30 7 ,
metal
25}
= lastic
E 20; : P ]
zZ nn
Id uites |
10} |
3o 0 50 100
SURFACE ORIENTATION (DEGREE)
Figure 2. The reflectance values of different surfaces
Later, all acquired data during the questioning of
unknown object will be matched with these graphs.
The closest graph will assumed to belong to the
examined object. As shown in Figure 2, each curve
has been determined by the polynomial curve fitting for
the corresponding object, which may be formulated as
= ao + 21% + 2% “+ aa X;° (5)
for each node. here ai are the coefficients of the
polynomial curves belong to each characteristic
surface. I}, and x; are the intensity and surface
orientation values. For four nodes of curve, as it is in
Figure 2, four equations (5) are needed.
3. OPERATION OF THE SYSTEM
The system software operational steps are as follows ;
The stereo images of the object is taken as it is
actively illuminated by six-point laser source. By the
properly selected threshold value, only six points
image coordinates are extracted and stereoscopically
matched. Then the Bundle Adjustment technique with
the additional parameters is used to determine X,Y,Z
coordinates of six points. Later, the six-point pattern on
the surface is replaced by circular light (e.g. filament
lamp) dot. Then randomly selected pixel intensity
values of the local surface patch (illuminated by the
dot), are collected and analysed. The set of composite
data (both geometric and photometric) yields sufficient
information of the examined object surface type.
Finally, object surface classification subprogram
utilizes the model surface parameters and attributes
them to the certain object type registered in database
for the ultimate recognition.
618
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B3. Vienna 1996
. REFERENCES
Hartt, E. and M. Carlotto, 1989. A method for
shape-from-shading using multiple images acquired
under different view and lighting conditions. Proc.
IEEE Comp. Vision and Pattern Recognition, San
Diego, CA, pp. 53-60.
Ikeuchi, K., 1980. Numerical shape from shading
and occluding contours in a single view. Artificial
Intell. Lab., Massachusetts Inst. of Technol.
Cambridge, Al-Memo. 566.
. Moffit, F.H. and E.M Mikhail, 1980.
Photogrammetry, Harper & Row publisher Inc.
Newyork - USA
Woodham, R. J., 1980. Photometric method for
determining surface orientation from multiple
images. Optical Engineering, Vol.19, No.1,
January.
KEY |
ABST
Stereo
inform
images
of ima
partict
for tof
may hi
paralla
matchi
Image
difficul
face r«
great :
vision :
tion ar
The pi
cause 1
a close
mental
thorou
anism
Due tc
hensiv
matchi
image
ing to
and st
Signal
The m
correla
is thus
etc. T
engine
Featur
Featur
the in:
matchi
primiti
studies
Global
Insteac
approa
homok
framew
laxatio
holm,
