Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data

Предложены эквивалентностные пространственно-инвариантные алгоритмы распознавания полутоновых изображений идентификационных объектов, представленные набором символов, и результаты экспериментальных исследований. Показаны интерфейсы программы и результаты ее тестирования, подтверждающие высокое быстр...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:Управляющие системы и машины
Datum:2013
Hauptverfasser: Krasilenko, V.G., Nikolsky, A.I., Bozniak, Yu.A.
Format: Artikel
Sprache:English
Veröffentlicht: Міжнародний науково-навчальний центр інформаційних технологій і систем НАН та МОН України 2013
Schlagworte:
Online Zugang:https://nasplib.isofts.kiev.ua/handle/123456789/83175
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Назва журналу:Digital Library of Periodicals of National Academy of Sciences of Ukraine
Zitieren:Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data / V.G. Krasilenko, A.I. Nikolsky, Yu.A. Bozniak // Управляющие системы и машины. — 2013. — № 4. — С. 12-19. — Бібліогр.: 18 назв. — англ.

Institution

Digital Library of Periodicals of National Academy of Sciences of Ukraine
id nasplib_isofts_kiev_ua-123456789-83175
record_format dspace
spelling Krasilenko, V.G.
Nikolsky, A.I.
Bozniak, Yu.A.
2015-06-16T14:38:33Z
2015-06-16T14:38:33Z
2013
Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data / V.G. Krasilenko, A.I. Nikolsky, Yu.A. Bozniak // Управляющие системы и машины. — 2013. — № 4. — С. 12-19. — Бібліогр.: 18 назв. — англ.
0130-5395
https://nasplib.isofts.kiev.ua/handle/123456789/83175
004.896: 681.5
Предложены эквивалентностные пространственно-инвариантные алгоритмы распознавания полутоновых изображений идентификационных объектов, представленные набором символов, и результаты экспериментальных исследований. Показаны интерфейсы программы и результаты ее тестирования, подтверждающие высокое быстродействие и низкий процент нераспознанных символов.
Equivalently space-invariant pattern recognition algorithms halftone identification of objects and the results of experimental studies are proposed. Interface of this software are shown. Results of its tests confirmed the high performance and low part unrecognized characters, less than 0,01%.
Запропоновано еквівалентнісні просторово-інваріантні алгоритми розпізнавання напівтонових зображень ідентифікаційних об'єктів, що представлені набором символів, та результати експериментальних досліджень. Показано інтерфейси програми та результати її тестування, що підтвердили високу швидкодію і низький відсоток нерозпізнаних символів.
en
Міжнародний науково-навчальний центр інформаційних технологій і систем НАН та МОН України
Управляющие системы и машины
Автоматическая обработка и распознавание изображений
Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
Алгоритмы распознавания полутоновых изображений многосимвольных идентификационных объектов с использованием нелинейных эквивалентностных метрик и анализ экспериментальных данных
Алгоритми розпізнавання напівтонових зображень багатосимвольних ідентифікаційних об’єктів з використанням нелінійних еквівалентнісних метрик та аналіз експериментальних даних
Article
published earlier
institution Digital Library of Periodicals of National Academy of Sciences of Ukraine
collection DSpace DC
title Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
spellingShingle Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
Krasilenko, V.G.
Nikolsky, A.I.
Bozniak, Yu.A.
Автоматическая обработка и распознавание изображений
title_short Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
title_full Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
title_fullStr Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
title_full_unstemmed Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
title_sort recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data
author Krasilenko, V.G.
Nikolsky, A.I.
Bozniak, Yu.A.
author_facet Krasilenko, V.G.
Nikolsky, A.I.
Bozniak, Yu.A.
topic Автоматическая обработка и распознавание изображений
topic_facet Автоматическая обработка и распознавание изображений
publishDate 2013
language English
container_title Управляющие системы и машины
publisher Міжнародний науково-навчальний центр інформаційних технологій і систем НАН та МОН України
format Article
title_alt Алгоритмы распознавания полутоновых изображений многосимвольных идентификационных объектов с использованием нелинейных эквивалентностных метрик и анализ экспериментальных данных
Алгоритми розпізнавання напівтонових зображень багатосимвольних ідентифікаційних об’єктів з використанням нелінійних еквівалентнісних метрик та аналіз експериментальних даних
description Предложены эквивалентностные пространственно-инвариантные алгоритмы распознавания полутоновых изображений идентификационных объектов, представленные набором символов, и результаты экспериментальных исследований. Показаны интерфейсы программы и результаты ее тестирования, подтверждающие высокое быстродействие и низкий процент нераспознанных символов. Equivalently space-invariant pattern recognition algorithms halftone identification of objects and the results of experimental studies are proposed. Interface of this software are shown. Results of its tests confirmed the high performance and low part unrecognized characters, less than 0,01%. Запропоновано еквівалентнісні просторово-інваріантні алгоритми розпізнавання напівтонових зображень ідентифікаційних об'єктів, що представлені набором символів, та результати експериментальних досліджень. Показано інтерфейси програми та результати її тестування, що підтвердили високу швидкодію і низький відсоток нерозпізнаних символів.
issn 0130-5395
url https://nasplib.isofts.kiev.ua/handle/123456789/83175
citation_txt Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data / V.G. Krasilenko, A.I. Nikolsky, Yu.A. Bozniak // Управляющие системы и машины. — 2013. — № 4. — С. 12-19. — Бібліогр.: 18 назв. — англ.
work_keys_str_mv AT krasilenkovg recognitionalgorithmsofmultilevelimagesofmulticharacteridentificationobjectsbasedonnonlinearequivalentmetricesandanalysisofexperimentaldata
AT nikolskyai recognitionalgorithmsofmultilevelimagesofmulticharacteridentificationobjectsbasedonnonlinearequivalentmetricesandanalysisofexperimentaldata
AT bozniakyua recognitionalgorithmsofmultilevelimagesofmulticharacteridentificationobjectsbasedonnonlinearequivalentmetricesandanalysisofexperimentaldata
AT krasilenkovg algoritmyraspoznavaniâpolutonovyhizobraženiimnogosimvolʹnyhidentifikacionnyhobʺektovsispolʹzovaniemnelineinyhékvivalentnostnyhmetrikianalizéksperimentalʹnyhdannyh
AT nikolskyai algoritmyraspoznavaniâpolutonovyhizobraženiimnogosimvolʹnyhidentifikacionnyhobʺektovsispolʹzovaniemnelineinyhékvivalentnostnyhmetrikianalizéksperimentalʹnyhdannyh
AT bozniakyua algoritmyraspoznavaniâpolutonovyhizobraženiimnogosimvolʹnyhidentifikacionnyhobʺektovsispolʹzovaniemnelineinyhékvivalentnostnyhmetrikianalizéksperimentalʹnyhdannyh
AT krasilenkovg algoritmirozpíznavannânapívtonovihzobraženʹbagatosimvolʹnihídentifíkacíinihobêktívzvikoristannâmnelíníinihekvívalentnísnihmetriktaanalízeksperimentalʹnihdanih
AT nikolskyai algoritmirozpíznavannânapívtonovihzobraženʹbagatosimvolʹnihídentifíkacíinihobêktívzvikoristannâmnelíníinihekvívalentnísnihmetriktaanalízeksperimentalʹnihdanih
AT bozniakyua algoritmirozpíznavannânapívtonovihzobraženʹbagatosimvolʹnihídentifíkacíinihobêktívzvikoristannâmnelíníinihekvívalentnísnihmetriktaanalízeksperimentalʹnihdanih
first_indexed 2025-11-26T16:17:17Z
last_indexed 2025-11-26T16:17:17Z
_version_ 1850627635924697088
fulltext 12 УСиМ, 2013, № 4 Автоматическая обработка и распознавание изображений UDC 004.896: 681.5 V.G. Krasilenko, A.I. Nikolsky, Yu.A. Bozniak Recognition algorithms of multilevel images of multicharacter identification objects based on nonlinear equivalent metrices and analysis of experimental data Предложены эквивалентностные пространственно-инвариантные алгоритмы распознавания полутоновых изображений иден- тификационных объектов, представленные набором символов, и результаты экспериментальных исследований. Показаны ин- терфейсы программы и результаты ее тестирования, подтверждающие высокое быстродействие и низкий процент нераспо- знанных символов. Ecvivalently space-invariant pattern recognition algorithms halftone identification of objects and the results of experimental studies are proposed. Interface of this software are shown. Results of its tests confirmed the high performance and low part unrecognized charac- ters, less than 0,01%. Запропоновано еквівалентнісні просторово-інваріантні алгоритми розпізнавання напівтонових зображень ідентифікаційних об'єктів, що представлені набором символів, та результати експериментальних досліджень. Показано інтерфейси програми та результати її тестування, що підтвердили високу швидкодію і низький відсоток нерозпізнаних символів. Introduction. The main perspective tendency of development of information-computating systems and computer technologies is making them intel- lectual and similar to human thinking and percep- tion. As the models of intellectual systems in the field of neurophisiology, artificial intelligence the following models of auto-processing nets are mainly used: connection models, neural nets models, mo- dels with parallel distributed processing. Basic researches reveal a number of important general principles besides spontaneous intellectual quali- ties. Among those features we can mention the principle of parallel processing of information on all the levels of its processing (global system be- haviour, considered as “intellectual”) and the con- cept of active and not passive memory. The great interest to neural model forces to revalue the fun- damental theses in many fields of knowledge, in- cluding computer techniques. Neural models of brain activity and cognitive processes, most pro- bably will cause perspective results in neurophy- siology, neurology, the advent of the systems with increased computing resources and intelligent qua- lities for solving problems of medical informatics and diagnostics. Many failures on the way of im- proving of artificial intelligence have appeared in recent years as firstly of all, the chosen computing techniques were not adequate to solve the impor- tant and complicated problems, and secondly, simp- le and not perfect neural models and nets were applied. Today, rapid progress in mathematical logic, especially matrix (multivalued, continuous, fuzzy, neural) [1–6], accumulation of data regard- ing continual (analog) and obviously nonlinear functions of neurons [7–9], elaboration of the neu- ral net theory, neurobiology and neurocybernetic, and adequate algebrologic instruments for mathe- matical description and modeling [6, 10–15], de- velopment of optical technologies have created conditions for construction of technical systems, adequate almost to any problem of artificial intel- ligence. The works [10, 16], and especially [11–14] solve the problem of increase of capacity in artifi- cial neural networks (ANN) and associate mem- ory (AM), even in cased storing of greatly corre- lated images, and the problem of convergence of methods and training rules, using multilevel repre- sentation of signals. The use of operations of neu- ral logic operations – equivalence and nonequiva- lence for construction of ANN and AM models is common for works [11–14]. In this connection such models and the theory were called “equivalental”. УСиМ, 2013, № 4 13 They showed, and described negative and inhibit- ting weights along with exciting ones at unipolar and bipolar coding. Neural biologic (NBL) [1, 3, 10–12] is integration (gnoseologically development loped and specified) of known logic: multivalued, hybrid, continuous, fuzzy etc. At the same time, the integrated operations in fuzzy logic are: operation of fuzzy negation, t-norms and s-norms, and they have the relation of dualism according to general form of De Morgan principle. The example of t-norms are: logical multiplication (min (a, b)), algebraic multiplication (a·b), limited multiplica- tion etc. The example of s-norms are: logical sum (max (a, b)), algebraic sum (a + b – a·b), limited sum (1(a + b)), contract sum etc [15, 17]. The basic operations of NBL, used in equivalental models NNAM [10–13], are binary operations of equivalence and nonequivalence, which have a few variants [12]. The variants of these equivalence operations on a carrier set ]1,0[ uC are shown on fig. 1 (a,b,c) respectively for: )},min(),,max{min(~1 bababaeq   ; bababaeq  ~2  ; babaeq   1~3 (1) baeq ~1  baeq ~2  0 0,5 1 0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 a a eq1 0 0,60 0,2 0,4 0,6 0,8 1 E2 b eq2 baeq  ~3 0 0,5 1 0 0,2 0,4 0,6 0,8 1 E3 c eq3 Fig. 1. Operations of equivalence Their negations are first, second and third non- equivalence respectively. Relation of these opera- tions when unipolar and bipolar coding is similar to relation of the signals and expressed as: 2/))/~((/~ (~)(~) bbuu baDba  , where (~) /~ is mean- ing valid for any of these operation variables, be- ing:  uuu Cba , ],0[ D , ],[, DDCba bbb  . In general case for scalar variables  Cba , = [A, B] – continuous line segment, signals them- selves and their functions and segment C can be brought to segments  DD, (or [–1, 1]) in bipolar coding and  D,0 (or [0, 1]) in unipolar coding. When shifting on interval, corresponding to  C = = [A, B], et at the value  (i.e. when  C = [A   , B  ])) set is chosen) the above mentioned equivalence and nonequivalence operations over variables    Cbbaa ^ , are expressed over the simple addition to equivalences and nonequi- valences from non-shifted variables the shift  , namely:  )~(~ baba  ;      )~(~ baba . Besides: for k > 0, )/~( (~) bak  )/~( (~) bkak  . Thus, equivalence (nonequivalence) of signals do- esn’t depend on constant shift (analogy to constant component), on scale factor k, on bipolar or uni- polar coding. These operations are the generaliza- tion of XNOR and XOR operations of binary logic and allow logical comparison of continuous-level (analog) and multilevel signals with unipolar and bipolar representation, including scalar, vector and matrix. For that purpose normalized equivalence and nonequivalence of vectors and matrices are used [11, 12, 14], which for these two matrices A =  JIija  , B =  JIijb  JI ]1,0[ are determined in the following way:     J j ij I i a JI 1 ij 1 n b~1~ BA ; (2)     J j ij I i a JI 1 ij 1 n b/~1/~ BA . (3) Normalized equivalence n~ and nonequivalence n/~ – are more general new complementary met- 14 УСиМ, 2013, № 4 rics in metrical space R. In particular,  n/~ is a normalized distance JId /),(1 BA between ma- trices and for  N1,0, BA it turns into norma- lized distance NdH /),( BA of Hamming. As it can be seen from expressions (2) and (3), component and component operation of NBL (~) or ( /~ ) is generalized for matrix case, and NBL logic be- comes matrix NBL (MNBL). The variants of op- erations (~),( /~ ) depend on different types of t-norms and s-norms operations used in them and integrated fuzzy-operations of union and intersec- tion [17]. Depending on the type or variants of equivalent algebra (EA) [11, 12] we can offer new algebro-logical instruments for creation of equiva- lental theory on the basis of MNBL. In works [11, 12] new notions of equivalental unidimension (1- D) and twodimension (2-D) functions were intro- duced:    N i ii ba N bafE 1 )~(1),()(~   ; .)~( *),(~),(~ 1 1 ,. ~      N n M m nnmn ba fE BABA (4) In this equation symbol ( ~ ) denotes the convo- lution (correlation) with operation of “equivalence”. Normalized space-dependent equivalence and nonequivalence functions were introduced in the work [14]   JI  /~~ BAe and     .1,0 ~][/* )1)(1( , /~/~         JMINne JI e1BAe These functions e~ and ~e reflect the measure of equivalence and nonequivalence of two images depending on their mutual spatial shifting. The connection of functions E~ (, ) with correlation functions (linear LCF and nonlinear NCF) was demonstrated [11, 14]:            JIJI /*/* ~ BABAe  );()( B,ABA, nn LCFLCF  (5) )()( ~ B,ABA,e nn LCFLCF    ; (6) .)()( )/()**( )/(*~~ minmin B,ABA, BABA BAe nn FCNFCN JI JI            (7) On the basis of these functions more genera- lized modified matrix-tensor neurological equiva- lental models (MTNLEMs) for space-invariant re- cognition of 2-D images were synthesized [14]. Making use of formulae transformation for cal- culation of linear and nonlinear correlative func- tions and criterial functions of mean absolute error (MAE) and mean square error (MSE), authors show in work [18] that these formulae can be reduced to two groups of mathematical constructions. These constructions determine two groups of architectu- res of parallel action: high-speed correlators with nonlinear and image morphological processing. In practice there often appears the problem dealing with recognition of multilevel images of the ob- jects being identified; on special seals, on engine- ering objects designed for various applications, do- cuments etc. That is why, taking into account the above mentioned facts and possibilities of equiva- lent functions applications shown earlier, we will suggest in the given work the recognition algo- rithms based on nonlinear equivalent metrics and will show the results of investigation of these al- gorithms. The algorithms of recognition without pre- liminary segmenting of input image (the algo- rithms of group I) The idea of NLEАs is concluded in segmenting (on the base of a priori information) into frag- ments, corresponding to separate symbols, source image of multicharacter object or reference image, composed of a set of alphabet of symbols, sub- jected to recognition. For each ith segment of source image nonlinear equivalence functions with general reference image for the first group of NLEAs realization variants are calculated. For the УСиМ, 2013, № 4 15 second group of NLEA realization variants NLEFs for each ith reference image segment with the ge- neral input multicharacter image are calculated. Inside of each NLEА group we apply several mo- difications of NLEFs, depending both on the type of "equivalence" ("nonequivalence") being used and on parameters, determining the type of non- linearity and a number of autoequivalence trans- formations. Moreover the latter can be applied both to separate components of images being compared and to integrate NLEF-evaluations and normalized NLEF as a whole. Let dimensions (in the number of pixels) K (in vertical position) and L (in hori- zontal position) in rectangular fragment Р of the image be chosen in such a way, that each of pos- sible Q reference images of )10(  Qq S symbols from selected alphabet could be paced in fragment re- gion (most closely and with minimal dimensions of rear plan). Then multi level image JIija  ][A to be recognized of multisymbol (R-symbol) iden- tification object О with horizontal row wise ac- commodation of symbols will include all R of )10(  Rr P fragments, each of which can be the image of one qS . of symbols. Total dimensions I, J of the image A must be greater than values K and R·L accordingly and meet the requirements: I = K + d1; J = R·L + d2, where d1 and d2 – total amounts of pixels in vertical and horizontal positions accord- ingly complementing the interfragmental space till dimensions A. Fig. 2,а shown initial multilevel (256 levels )2550( ija ) image A = {0,255}IxJ, to be recognized, images of arranged set (alphabet) of symbols   LK lkq S  255...1,0][ ,S (see fig. 2,b). In this case, criterial function (space-dependent) )1()1( , ][),(  LJKI q q b B can be determined for each qS reference, as it is shown in fig. 2,с. Knowledge algorithms regarding the number of R symbols in identification object (number of frag- ments in the image A) and possible deviations of symbols positions taking into account the gaps permit to divide the image of two-dimensional criterial function into R regions. In each R-th re- gion there is an extremum (min or max) of crite- rial function qB namely qr max(min)B and its coordi- nates. Comparing all q max(min)B within r-th region, we can find index q’ (number of symbol referen- ce), which gives us the greatest (the least) value of criterial function in this region. Thus we deter- mine (recognize) symbol-reference in each r-th region, and using the coordinate )(, ',', qrqr yx of the extremum of the best equivalence (similarity) or nonequivalence (nonsimilarity) the position of recognized reference symbol in the object is de- termined, and coordinates of the following region (sign location) are being searched. For the analysis of NLEF functions obtained after the first step, both of integral signs and for decision taking based on their collection, we perform transformation of 2-D NLEFs into 1-D NLEFs, carrying out by component operations of minimum (maximum) over vector arrays, cor- responding columns (or lines) of initial 2-D NLEFs. Thus we select local spatial extremums of NLEF. Such compression of information is possible and reasonable and permits to simplify considerably further analysis of recognition. On the following step by the result of comparisons of 1-D transformed NLEF local extremums (lighting countings out) (number of which cor- responds to a number of reference recognized symbols), we come to a conclusion on the pres- ence or absence (the most probable) of one of the reference symbols and on its coordinate lo- cation in initial image being recognized. The modification of the given algorithms is the algo- rithms, when criterial functions are calculated only in R -th local regions ( RR  ' )in the proximity of arising determined possible posi- tion of the extremum. It considerably decreases temporal expenses needed for its realization. Formula for calculation of maximorum (mini- morum) of criterial function (q-th) in R region is the following: )( (min) max r qr xB ,...,(max 10 ,,(min) q yx rq yx r rr bb ), 1, q yx r KIr b  . Formula for q selection is the following: qq  , if q x rQq x rq x rq x r rrrr Bbbb  ),......,(max 110 (min) . (8) 16 УСиМ, 2013, № 4 Fig. 2. Formation process of criterial functions Recognition algorithms applying the method of preliminary segmentation of input image (algorithms of group II) First step: Image Р0 is formed as arranged set of all reference symbols:   ))(( ,,0 1 01 0 21 dLQdK ji Q q q Q i i P      SPP . (9) Second step: The image to be recognized A   JI 255,0 is segmented into R sections, re- gions (sign places of symbols) having the dimen- sions of I  J pixels taking into account a priori information-regarding the location of symbols and possible deviations. Third step: Criterial functions ),(),( 0PAB rF for each rth segment of input image A and Р0 (as the set of references) or criterial functions ),( 0PA qrF are calculated for each rth segment Ar and arranged set qth conventional region Р0, namely qP0. Fourth step: For each segment Ar (its serial number) the number q of conventional region of reference arranged set Р0, is determined, this number gives the greatest (the least) value of cri- terial function for the best special mutual shift. By this number q the recognition of rth segment of identification object image is carried out (com- puter code or reference symbol is put on the posi- tion, coordinate of rth segment). Criterial functions For all possible 3 groups of algorithms we ap- ply the following criterial functions from two im- ages NM ija  ][A and LK klp  ][P , where aij, ,0{klp 1,, 255}, M > K, N > L: a)                  1 0 1 0 ,,,, , 255 )255)(255(1 )( K k L l lklklklk a papa LK B , (10) moreover )(0 KM  ; )(0 LN  ; b)      ,)255,255min( ),min(2551 ),min(),max(1 1)( ,, 1 0 1 0 ,, 1 0 1 0 ,,,, 1 0 1 0 ,,, lklk K k L l lklk K k L l lklklklk K k L l lklkb pa pa LK papa LK pa LK                                 B (11) Fig. 3. Part of lD NLEF function )(7,5,2,1 max 2,1  qr b B (Xr=1= 8, Xr=2= 26) a Input recognized multilevel image b Equivalence function (with character “1”) c Dependency of function )(max r qr c XB Fig. 4. Character «1» recognition processes are shown УСиМ, 2013, № 4 17 c) )]([)( ,,   bf anc B , (12) d) )]([)( ,,   bf bnd B , (13) where nf – nonlinear function used in work [14] and named “autoequivalence”  aafn ),(          ;5,0,/~.../~ ;5,0,~~ ... aforaa aforaa timesa timesa     (14) e) ]([)( ,,,, lkne apf  B . (15) (15) In criterial function ),( Be nonlinear trans- formation is performed over the whole set of sam- ples, obtained before addition over the region of indices k, l change. Formation of reference images of symbols Unlike the recognition algorithms suggested before, which are based on equivalence models of neuronets [14], input images being recognized and images of reference symbols (while teaching and input) are not converted into the set of two- grading images, but are processed in the initial gray scale format. In order to form reference symbols, mathematical expectations are determi- ned from the sets of representatives for each qth symbol. Averaged on several realizations frag- ments of separate symbols, were used for shaping of reference base and reference general image. Taking into account slight possible shifts the in- distinct (blurring) representation of each qth sym- bol is formed. Computer modeling and the results For the selection of the best optimal criterial function, taking into account the influence of va- rious noises and interference, distorting the image being recognized, selection of the needed algo- rithm from the group of algorithms, we have de- veloped the program aimed at modeling of sug- gested recognition algorithms. The suggest pro- gram permitted to establish the needed, form, al- gorithm type, criterial function, display the results (final and intermediate) in graphic interface (see fig. 5), suitable for researcher. Image, being recog- nized, in particular, of 16446 pixels of dimensi- onality were coded by 256 gradation levels, presen- ting 7 symbol identification number of the object. Other images were also used, for instance, 14-sym- bol object with spaces etc. In the first case dimen- sionality of reference image was 3018 pixels. Fig. 5. Graphical interface of research program for recognition algorithms simulation Deviation from forecasted distance between symbols at segmenting and processing we choose ±10 pixels. On image being recognized we super- imposed a noise with different parameters and for each value of parameter and each type of NLEFs 1D-transformed NLEFs were computed, corres- ponding to all possible references. Fig. 3, 4 showed 2D and 1D NLEFs. When increasing a parameter, characterizing the degree of nonlinearity in 2D NLEFs, type of the latter can be controled. NLEА allow recognize under noising up to 50% (normal law). Correlation of peaks to lateral petals can be increased, selecting the type of function. Noised multilevel images and recognition results are shown by fig. 6. Consider the possible hardware imple- mentation. For calculation of criterial functions equivalentors of images, introduced in [11, 12] can be used. They are correlators (convolution opera- tion systems) realizing linear LCF and nonlinear NCF space-dependent functions from source and additional images (see (5)(7)). The realization of such devices is described in [14, 18] For calcula- tion of nonequivalent functions the same devices are used, but while writing in LCTV two images are written (one original image and the second – complementary image) in accordance with for- mula (6). The important problem is the deve- lopment of parallel operating real-time high per- 18 УСиМ, 2013, № 4 formance digital convolvers and correlators, func- tioning in signal region. This problem will be re- ported in further publications. To test the proposed approach has been developed a real program for the rapid identification numbers on buildings locking sealing devices. The program production tested at the state enterprise "Vіnnitsatransprilad" South-Western Railway. The system has a number of advantages: 1) Enters the 20 images on the bodies of locking sealing devices with a matrix that is installed on the scanner; 2) Recognize the images invariant to shift and turn on the bodies of locking and sealing devices, and not sensitive to shift and rotation of the buildings themselves locking sealing devices in the matrix; 3) Signals of images to recognize bad and does not recog- nize at all; 4) Recognize the 7 digits on the image and exports them to bill submitted machine codes; 5) Refer recognized numbers of locking sealing devices with numbers already in a database; 6) Recognize the image of a seven-digit one set of 20 images of locking sealing device in 100-200 seconds; 7) Resistant to very noisy images of locking sealing devices and lighting glare on the surface. Graphical interface of the experimental manufacturing program for recognition al- gorithms simulation are shown on fig. 7. Results of computer modeling and laboratory studies, conducted on real objects have con- firmed advantages of such algorithms. Addi- tional entering of correlation factors in rec- ognition models improves the quality and validity of NLE-recognition algoritms. As our research shows, NLEА possess higher discriminant properties than correlation and other conventional algorithms. When adding background component in the image being recognized in reasonable measues the character and quality of recog- nition doesn’t change applying NLEA. As a result of modeling the hypothesis that the preliminary processing of images, containing considerable level of noise, in particular, seperation of sidebars does not improve but worsens the quality of recognition. Conclutions. The suggested nonlinear- equivalent recognition algorithms of multilevel images of identification objects possess good re- cognition quality, especially if the objects to be re- cognized are in noisy environment, if background noises have been added. NLE-algorithms permit to recognize, as it has been proved by lab tests, in such noise conditions when traditional (correla- Number of recognized characters Input images Noise, % I group of algorithms II group of algorithms Diagrams For ),( Bb , noise – normal 20 7 7 40 7 7 50 7 7 60 7 7 70 7 7 80 7 7 90 7 7 0 1 2 3 4 5 6 7 20 40 50 60 70 80 90 Noise, % N um be r o f r ec og ni ze d ch ar ac te rs I groupe of algorithms II groupe of algorithms (modif ied) For ),( Bb , noise – by Gause 20 7 7 40 7 7 50 7 7 60 6 6 70 3 5 0 1 2 3 4 5 6 7 20 40 50 60 70 Noise, % N um be r o f r ec og ni ze d ch ar ac te rs I group of algorithms II group of algorothms (modif ied) Fig. 6. Noised multilevel images and recognition results Fig. 7. Graphical interface of testing program for recognition algo- rithms simulation УСиМ, 2013, № 4 19 tion) algorithms fail. The experiments have shown that preliminary processing of images being rec- ognized, in particular those, whose outlining, when they are saturated with noise and decreased levels number, does not lead to increase of recognition quality. The results of theoretical research are im- plemented in a software product designed for the rapid identification numbers on the sealing de- vices. The software product was tested on produc- tion enterprise "Vinnitsatranspribor" South-Wes- tern Railway. Interface of this software are shown. Results of its tests confirmed the high perform- ance and low part unrecognized characters, less than 0,01%. 1. Krasilenko V.G., Kolesnitsky O.K., Bogukhvalsky A.K. Creation Opportunities of Optoelectronic Continuous Logic Neural Elements, Which are Universal Circuitry Macrobasis of Optical Neural Networks // Proc. SPIE. – 1995. – 2647. – P. 208–217. 2. Levin V.I. Continuous Logic, Its Generalization and Application // Automatica and Telemechanica. – 1990. – N 8. – P. 3–22. 3. Krasilenko V.G., Kolesnitsky O.K., Mikhalnichenko N.N. Lines of Optoelectronic Neural Elements with Optical Inputs/Outputs Based on BISPIN-devices for Optical Neural Networks // Proc. SPIE. – 1995. – 2647. – P. 264– 272. 4. Krasilenko V.G., Magas A.T. Fundamentals of Design of Multifunctional Devices of Matrix Multilevel Logic with Fast Programmed Adjusting // Measuring and Com- puter Technique in Technological Processes. – 1999. – N 4. – P. 113–121. 5. Awwal A.S. Abdul., Khan M. Iftekharuddin. Computer Arithmetic for Optical Computing. Special Section // Optical Eng. – 1999. – 38, N 3. – P. 400–402. 6. Volgin L.N. Complementary Algebra and Relative Models of Neural Structures with Coding of Channel Numbers // Electrical Modeling. – 1994. – 16, N 3. – P. 15–25. 7. Sokolov E.N., Vaytkyavichus G.G. Neurointelligence: from Neuron to Neurocomputer. – M.: Nauka, 1989. – 238 p. 8. Pozin N.V. Modeling of Neural Structures. – M.: Nauka, 1970. – 264 p. 9. Antomonov Yu.G. Principles of Neurodynamics. –Kiev.: Nauk. dumka, 1974. – 200 p. 10. Krasilenko V.G., Bogukhvalskiy A.K., Magas A.T. Equi- valental Models of Neural Networks and Their Effec- tive Optoelectronic Implementations Based on Matrix Multivalued Elements // Proc. SPIE. – 1996. – 3055. – P. 127–136. 11. Krasilenko V.G., Kolesnitsky O.K., Bogukhvalsky A.K. Applications of Nonlinear Correlation Functions and Equivalence Models in Advanced Neuronets // Proc. SPIE. – 1997. – 3317. – P. 211–222. 12. Continuous Logic Equivalental Models of Hamming Network Architectures with Adaptive-Correlated Weig- hting / V.G. Krasilenko, F.M. Saletsky, V.I. Yatskov- sky et al. // Proc. SPIE. – 1997. – 3402. – P. 398–408. 13. Demonstration of neural-network efficiency models with adaptive equivalentaly weightings and interconnection matrixes adjustment / V.G. Krasilenko, A. I. Nikolsky, V. M. Voloshin et al. // Measuring and Comp. Techn. in Technol. Processes. – 2000. – N 4. – P. 119–122. 14. Optical pattern recognition algorithms on neural-logic equivalental models and demonstration of their prospec- tivaness and possible implementations / V.G. Krasilen- ko, A.I. Nikolskyy, A.V. Zaitsev et al. // Proc. SPIE. – 2001. – 4387. – P. 247–260. 15. Gnoseological Approach to Search of Most General Functional Model of Neuron / V. Krasilenko, A. Nikol- skyy, V. Boyko et al. // Measuring and Comp. Techn. in Technol. Processes. – 2000. – N 7. – P. 23–27. 16. Krasilenko V.G., Nikolsky A.I., Pavlov S.N. The asso- ciative 2D-memories based on matrix-tensor equiva- lental models // Radio Electronics, Comp. Sci., Con- trol. – 2002. – N 2 – P. 45–53. 17. Kuzmin V.B., Travkin S.I. The theory of fuzzy signals in control tasks and construction principles of fuzzy processors. Survey of foreign literature // Automatic and telemechanica. – 1992. – N 11. – P. 3–36. 18. Krasilenko V., Motygin V., Pastushenko A. The archi- tecture of high-speed correlators with non-linear and morphological image processing // Proc. SPIE. – 1994. – 2321. – P. 538–541. Тел. для справок: +38 0432 519-249, +38 098 370-7440, 096 193-7836, 097 261-9784 (Винница) E-mail: krasilenko@mail.ru, nikolskyy@i.ua, it@minisoft.com.ua © В.Г. Красиленко, А.И. Никольский, Ю.А. Бозняк, 2013  << /ASCII85EncodePages false /AllowTransparency false /AutoPositionEPSFiles true /AutoRotatePages /None /Binding /Left /CalGrayProfile (Dot Gain 20%) /CalRGBProfile (sRGB IEC61966-2.1) /CalCMYKProfile (U.S. Web Coated \050SWOP\051 v2) /sRGBProfile (sRGB IEC61966-2.1) /CannotEmbedFontPolicy /Error /CompatibilityLevel 1.4 /CompressObjects /Tags /CompressPages true /ConvertImagesToIndexed true /PassThroughJPEGImages true /CreateJobTicket false /DefaultRenderingIntent /Default /DetectBlends true /DetectCurves 0.0000 /ColorConversionStrategy /CMYK /DoThumbnails false /EmbedAllFonts true /EmbedOpenType false /ParseICCProfilesInComments true /EmbedJobOptions true /DSCReportingLevel 0 /EmitDSCWarnings false /EndPage -1 /ImageMemory 1048576 /LockDistillerParams false /MaxSubsetPct 100 /Optimize true /OPM 1 /ParseDSCComments true /ParseDSCCommentsForDocInfo true /PreserveCopyPage true /PreserveDICMYKValues true /PreserveEPSInfo true /PreserveFlatness true /PreserveHalftoneInfo false /PreserveOPIComments true /PreserveOverprintSettings true /StartPage 1 /SubsetFonts true /TransferFunctionInfo /Apply /UCRandBGInfo /Preserve /UsePrologue false /ColorSettingsFile () /AlwaysEmbed [ true ] /NeverEmbed [ true ] /AntiAliasColorImages false /CropColorImages true /ColorImageMinResolution 300 /ColorImageMinResolutionPolicy /OK /DownsampleColorImages true /ColorImageDownsampleType /Bicubic /ColorImageResolution 300 /ColorImageDepth -1 /ColorImageMinDownsampleDepth 1 /ColorImageDownsampleThreshold 1.50000 /EncodeColorImages true /ColorImageFilter /DCTEncode /AutoFilterColorImages true /ColorImageAutoFilterStrategy /JPEG /ColorACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /ColorImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000ColorACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000ColorImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasGrayImages false /CropGrayImages true /GrayImageMinResolution 300 /GrayImageMinResolutionPolicy /OK /DownsampleGrayImages true /GrayImageDownsampleType /Bicubic /GrayImageResolution 300 /GrayImageDepth -1 /GrayImageMinDownsampleDepth 2 /GrayImageDownsampleThreshold 1.50000 /EncodeGrayImages true /GrayImageFilter /DCTEncode /AutoFilterGrayImages true /GrayImageAutoFilterStrategy /JPEG /GrayACSImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /GrayImageDict << /QFactor 0.15 /HSamples [1 1 1 1] /VSamples [1 1 1 1] >> /JPEG2000GrayACSImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /JPEG2000GrayImageDict << /TileWidth 256 /TileHeight 256 /Quality 30 >> /AntiAliasMonoImages false /CropMonoImages true /MonoImageMinResolution 1200 /MonoImageMinResolutionPolicy /OK /DownsampleMonoImages true /MonoImageDownsampleType /Bicubic /MonoImageResolution 1200 /MonoImageDepth -1 /MonoImageDownsampleThreshold 1.50000 /EncodeMonoImages true /MonoImageFilter /CCITTFaxEncode /MonoImageDict << /K -1 >> /AllowPSXObjects false /CheckCompliance [ /None ] /PDFX1aCheck false /PDFX3Check false /PDFXCompliantPDFOnly false /PDFXNoTrimBoxError true /PDFXTrimBoxToMediaBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXSetBleedBoxToMediaBox true /PDFXBleedBoxToTrimBoxOffset [ 0.00000 0.00000 0.00000 0.00000 ] /PDFXOutputIntentProfile () /PDFXOutputConditionIdentifier () /PDFXOutputCondition () /PDFXRegistryName () /PDFXTrapped /False /CreateJDFFile false /Description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> /CHS <FEFF4f7f75288fd94e9b8bbe5b9a521b5efa7684002000410064006f006200650020005000440046002065876863900275284e8e9ad88d2891cf76845370524d53705237300260a853ef4ee54f7f75280020004100630072006f0062006100740020548c002000410064006f00620065002000520065006100640065007200200035002e003000204ee553ca66f49ad87248672c676562535f00521b5efa768400200050004400460020658768633002> /CHT <FEFF4f7f752890194e9b8a2d7f6e5efa7acb7684002000410064006f006200650020005000440046002065874ef69069752865bc9ad854c18cea76845370524d5370523786557406300260a853ef4ee54f7f75280020004100630072006f0062006100740020548c002000410064006f00620065002000520065006100640065007200200035002e003000204ee553ca66f49ad87248672c4f86958b555f5df25efa7acb76840020005000440046002065874ef63002> /CZE <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> /DAN <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> /DEU <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> /ESP <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> /ETI <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> /FRA <FEFF005500740069006c006900730065007a00200063006500730020006f007000740069006f006e00730020006100660069006e00200064006500200063007200e900650072002000640065007300200064006f00630075006d0065006e00740073002000410064006f00620065002000500044004600200070006f0075007200200075006e00650020007100750061006c0069007400e90020006400270069006d007000720065007300730069006f006e00200070007200e9007000720065007300730065002e0020004c0065007300200064006f00630075006d0065006e00740073002000500044004600200063007200e900e90073002000700065007500760065006e0074002000ea0074007200650020006f007500760065007200740073002000640061006e00730020004100630072006f006200610074002c002000610069006e00730069002000710075002700410064006f00620065002000520065006100640065007200200035002e0030002000650074002000760065007200730069006f006e007300200075006c007400e90072006900650075007200650073002e> /GRE <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a stvaranje Adobe PDF dokumenata najpogodnijih za visokokvalitetni ispis prije tiskanja koristite ove postavke. Stvoreni PDF dokumenti mogu se otvoriti Acrobat i Adobe Reader 5.0 i kasnijim verzijama.) /HUN <FEFF004b0069007600e1006c00f30020006d0069006e0151007300e9006701710020006e0079006f006d00640061006900200065006c0151006b00e90073007a00ed007401510020006e0079006f006d00740061007400e100730068006f007a0020006c006500670069006e006b00e1006200620020006d0065006700660065006c0065006c0151002000410064006f00620065002000500044004600200064006f006b0075006d0065006e00740075006d006f006b0061007400200065007a0065006b006b0065006c0020006100200062006500e1006c006c00ed007400e10073006f006b006b0061006c0020006b00e90073007a00ed0074006800650074002e0020002000410020006c00e90074007200650068006f007a006f00740074002000500044004600200064006f006b0075006d0065006e00740075006d006f006b00200061007a0020004100630072006f006200610074002000e9007300200061007a002000410064006f00620065002000520065006100640065007200200035002e0030002c0020007600610067007900200061007a002000610074007400f3006c0020006b00e9007301510062006200690020007600650072007a006900f3006b006b0061006c0020006e00790069007400680061007400f3006b0020006d00650067002e> /ITA <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> /JPN <FEFF9ad854c18cea306a30d730ea30d730ec30b951fa529b7528002000410064006f0062006500200050004400460020658766f8306e4f5c6210306b4f7f75283057307e305930023053306e8a2d5b9a30674f5c62103055308c305f0020005000440046002030d530a130a430eb306f3001004100630072006f0062006100740020304a30883073002000410064006f00620065002000520065006100640065007200200035002e003000204ee5964d3067958b304f30533068304c3067304d307e305930023053306e8a2d5b9a306b306f30d530a930f330c8306e57cb30818fbc307f304c5fc59808306730593002> /KOR <FEFFc7740020c124c815c7440020c0acc6a9d558c5ec0020ace0d488c9c80020c2dcd5d80020c778c1c4c5d00020ac00c7a50020c801d569d55c002000410064006f0062006500200050004400460020bb38c11cb97c0020c791c131d569b2c8b2e4002e0020c774b807ac8c0020c791c131b41c00200050004400460020bb38c11cb2940020004100630072006f0062006100740020bc0f002000410064006f00620065002000520065006100640065007200200035002e00300020c774c0c1c5d0c11c0020c5f40020c2180020c788c2b5b2c8b2e4002e> /LTH <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> /LVI <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> /NLD (Gebruik deze instellingen om Adobe PDF-documenten te maken die zijn geoptimaliseerd voor prepress-afdrukken van hoge kwaliteit. De gemaakte PDF-documenten kunnen worden geopend met Acrobat en Adobe Reader 5.0 en hoger.) /NOR <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> /POL <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> /PTB <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> /RUM <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> /RUS <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> /SKY <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> /SLV <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> /SUO <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> /SVE <FEFF0041006e007600e4006e00640020006400650020006800e4007200200069006e0073007400e4006c006c006e0069006e006700610072006e00610020006f006d002000640075002000760069006c006c00200073006b006100700061002000410064006f006200650020005000440046002d0064006f006b0075006d0065006e007400200073006f006d002000e400720020006c00e4006d0070006c0069006700610020006600f60072002000700072006500700072006500730073002d007500740073006b00720069006600740020006d006500640020006800f600670020006b00760061006c0069007400650074002e002000200053006b006100700061006400650020005000440046002d0064006f006b0075006d0065006e00740020006b0061006e002000f600700070006e00610073002000690020004100630072006f0062006100740020006f00630068002000410064006f00620065002000520065006100640065007200200035002e00300020006f00630068002000730065006e006100720065002e> /TUR <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> /UKR <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> /ENU (Use these settings to create Adobe PDF documents best suited for high-quality prepress printing. Created PDF documents can be opened with Acrobat and Adobe Reader 5.0 and later.) >> /Namespace [ (Adobe) (Common) (1.0) ] /OtherNamespaces [ << /AsReaderSpreads false /CropImagesToFrames true /ErrorControl /WarnAndContinue /FlattenerIgnoreSpreadOverrides false /IncludeGuidesGrids false /IncludeNonPrinting false /IncludeSlug false /Namespace [ (Adobe) (InDesign) (4.0) ] /OmitPlacedBitmaps false /OmitPlacedEPS false /OmitPlacedPDF false /SimulateOverprint /Legacy >> << /AddBleedMarks false /AddColorBars false /AddCropMarks false /AddPageInfo false /AddRegMarks false /ConvertColors /ConvertToCMYK /DestinationProfileName () /DestinationProfileSelector /DocumentCMYK /Downsample16BitImages true /FlattenerPreset << /PresetSelector /MediumResolution >> /FormElements false /GenerateStructure false /IncludeBookmarks false /IncludeHyperlinks false /IncludeInteractive false /IncludeLayers false /IncludeProfiles false /MultimediaHandling /UseObjectSettings /Namespace [ (Adobe) (CreativeSuite) (2.0) ] /PDFXOutputIntentProfileSelector /DocumentCMYK /PreserveEditing true /UntaggedCMYKHandling /LeaveUntagged /UntaggedRGBHandling /UseDocumentProfile /UseDocumentBleed false >> ] >> setdistillerparams << /HWResolution [2400 2400] /PageSize [612.000 792.000] >> setpagedevice