723
the image A. In the reflected signal in
tensity formation it was supposed that
the propagation and reflection conditi
ons in every pixel are identical. Pig.3
gives the image F, All images are quan
tized by 16 levels. The dynamic range of
images includes 256 levels. The backgro
und frequency was set at f/ =100. In
this case objects frequencies values are
equal to =113, f«=1l6, f, =119, f* =122.
A in exponential distribution is set
equal to 100. In (1) /À7/T=40. Pig.4
shows the image after the threshold pro
cessing. This image is an initial seg
mentation for the subsequent algorithm
implementation. Pig.5 presents the re
sultant image, obtained after 5 iterati
ons of algorithm implementation, with
5x5 pixels window size. Objects are sa
tisfactory discriminated from the back
ground. Individual points on the image
edges are retained owing to the edge
effect.
Pig.5-10 show similar images for the
case of objects arbitrary motion. Ob
jects Doppler frequencies are chosen so
that a part of pixels of their patterns
have a frequencu which coincides with
a background frequency and that is the
most complicated case for the segmenta
tion. The initial segmentation (fig.9)
is implemented through two-threshold
processing of image P (fig.8). The re
sultant image (fig.10) is obtained in
10 iterations of the algorithm imple
mentation with 5x5 pixels window size.
All objects are satisfactory extracted.
Their shape is retained. The small-size
regions which appeared due to false al
arms may be easily removed at the sub
sequent stages of image processing.
On the whole the mathematical modeling
confirms the effectiveness of proposed
algorithms. The choice of variable pa
rameters can be made only in the process
of specific applied problems solving.
The proposed algoritlims provide the ma
thematically validated means of DOI seg
mentation problem solution. The impor
tance of the obtained results for prac
tical application is obvious as far as
two-dimensional convolution operations
are easily implemented on specialized
processors and the P image threshold
processing used as the initial segmenta
tion can be implemented with, the rate of
information receipt. Taking into account
the low rate of information receipt the
binary segmentation algorithm implemen
tation doesn’t require the VHSIC tech
nology application.
4. Dansac,J.,Meyzonnette,J.L.,1985.non
laser Doppler rengefinding with hetero-
dine detection and chirp pulse compres
sion. Proceedings of Ihe SPIE,v.590:
380-388.
5. Derin, H., Cole,VV.S., 1986.Segmentation
of textured images using Gibbs randomfi-
elds.Computer vision, graphics and image
processing,v.35:72-98.
6. Prey,R,W.,1983. laser imaging radar
sensor,Proceedings of the SPIE,V,415:
2-12.
7. Goodman,J.W., 1985.Statistical Optics.
Wiley,New York.
8. Hanson,P.R.,Elliott,H.,1982.Image se
gmentation using simple Markov fields
models.Сотри ter vision, graphics and im
age processing,V.20:101-132.
9. Harney,R.C.,Hull,R.J.,1980.Compact
infrared radar technology.Proceedings of
the SPltí,V,227:162-171.
10. Harney,R.C.,1981.Military applicati
ons of coherent infrared radar.Proceed
ings of the SPIE,v.300:19£32.
11. Hull,R.J.,1982.A place for laser in
radar.Photonics spectra,v.16(4):60-65.
12. Лешн Б.Р. ,1974.Теоретические основы
статиотическои радиотехники.Том.I.С о-
зетокое радио»Москва.
13. Letalick,D.,1986.Measured signal am
plitude distributions for a coherent FM-
CW C0 9 laser radar.Applied optics,v.25
(21) 0927-3938.
14. Papurt,D,Shapiro, J. , 1981 .Atmospheric
propagation effects on coherent laser
radars.Proceeding of the SPIE,v.300:86
-99.
15. Pratt,W.K.,1978.Digital Image Proces
sing. Wiley, New York,
16.Sullivan,D.R.,1980.Active 10.6 m im
age processing.Proceeding of the SPIE,
v. 238:103-118.
17. Therrien,C.W.,Quatieri,T,P.»Dudjion,
T.E,,1986.Statistical model-based algo
rithm for image analysis.Proceeding of
the IEEE,v.74(4):532-551.
18. Therrien,C.W.,1983.An estimation-
theoretic approach to terrian image
segmentation.Computer vision, graphics
and image processing,v.22:313-326.
19. Wang,J.Y.,1984.Image C0 9 laser radar
field test.Applied optics,v.23(15):2565
-2571.
9. REFERENCES
1. Albert,A.,1972.Regression and the
Moore-Penroce Pseudoinverse.Academic
Press,New York.
2. Besag,J.,1974.Special interactions
and statistical analysis of Lattice sys
tems. Journal of the Royal Statistical
Society,ser.B,v.36:192-236.
3. БорисенкоИ.B 0 »Златопольскик А.А.»Муч
ник А.Б.,1987.Сегментация изображения.
Автоматика и телемеханика,74 (4):4-25