WO2006027158A1 - Verfahren zur zuordnung eines digitalen bildes in eine klasse eines klassifizierungssystems - Google Patents
Verfahren zur zuordnung eines digitalen bildes in eine klasse eines klassifizierungssystems Download PDFInfo
- Publication number
- WO2006027158A1 WO2006027158A1 PCT/EP2005/009427 EP2005009427W WO2006027158A1 WO 2006027158 A1 WO2006027158 A1 WO 2006027158A1 EP 2005009427 W EP2005009427 W EP 2005009427W WO 2006027158 A1 WO2006027158 A1 WO 2006027158A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- class
- image
- values
- shape
- numerical
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/18—Extraction of features or characteristics of the image
- G06V30/186—Extraction of features or characteristics of the image by deriving mathematical or geometrical properties from the whole image
- G06V30/188—Computation of moments
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Definitions
- the invention relates to a method for assigning a digital image to a class of a classification system.
- Optical error detection methods were carried out in the past by quality assurance personnel. These have considered the object to be checked or an image representation of the object to be checked and recognized possible errors.
- weld seams are checked for types of defects by means of X-ray images, such as, for example, cracks, insufficient penetration, binding defects, slag, slag lines, pores, hose pores, wort I, root defects, heavy metal inclusions and edge offset.
- X-ray images such as, for example, cracks, insufficient penetration, binding defects, slag, slag lines, pores, hose pores, wort I, root defects, heavy metal inclusions and edge offset.
- it is known to view radioscopic images of castings in order to detect defects in the casting, for example foreign inclusions, gas inclusions, voids, such as yarn pockets or spongy voids, cracks or core supports.
- the industry standard EN 1435 describes the classification system for weld defects.
- the errors detected in welds and identified by means of X-ray images are divided into the 30 different classes, for example classes for the faults such as cracking
- the faults such as cracking
- each letter forms its own class, so that, for example, there are 26 classes for the capital letter alphabet, namely for the characters (A, B, C, D, E, F, G, H, I 1 J, K, L, M, N, O, P, Q, R, S, T, U, V, W, X, Y, Z).
- OCR technologies Optical Character Recognition
- the published European patent specifications 0 854 435 B1 and 0 649 113 B1 cover, for example, the technical field of optical character recognition (Optical Character Recognition).
- Image preprocessing is the computer-aided improvement of the quality (processing: denoising, smoothing) of the respective digital image for easier visual perception of the information content of this image for the viewer.
- Image analysis is the computer-aided evaluation of the information content of the respective digital image by automatic and reproducible structuring, recognition and understanding of this image.
- the analysis of image sequences means the computer-aided evaluation of the information content of the respective sequence of digital images by automatic and reproducible structuring, recognition and understanding of all individual images of this sequence and by automatic and reproducible understanding of the context of the sequence of individual images follow-up.
- Image archiving is the computer-aided compression and storage of digital images together with indexing search descriptors from a controlled vocabulary.
- Imaging means the computer-aided generation of synthetic graphics and digital images for the visualization and explanation of the information content of complex processes on image and symbol level for the human observer.
- the technique of associating the contents of digital images with a class of a classification system is a method of image analysis. This can be subdivided into three subareas: segmentation, object recognition and image understanding.
- Segmentation is understood to mean the automatic and reproducible structuring of the respective digital image by separating the objects relevant for the analysis of the image from each other and from the background of the image.
- Object recognition is the automatic and reproducible zierbare classification of the separated objects.
- Image comprehension can be understood as the automatic and reproducible interpretation of the respective digital image by context evaluation of the classified, separated objects.
- the technique of associating digital images with a class of a classification system is a method of object recognition.
- the object recognition can be understood as a subarea of the pattern recognition and indeed as the subarea of the pattern recognition, which recognizes as pattern only planar objects in images.
- the images are regularly displayed by an image composed of pixels, whereby the content of each pixel and its position in the image must be known in order to display the image.
- the images can be stored in color images, grayscale images and binary images, with binary images as the content attribute, for example, only taking the values 0 and 1 for black and white.
- malsvektoren ⁇ ( ⁇ ', ⁇ 2, ⁇ 3, ⁇ - *, ⁇ s, ⁇ , ⁇ 7), whose coordinates are dimensionless shape features done so that in particular size differences between the objects to be recognized and the objects used for the preparation of Ver ⁇ equal table meaningless. Furthermore, within the set of dimensionless shape features .phi. / By the coordinate reference to the feature vector .phi., An unambiguous order with respect to the relevance of the features for the object recognition in digital image processing is given, so that it is clear that the first feature .phi..sub.r is the most important is.
- the object of the invention is to propose a method for assigning the content of a digital image to a class of a classification system, with which also characters of more complex form can be reliably recognized.
- the invention is based on the idea of determining the image to be analyzed of a predetermined number of numerical shape features ⁇ m with m as count index running from 1 to F, where ⁇ m is a transformed expression of dimensionless, scaled, normalized, centralized pola ⁇ ren moment p m is.
- These mutually independent shape features ⁇ m can be compared for matching the content of the digital image with stored in a table values for these shape features. If the values of all determined F shape features ⁇ m coincide with the F shape features ⁇ m stored for a class in the table, the image content of the analyzed image belongs to this class. Due to digital In this case, it is preferable to work with approximate values, so that a class assignment is also already output when the calculated F shape features ⁇ m approximately coincide with the F stored features ⁇ m of a class.
- the numerical shape features ⁇ m proposed according to the invention for image analysis are independent of one another in such a way that a large number of shape features can be established without the dependency of the shape features on one another arises. As a result, an unambiguous assignment of the image contents to be recognized to a designated class can be achieved.
- the method according to the invention is, in particular, independent of the relative position of the content to be recognized relative to the receiving device. Also, for example, rotated by 60 ° or 180 ° objects can be clearly assigned zuge ⁇ .
- the method is based on the computation of a sequence of F functionally independent, dimensionless features of the separated, bounded content in the respective image.
- the image is displayed in a conventional manner by N pixels, wherein a pixel in a predetermined coordinate system at the location (Xj, y j), and the image is on the coordinates (0,0) to (Xi ma ⁇ , yim a x) and imax is the maximum number of pixels in the direction of the x-coordinate and ymax is the maximum number of pixels in the direction of the y-coordinate and each pixel is assigned a content attribute data [j, i].
- the content attribute is for a binary-represented image in which the respective pixel content assumes, for example, either the value 1 or 0 for black or white, for example a single value deposited in a table, and data [j, i] representatively for the value in this table at the location associated with the pixel.
- the content attribute data [j, i] is representative of a vector which contains these three values for the respective pixel contains.
- Data [j, i] can also be representative of other vectors, if other color representations are used, or greyscale representations, data [j, i] can also be representative of the magnitude of such a vector, if a multi-color representation is converted from a multi-color representation, for example an RGB representation, into a greyscale or even binary representation before the use of the classification method according to the invention.
- data [j, i] can also stand for the individual value of the red representation, or green representation, or blue representation in the pixel.
- the classification method is then carried out, for example, exclusively on the basis of a representation, for example the red representation, the method then being carried out in the same way as for the binary representation above. It is then also possible to use binary values 1 and 0 for data [j, i] at the image point, where 1 stands, for example, for red and 0 for empty.
- the classification method can be performed in parallel for the different color representations, ie parallel for a binary red representation, a binary green representation and a binary blue representation. This increases the accuracy of the classification.
- W 0 1 (Abf * ⁇ * ⁇ ⁇ (y -0,5) * ⁇ / ⁇ 4 /, /]
- ⁇ a width of the pixel in x-coordinates direction
- ⁇ b width of the pixel in y-coordinates direction data
- ö, i] content attribute of the pixel at the position (y j , X 1 )
- m continuous number from 1 to F.
- the predetermined coordinate system is particularly preferably a Cartesian coordinate system, since most digital images define the pixels via a Cartesian coordinate system.
- an image content is defined by the arrangement of pixels of the same content attribute.
- the F-determined shape features of the image content yield a feature vector in a bounded, F-dimensional subarea (unit hypercube) of the F-dimensional feature space.
- the content classification finally takes place by problem-specific clustering of this n-dimensional unit hypercube.
- the classification system can be, for example, a given industry standard, such as EN 1435.
- each person can form his own class.
- the F shape features ⁇ m which characterize the fingerprint or the iris image of the person to be recognized, are filed.
- the image of the iris taken by a recording unit for example a camera, is analyzed by the method according to the invention, wherein the F shape features ⁇ m of the recorded iris are calculated and compared with the shape feature values stored in the table. If there is an (approximate) agreement with all the values of the features ⁇ m of a class, then the system recognizes the person who is characterized by this class.
- a method of least squares for example according to Gauss, can preferably be used.
- the aforementioned method steps can be carried out for a plurality of groups with F numerical shape features ⁇ m , for example in a group for values of a red representation, in a group for values of a green representation and in a group for values of a blue representation.
- the aforementioned method steps can also be carried out on content attributes data [j, i] which contain as vector the individual values of the individual color representations. Divisional calculation operations are then preferably carried out on the amounts of the vectors.
- the shape feature ⁇ m is the transformation
- the shape characteristic to be compared with the values stored in the table is preferably the shape feature ⁇ m obtained by means of the aforementioned transformation.
- the series of F shape features can be subjected to an orthogonalization method, as is carried out, for example, according to E. Schmidt.
- the shape characteristic to be compared can be converted in particular such that a series of F shape features ⁇ i, ⁇ 2, ⁇ 3 for a circle. ⁇ 4, ⁇ s ••• ⁇ F with values of 1, 0,0,0,0 ... 0 results.
- the number F of the shape features is increased until the values of the shape features having the highest order numbers m decrease in all classes with increasing order number.
- the values of the respective shape feature ⁇ m determined for the at least 29 samples per class can be arithmetically averaged to determine a value to be used for this class for this feature.
- the number F can also be determined by way of a rotational ellipse determination method.
- Such "cluster processes” are described, for example, in H. Niemann, Klasstechnik von Muster, Springer Verlag, Berlin, 1983, page 200ff.
- the inventive method for assigning the content of a digital image in a class of a classification system is preferably used in the optical inspection of components, in particular in the optical surface chenenspektion. Furthermore, the method can be used for quality assurance, texture, shape and contour analysis, photogrammetry, character and character recognition, person recognition, robotic vision or the evaluation of radiographic or radioscopic images, ultrasound images and nuclear spin Tomography are used.
- the images with respect to which the object recognition is carried out are "optical" images from the spectral range of visible light or radiographic or radioscopic Images or even synthetic imagery.
- the method can therefore also be used in the field of optical surface inspection, such as in quality assurance, texture, shape and contour analysis, photogrammetry, character and character recognition, person recognition, robotic vision or the evaluation of radiographic or radioscopic images, ultrasound images and magnetic resonance tomography.
- Fig. 1 shows three different representations of a first character to be recognized
- Fig. 2 shows three representations of a second character to be recognized
- FIG. 3 shows three representations of a third character to be recognized.
- the letters A, B and C are respectively in three forms of representation i) normal, ii) normal but rotated by 90 °, iii) same orientation as normal but smaller. Furthermore, in addition to the centered orientations shown in the figures, one positioning on the left and one positioning on the right were examined.
- ⁇ a width of the pixel in x-coordinate direction
- ⁇ b width of the pixel in y-coordinate direction
Abstract
Description
Claims
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
US11/574,698 US7876964B2 (en) | 2004-09-03 | 2005-09-01 | Method for associating a digital image with a class of a classification system |
AT05776279T ATE440342T1 (de) | 2004-09-03 | 2005-09-01 | Verfahren zur zuordnung eines digitalen bildes in eine klasse eines klassifizierungssystems |
CN2005800334923A CN101048784B (zh) | 2004-09-03 | 2005-09-01 | 用于将数字图像对应到分类系统的类别中的方法 |
DE502005007957T DE502005007957D1 (de) | 2004-09-03 | 2005-09-01 | Verfahren zur zuordnung eines digitalen bildes in |
EP05776279A EP1789909B1 (de) | 2004-09-03 | 2005-09-01 | Verfahren zur zuordnung eines digitalen bildes in eine klasse eines klassifizierungssystems |
Applications Claiming Priority (4)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE102004043149.3 | 2004-09-03 | ||
DE102004043149 | 2004-09-03 | ||
DE102005001224.8 | 2005-01-10 | ||
DE102005001224A DE102005001224A1 (de) | 2004-09-03 | 2005-01-10 | Verfahren zur Zuordnung eines digitalen Bildes in eine Klasse eines Klassifizierungssystems |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2006027158A1 true WO2006027158A1 (de) | 2006-03-16 |
Family
ID=35395937
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/EP2005/009427 WO2006027158A1 (de) | 2004-09-03 | 2005-09-01 | Verfahren zur zuordnung eines digitalen bildes in eine klasse eines klassifizierungssystems |
Country Status (6)
Country | Link |
---|---|
US (1) | US7876964B2 (de) |
EP (1) | EP1789909B1 (de) |
CN (1) | CN101048784B (de) |
AT (1) | ATE440342T1 (de) |
DE (2) | DE102005001224A1 (de) |
WO (1) | WO2006027158A1 (de) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR101461879B1 (ko) * | 2012-12-17 | 2014-11-13 | 현대자동차 주식회사 | 용접검사 장치 및 방법 |
CN104519416B (zh) * | 2014-11-28 | 2018-04-06 | 四川长虹电器股份有限公司 | 一种数字电视竖直指纹的实现方法和系统 |
Family Cites Families (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
EP0854435B1 (de) | 1991-11-04 | 2002-01-16 | Canon Kabushiki Kaisha | Gerät und Verfahren zur optischen Zeichenerkennung |
US5539840A (en) | 1993-10-19 | 1996-07-23 | Canon Inc. | Multifont optical character recognition using a box connectivity approach |
US5963670A (en) * | 1996-02-12 | 1999-10-05 | Massachusetts Institute Of Technology | Method and apparatus for classifying and identifying images |
US6404923B1 (en) * | 1996-03-29 | 2002-06-11 | Microsoft Corporation | Table-based low-level image classification and compression system |
US6470094B1 (en) * | 2000-03-14 | 2002-10-22 | Intel Corporation | Generalized text localization in images |
US6580824B2 (en) * | 2001-05-14 | 2003-06-17 | Hewlett-Packard Development Company, L.P. | Classification of photos with sepia tones |
-
2005
- 2005-01-10 DE DE102005001224A patent/DE102005001224A1/de not_active Withdrawn
- 2005-09-01 DE DE502005007957T patent/DE502005007957D1/de active Active
- 2005-09-01 CN CN2005800334923A patent/CN101048784B/zh not_active Expired - Fee Related
- 2005-09-01 EP EP05776279A patent/EP1789909B1/de not_active Not-in-force
- 2005-09-01 WO PCT/EP2005/009427 patent/WO2006027158A1/de active Application Filing
- 2005-09-01 AT AT05776279T patent/ATE440342T1/de not_active IP Right Cessation
- 2005-09-01 US US11/574,698 patent/US7876964B2/en not_active Expired - Fee Related
Non-Patent Citations (3)
Title |
---|
BELKASIM S O ET AL: "PATTERN RECOGNITION WITH MOMENT INVARIANTS: A COMPARATIVE STUDY AND NEW RESULTS", PATTERN RECOGNITION, ELSEVIER, KIDLINGTON, GB, vol. 24, no. 12, January 1991 (1991-01-01), pages 1117 - 1138, XP000248340, ISSN: 0031-3203 * |
COEN, GÜNTHER: "Eine neue Methode der Musterklassierung in der digitalen Bildverarbeitung", 18 May 2004, DEUTSCHE GESELLSCHAFT FÜR ZERSTÖRUNGSFREIE PRÜFUNG, DACH-JAHRESTAGUNG 2004, SALZBURG, XP009058053 * |
MINDRU F ET AL: "Moment invariants for recognition under changing viewpoint and illumination", COMPUTER VISION AND IMAGE UNDERSTANDING, ACADEMIC PRESS, SAN DIEGO, CA, US, vol. 94, no. 1-3, April 2004 (2004-04-01), pages 3 - 27, XP004501957, ISSN: 1077-3142 * |
Also Published As
Publication number | Publication date |
---|---|
EP1789909A1 (de) | 2007-05-30 |
DE102005001224A1 (de) | 2006-03-09 |
US7876964B2 (en) | 2011-01-25 |
ATE440342T1 (de) | 2009-09-15 |
DE502005007957D1 (de) | 2009-10-01 |
CN101048784A (zh) | 2007-10-03 |
US20090214123A1 (en) | 2009-08-27 |
EP1789909B1 (de) | 2009-08-19 |
CN101048784B (zh) | 2010-05-05 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
DE60123378T2 (de) | Digitales Bildverarbeitungsverfahren mit verschiedenen Arbeitsweisen zum Erkennen von Augen | |
DE102007013971B4 (de) | Verfahren und Vorrichtung zur Ermittlung einer Zellkontur einer Zelle | |
DE102009023756B4 (de) | Verfahren zum Herkunftsnachweis und zur Urheberschaft von Bildern | |
WO1996007976A1 (de) | Verfahren zur rekonstruktion von in rasterform vorliegenden linienstrukturen | |
DE4133590A1 (de) | Verfahren zur klassifikation von signalen | |
DE112019005143T5 (de) | System zur co-registrierung medizinischer bilder unter verwendung eines klassifikators | |
DE102012002321B4 (de) | Verfahren zur Erkennung eines vorgegebenen Musters in einem Bilddatensatz | |
DE102021100444A1 (de) | Mikroskopiesystem und verfahren zum bewerten von bildverarbeitungsergebnissen | |
DE102019133685A1 (de) | Informationsverarbeitungssystem und -verfahren | |
DE102005025220B4 (de) | Gerät, Verfahren und Programm zum Beseitigen von Poren | |
DE102005049017B4 (de) | Verfahren zur Segmentierung in einem n-dimensionalen Merkmalsraum und Verfahren zur Klassifikation auf Grundlage von geometrischen Eigenschaften segmentierter Objekte in einem n-dimensionalen Datenraum | |
EP3123393B1 (de) | Verfahren zur optischen erkennung von zeichen | |
DE3246631C2 (de) | Zeichenerkennungsvorrichtung | |
WO2006027158A1 (de) | Verfahren zur zuordnung eines digitalen bildes in eine klasse eines klassifizierungssystems | |
DE102015205502A1 (de) | Bildverarbeitungsverfahren und Bildverarbeitungssystem zum Extrahieren verzerrter kreisförmiger Bildelemente | |
DE19801400A1 (de) | Verfahren und Anordnung zur automatischen Erkennung, Eigenschaftsbeschreibung und Interpretation von Hep-2-Zellmustern | |
DE112019004112T5 (de) | System und verfahren zur analyse mikroskopischer bilddaten und zur erzeugung eines annotierten datensatzes zum training der klassifikatoren | |
DE19834718C2 (de) | Digitale Bildverarbeitung für ein Qualitätskontrollsystem | |
DE102017116882B4 (de) | Druckinspektionsvorrichtung zur optischen Inspektion eines Druckbildes eines Druckobjekts | |
DE4495111C2 (de) | Verfahren zur Bestimmung einer Menge von charakteristischen Merkmalen im Rahmen einer Objekterkennung | |
DE102004018174B4 (de) | Verfahren zur Akquisition von Formen aus Bildern mit Fällen und zum fallbasierten Erkennen von Objekten in digitalen Bildern, Computer-Programm-Produkt und digitales Speichermedium zur Ausführung dieses Verfahrens | |
DE19939732C2 (de) | Verfahren zur automatischen Detektion und Identifikation von Objekten in multispektralen Fernerkundungsdaten | |
DE102022206025A1 (de) | Populationsklassifikation mittels Punktwolken-Merkmalen | |
DE112022001435T5 (de) | Teilsucheinrichtung, teilsuchprogramm und teilsuchverfahren | |
Inkielman et al. | Microscopic image classification based on descriptor analysis |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KM KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NG NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SM SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IS IT LT LU LV MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application | ||
WWE | Wipo information: entry into national phase |
Ref document number: 2005776279 Country of ref document: EP Ref document number: 200580033492.3 Country of ref document: CN |
|
WWP | Wipo information: published in national office |
Ref document number: 2005776279 Country of ref document: EP |
|
WWE | Wipo information: entry into national phase |
Ref document number: 11574698 Country of ref document: US |