internal architecture of distributed real time system of image processing and pattern recognition
DESCRIPTION
Internal architecture of distributed real time system of image processing and pattern recognition. Gostev I. M. Sevastianiv L. A MIEM-PFUR Moscow 2005. Based supposition(1). Pattern – is some description of object ! - PowerPoint PPT PresentationTRANSCRIPT
11
Internal architecture of distributed real Internal architecture of distributed real time system of image processing and time system of image processing and
pattern recognitionpattern recognition
Gostev I. M. Sevastianiv L. AGostev I. M. Sevastianiv L. AMIEM-PFUR MoscowMIEM-PFUR Moscow
20052005
22
BasedBased supposition(1)supposition(1)
• Pattern – is some description of objectPattern – is some description of object! ! • Pattern recognition – separation of input object in predetermined Pattern recognition – separation of input object in predetermined
class under its features or characteristicsclass under its features or characteristics..
• We use We use contour of objectcontour of object at the heart of recognized objects. (It is at the heart of recognized objects. (It is base of gestalt psychology, and base of human perception of base of gestalt psychology, and base of human perception of object). object).
• Contour have considerable proportion of information about Contour have considerable proportion of information about graphical object.graphical object.
33
BasedBased supposition supposition (2)(2)
Development methodology of graphical pattern Development methodology of graphical pattern recognition to invariance to 2D affine transform recognition to invariance to 2D affine transform (translation, scaling and rotation(translation, scaling and rotation ) with receiving as ) with receiving as result object’s coordinates and its angle of rotation result object’s coordinates and its angle of rotation relatively sample. relatively sample.
44
Plan of based steps from methodology of Plan of based steps from methodology of image processing and pattern recognitionimage processing and pattern recognition
Preliminary processing
Pattern recognition
process
Input criteria and samples
Binarisation
Receive sample contour
Contour tracing Clusterisation
I II III
55
DeltaDelta--segmentation principlessegmentation principles
• Using fly window in which calculated statistical Using fly window in which calculated statistical parameters of signals on based which is assignment parameters of signals on based which is assignment value of cutting level.value of cutting level.
• UsingUsing Delta – modulation with only two resulting value Delta – modulation with only two resulting value of signalof signal..
Input Image
Output image
Differentlevel
of intensity
66
Delta segmentation resultsDelta segmentation results ( (11))
Input image and image after delta - segmentation
PS. Intermediate image processing is absent
77
Compare delta-segmentation Compare delta-segmentation to another methodsto another methods
a) Input Image.
b) Image after processing SUSAN method.
c) Image after processing Delta segmentation method.
d) Image after processing Canny method.
a) b)
c) d)
88
Image contour tracing.
Two fragment of image after step of
contour tracing
N.B. Characteristic feature is receive closed contours of object always.
Zooming of fragment ofimage
99
Step conclude is:
Clusters construction – building verbal description of isolated closed contour of objects and saved its to a file. Any of objects may use as sample for process of pattern recognition
Clustering and samples
Contour of recognition object(Cluster)
Noise’s object (Cluster).
Noise on image (Clusters)
1010
Common method of image processingCommon method of image processing..
0Im 1Im Imn
1f
2f
nf
1f 2f nf
1Im Im , , 1,i ifi i i ff i k
1 2
1 2
1, 2, ,
( ) { , ,... }
{ , ,... }
{ ... }
i m
i m
f f f f
f f fl
f F f f f
Im { , , }, 1,f fI N
1111
Understanding of sample Understanding of sample
Sample – this is verbal description of aggregate of groups of parameters, which Sample – this is verbal description of aggregate of groups of parameters, which unambiguously described a object.unambiguously described a object.
On such ofOn such of groups are: groups are: • Processing image’s methods and condition of refinement of object;Processing image’s methods and condition of refinement of object;• Coordinate parts of objects;Coordinate parts of objects;• Pattern recognition’s methods;Pattern recognition’s methods;• Classification thresholds in pattern recognition’s methods;Classification thresholds in pattern recognition’s methods;
Example of Samples
( , ( ))s s s sx I
( )
0
;n
i
i
I I
( ) ( 1)( ), 1,i iiI I i n
( ) ( ) , 1,i ili I l k
1212
Example of implementation subset of sampleExample of implementation subset of sample
1, , ,i ix y i kThe set 0( )I
primary vectors of properties is as set of points in 2d space:
0 0 1( ) ( ){ ( , ), , }l l lI i x y l k
The set of secondary properties is 1 0 1
0 0 ( ) ( ) ( )( )i I I
Let as named “centre of mass” element of set
10
1 1
1 1( ) ( , );k k
c l c ll l
i x x y yk k
is
1( )I1( )I
1313
Understanding “contour function”Understanding “contour function”
2.
3.
4.
0 13 2 1 ( ) ( )( ( ( , )))r I I
0 0 11
( ') ( ) ( )( , )I I I1
2 2 2( , ) {(( ) ( ) ) , }c c
xR x x y y arctg
y
0 1 ( ') ( , ), ,l lI R l k1.
0 02
( '') ( ')( )I I - sorted function of set 0( ')I on angle
in order its increasing. 0
3( '')( )r I - function of interpolation of points of object
max1/ r /í î ðr r - normalization of contour function.
1414
Third step of methodologyThird step of methodology - - recognition processrecognition process
Input clusters for pattern recognition
System identification is based on
consecutive compare clusters whith
preload sample in set of methods. Each
next method is more calculation
difficult and more precision.
Recognized objects
1
2
3
4
““sieve of the sieve of the methods”methods”
1515
MethodMethod of consecutive weighingof consecutive weighing..
0Im
0Im
1
2
n
1 2 n
1Im Im( , , , ), 1, ;
q q
q qq sI q Q
1 0Im Im Imq q
1Im
Imn
1 2, , ,q m
1Im Im
, , , , , , , ;
s ssI I N
1 2{ , , ..., }Q MCWMCW – – this isthis is::
1818
Geometric correlation #Geometric correlation # 1 ( 1 (GCGC1)1)
Let us ( , )xy as function if “difference values” ( )x ( )yand
( , ) ( ) ( )xy x y where , 0, 360
Let us “function of deviation” as ( )xy
360
1
1( ) ( , )
360xy xy
where 0,360
The function of recognition on based of geometric correlationThe function of recognition on based of geometric correlation ##11 is is
1
11
1, min ( );
0, min ( ) .
Ý Oí î ð í î ð
Ý Oí î ð í î ð
Gr r
GGr r
0,360 where
1919
Geometric correlation #Geometric correlation # 22 ( (GC2GC2))
, 0, 360
Let us function
where
The function of recognition on based of geometric correlationThe function of recognition on based of geometric correlation #2 is#2 is
where
360
1
1( ) | ( ) ( , ) |
360xy xy xy
( , )xy
2
22
1, min ( )
0, min ( )
Ý Oí î ð í î ð
Ý Oí î ð í î ð
Gr r
GGr r
0,360
as mean deviation ( , )xy ( )xy
from
2020
Example of pattern recognition (1)
Sample of objects (zoomed)
Image in real size
Zooming fragment
Message about result of recognition
2121
Example of pattern recognition (2)
Sample of objects.
Recognized objects
2222
Example of pattern recognition (3)
Sample of objects
2323
Parallel processing of input Parallel processing of input graphical flowgraphical flow
V
Data General
Data General
Data General
Data General
Data General
Cutting
Parallelprocessing
Result of image processing and pattern recognition
Input information flow is cutting on part and each part is processing into parallel process.
2424
Conveyer processing image and recognition inConveyer processing image and recognition in STIPR2000STIPR2000
k
l
m
n
1-st Method
2-nd method 3-d
Method
N- method
Input image
k, l, m, n – required numbers of line in a method
Outputresults
Conveyer of methods image
2525
Controller of flowController of flow
Controller of flow
Method 1 Method 1 Method 1Buffer Buffer ...Input
graphicflow
Result ofprocessing
andrecognition
2626
Work diagram of the system for dynamic Work diagram of the system for dynamic identification graphic object (identification graphic object (11))
(x, y ), α
1-st host 2-nd host
2727
Work diagram of the system for dynamic Work diagram of the system for dynamic identification graphic object (2)identification graphic object (2)
(x, y ), α
1-st host
2-nd host
2828
Address:
Moscow 117419
Ordzhonikidze 3
Тел: |(095) 916 -8886
mob 7 916-610-7801
Fax: (095) 952-2823
E-mail: [email protected] [email protected]
The work is carried out by
the associated professor
Gostev Ivan Michailovich
and his post-graduates students.
Thank You!