It refreshes its digits every 1/100s, and could be used as
a code of the sequence. As an auxiliary method, the
frame advance function of the VCR can also be used for
that purpose provided a significant starting point was
identified in each image.
5. TEST RESULTS AND ANALYSIS
5-1. Radiometric & Geometric Characteristics
Radiometric Characteristics: As Curry et al(1986)
have noted, radiometric criteria include sensor response
linearity, system noise, and system stability over time.
System noise is defined as fluctuations in pixel gray
levels caused by random and systematic perturbations in
the CCD sensor and A/D converter. Systematic noise is
seen as a repeated spurious response by the sensor
element. For detecting the system noise of the cameras
used, dark images were captured by running the camera
in the dark(lens cap on) and measuring the output for
each pixel. When the gray value of a pixel exceeded 3
times the standard deviation from the mean of the
image, it was assumed to be a blemished pixel. Fig. 2
shows the results. In Fig. 2(A), black spots mean too low
gray values, and white spots mean too high gray values.
We found that every dark image of a video camera at an
instance has some noise, but the position of noise pixels
changes with time, and rarely fell on the same pixel
position in different images. Moreover, the image of white
paper doesn’t show the noise pixel. In Fig. 2(B), two
groups of black spots on the lower corner of each side
are caused by unbalanced illumination, not noise. In this
context, the images of the video cameras don't have
serious intrinsic noise that should be corrected before
target location determination.
Fig. 2. Noise Pixels of Dark and White Image
Generally, shutter speed influences the image quality. In
order to investigate the influences, a rectangular window
consisting of 25*20 pixels was picked out from the black
target area, and another same size window was picked
out from the white background area . The test images of
target and background were captured under indoor
florescent light and outdoor sun shine at different shutter
speeds, then the ranges of gray values for each window
were collected, and plotted in Fig. 3. The images
captured outdoors show uniform ranges of gray values in
spite of varying shutter speed, but the gray values of
images captured indoors changed with the shutter speed.
The sensitivity of the sensor gradually decreased when
the shutter speed increased. Therefore, when
photographing indoors with short exposure times(e.g.
less than 1/2000s), we should provide enough
illumination according to the surrounding conditions.
Geometric Characteristics: Compared to metric
cameras, non-metric cameras’ lenses are designed for
high resolution at the expense of geometric quality.
Interior orientation parameters lack stability, and there
are no fiducial marks. In addition to the above
disadvantages, the size of unit pixels of video images
changes in the course of digitizing, and the scale changes
with the display format. We only have approximate data
for the object distance, focal length, and pixel size of the
video images. We therefore derived lens distortion
based on some approximations. The horizontal and
vertical pixel sizes of the CCD were deduced from
dividing the length of each side by the total number of
pixels in the corresponding direction. As seen in Fig. 4(A),
the camera has different distortions in vertical, horizontal,
and diagonal directions. This means it has radial and
decentering distortions. But, it is hard to estimate the
amounts of distortion. We derived another lens
distortion curve showing typical radial distortions by using
the calibrated focal length and adjusting the pixel
sizes(Fig. 4(B)) The maximum distortion was 3.5um,
which is quite acceptable.
Black/White Gray Value with Shutter Speed
@8-Blackt
@8-BlackH
250 Os-wntel
32 OS-WhiteH
$150 8 8 8 8 9 g 8 8 8 0
g00 B WF Black,
e n 9 ü 8 ü e B 8 ü e ü 0 i BF -BlackH
OF-Light-L
3 es oË ont
8 2 OF-Lignt-H
Shutter Speed(Sec)
Fig. 3. Change of Gray Value with Shutter Speed
Lens Distortions(Camera:H-33)
0.1
E 00 |... 2. sire tee aree
HE e A e Verica
s 006 ]--------- ic RR ReÓMeeeMell3edÉe..-- Directioion
5 a te A tte. & 3 À tait 8 n B Horizontal
6 u a Direction
ë 0024-9 MMmMm$MMMeee ADiagonal
0 a Direction
- - o = -
: ? o = e
Radial Distances (mm)
Lens Distortions
@ Vertical
Direcboion
8 Horizontal
Direction
A Diagonal
Direction
Distortions (mm)
Radial Distances (mm)
Fig. 4. Lens Distortions(Camera: H-33)
Response to Moving Objects: We expected no friction
when the target board was falling through the stand, but
the images show that was not quite the case. The target
plate shown in Fig. 5 fell from 40cm height above the
optical axis. Theoretically the speed should be 2.8m/s,
but its actual speed was about 1.7m/s. The image blur
156
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B5. Vienna 1996
was improved
during exposul
3in(C):. Am
(C). Vicek(19
during the ex;
clear image. A
expect that r
provide a clea
motion, a shor
(A) Shutter 1/1
s st | "ER
(C) Shutter 1/7
Fig. 5. Falling
5-2 Dynamic
Source Sequt
left and right
VHS VCR.
installed frarr
stopwatch's 1/
However its le
was attached
rear of the car
the frame adv:
capturing the
scenes(e.g. lig
designated as
fields between
exceed 2 fiel
context, we «
synchronized
movement of
board, and th
the frame gr
image was sa
This image
format, and v
images in the
no significant
VHS and VHS
recorded by S
pairs of sequ
VCR. Image «
Were processe
Sub-Pixel Tz
coordinates c
preprocessed
e Noise ren
