-VHS
Sony V
PC Dell, Intel 100MHz Pentium,
Dimension P100C 16Mb Ram
Image Processor [Clark University, | Window 95,
IDRISI compatible with TIFF
Table 1. System Components
4-2. 3-D Measurements of a Moving Car
Outline of Model: The objective of this research
is to find out the performance of a conventional video
System, and to suggest better ways to get higher
performance and accuracy in monitoring moving objects
and vibrations. To achieve this objective, we selected a
car as the vibrations-and moving object, and planned to
take images of the process when a driver and passenger
get into the car, start the ignition, then drive ahead. We
expected 3 directional movements from small to large
size in the process.
As shown in Fig. 1, two pairs of targets were attached to
the left and right side of the rear bumper, and one pair
was attached to the license plate of a car. For control, we
put a 3 dimensional control frame in front of the car. The
control frame has 20 control points distributed randomly,
which were coordinated precisely with the Kern E2
Electronic Theodolite System.
Fig. 1. Right Image of a Car and a Control Frame.
Imaging Geometry: The adopted imaging geometry
for a network is a central factor in determining the object
point positioning accuracy. To obtain a near-
homogeneous distribution of object space precision, a
convergent multi-imaging geometry is mandatory. In the
case of a series of images for a moving object, we must
take into account the amount of image manipulating,
therefore we used the minimum number of cameras,
namely two.
An increase in the B/D ratio for near normal imaging
configurations is accompanied by both an improved level
of mean object point precision and enhanced reliability.
In this context, we took images at three different base
lengths, to compare the precision, and chose the best
base length for the test .
Photographing Condition: ^ Based on the preliminary
tests we fixed the shutter speed at 1/1000 s. Considering
the field of the view of the camera(17-20?) and the size of
the control frame sufficient to encompass the car, we
decided on the camera stations. The diameter of circular
targets that could be imagined larger than 3 pixels in x, y
directions was calculated. We used black circular stickers
on white background as targets. The diameter of the
stickers was 19mm and the width of the background was
3 times that of the sticker's. The average distance from
the camera to the objects was 2.7m. resulting in a photo
scale of 1: 270. Additionally, to get a clean image of a car
outdoors without artificial illumination, we needed sun
shine. Unbalanced or poor illumination causes a lot of
problems in target detection and location processes.
Synchronization for Sequential Images: It is very
important to synchronize the left and right images.
Baltsavias & Stallmann(1991) used one electronic device
between the camera and the VCR to add a vertical line of
high contrast and binary pattern to the video signal. The
binary pattern provided a code for the sequence of each
image during recording and permitted safe identification
of each image during the digitization. In this research, a
stopwatch was attached to the front of the control frame.
155
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
