is have many
1s. They also
sists of two
grabber were
id digitized in
id the object
togrammetric
1e dependent
, sind Video -
hrieben. Das
kameras und
iserie für ein
der Direkten
t, dass eine
ystem für die
ds of science
computer
evelop a fully
m capable of
| reach high
able of taking
rage devices
| speed, and
e interesting
methods and
on techniques
requirements.
netrists have
mentioned
irin, 1993).
ience on this
existing ones,
and possible
solutions. As a pilot system, hand-held camcorders,
VCRs, frame grabber, and PC were used to take
sequential images. Basic image processing software,
DLT, and UNBASC2 were used to derive object
coordinates. To analyze the characteristics of the
camcorders, preliminary tests were conducted with still
and moving targets. Then, to estimate the performance of
this system for vibrations, a car was imaged covering
several phases when starting to drive. The images were
recorded on VCR. 15 images per second were digitized in
an off-line mode by frame grabber and the targets
attached to the rear bumper were traced using Direct
Linear Transformation(DLT).
2. VIDEO SYSTEM
2-1. Camcorder
In a video system, the sensor converts light energy into
electrical signals. As a sensor, a video camera is an
image plane scanner unlike the line and push-broom
scanners that operate as object plane scanners. This
means that a 2-dimensional image of the object space,
defined by the field of view of the camera, is formed on
the imaging surface of the tube or a solid state array. This
latent image, scanned by an electron beam(tube camera)
or the photovoltaic seats of the sensor array(solid state
camera) are read-out to generate a video signal, an
amplitude modulated voltage signal. The video signal is
then formatted to produce a 2 dimensional raster image
by means of accurately supplied synchronization
pulses(Vicek, 1988).
The video image is made up of a series of horizontal scan
lines, scanned left to right and from top to bottom much
as a page of text is read. Within the scan, the voltage
level of the signal is varied in proportion to the image
brightness. In most systems the use of television
standards plays a role because the image is often
transferred from camera to the frame grabber by an
analogue television signal. The television signal can be a
source of errors and excludes the use of sensors with
more rows than defined in the standard. The video
industry uses several standards for encoding and
decoding video signals. For example, the NTSC
standards calls for a video frame, or image, to be
scanned every 1/30s. Because this is a rate sufficiently
slow for flicker to be apparent, the frame is divided into
two fields, containing alternate scan lines, each displayed
every 1/60s. This concept, called an interlaced signal,
eliminates the flicker problem in ordinary television
viewing. Two types of CCD camcorders supporting
NTSC standards were used in this research.
2-2. VCR
Video recording is a form of electronic imaging whereby
standard analog television signals are recorded on
magnetic tape or disks. The raw video signal is a
fluctuating DC voltage. The amplitude-modulated, DC-
voltage video signal is converted into a frequency-
modulated signal by video cassette recorder(VCR).
As the video image is used to measure geometric figures,
we have to consider its resolution. The vertical resolution
of recorded video imagery is determined by the fixed
153
number of interlaced scan lines. The horizontal
resolution, in comparison, is determined largely by the
maximum recording frequency. For broadcast television,
this frequency bandwidth is fixed by FCC(Lusch, 1988).
In photogrammetric applications, the video image is
generally not collected for broadcast purposes. In this
context, the horizontal resolution of video image is
determined by the quality of video camera/lens and the
VCR. While VHS format provides a maximum recorded
horizontal resolution of about 240 lines, regardless of how
many lines the camera produces, S-VHS format provides
more than 400 lines of horizontal resolution. Two S-VHS
VCRs were used to record the left and right images in this
experiment.
The frequency-modulated video signal, called the
luminance or Y signal, carries the black and white
information of the scene. This achromatic data provides
the spatial detail in the video image. The color
information in the video signal, the crominance(or croma)
signal, is converted by the VCR to a lower-frequency,
non-modulated subcarrier. Two different color carrying
formats, composite video and S-Video, are mainly used,
depending on the recording format. We used S-Video to
get a high spatial quality from the video image.
2-3. Frame Grabber
CCD sensors most commonly use one of three
addressing strategies: interline transfer, frame transfer,
column-row transfer. In the frame transfer organization,
the sensor consists of vertical columns of CCD shift
registers divided into two zones. One zone, where charge
accumulates during integration time, is photosensitive.
When integration is complete, the whole array is
transferred in parallel to the opaque storage area of the
second zone(Dunbar, 1986).
The A/D converters of most frame grabbers resolve the
analogue light intensity signal coming from the camera
into 8 bit pixels, which is equivalent to 256 shades of
gray, ranging from O for black to 256 for white or vice
versa. A separate device, fabricated by a completely
different integrated circuit process than the imager,
performs the conversion. It was not until 1978 that the
single-chip video rate "flash" analog to digital converter
was invented.
If ADC and /or memory is too slow, every image frame
will not be stored. At a frame rate of 30 frames per
second, it becomes necessary to cope with data sets of
640*480*8 bits in 1/30 s. No conventional PC can cope
with these demands within 1/30 s and it is necessary to
introduce additional processing power into the PC. This
can be done by installing a video digitizer board, also
known as frame grabber. This device converts a frame of
imagery to digital data, stores it in computer memory,
and regenerates a video image from the stored data. We
used a popular frame grabber for the PC, Video-
BlasterSE. It is designed to work with VGA display
standards, and support NTSC, S-Video, and TIFF file
formats.
3. MATHEMATICAL MODEL
3-1. Sub-Pixel Target Location
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
