WO1997041529A1 - Verfahren zum klassifizieren und wiedererkennen von mustern, wobei eine signatur durch das glätten eines polygonzugs erzeugt wird - Google Patents
Verfahren zum klassifizieren und wiedererkennen von mustern, wobei eine signatur durch das glätten eines polygonzugs erzeugt wird Download PDFInfo
- Publication number
- WO1997041529A1 WO1997041529A1 PCT/DE1997/000836 DE9700836W WO9741529A1 WO 1997041529 A1 WO1997041529 A1 WO 1997041529A1 DE 9700836 W DE9700836 W DE 9700836W WO 9741529 A1 WO9741529 A1 WO 9741529A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- point
- polygon
- polyline
- property
- pattern
- Prior art date
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
Definitions
- the present invention relates to a method for classifying and recognition of patterns and insbeson particular a method with which a pattern with the aid of a classification vector, which differs according to the fractality of each pattern are classified "can.
- Pattern recognition is a central problem in many technical fields. Such pattern recognition is intended to enable m-dimensional objects to be detected by means of a data processing device in such a way that the data processing device can determine with as high a precision as possible which m -dimensional object the respective pattern is to be assigned.
- a highly precise pattern recognition of m-dimensional objects would, for example, allow vehicles of any kind to be automatically steered, so that accidents caused by human error could largely be avoided.
- the production of automatic machines or robots equipped with intelligent sensors would also be no problem with a highly precise pattern recognition.
- Further areas of application lie, for example, in the detection of contour lines in the form of red-green-blue color data, a symbol when entering a symbol via a pressure-sensitive graphics tablet or mono audio data, etc. Many others Applications are also conceivable.
- the object of the present invention is accordingly to provide a method for classifying and recognizing patterns, which can be used with any type of object and which is based on a classification vector which differs according to the fractality of the object, patterns classified and recognized.
- a pattern of an m-dimensional object to be classified and recognized is created in the form of an m-dimensional polygon.
- a property is recorded that clearly defines a Drawing a respective point of the polygon to the point preceding it and the point following it reflects it.
- These properties for each point are linked to one another in order to obtain an overall property which uniquely characterizes the polygon.
- the polyline is smoothed.
- the steps of capturing, linking and smoothing (k1) are repeated below, where k represents an integer.
- a signature for the pattern is generated using the overall properties obtained, which are present in the number k. This signature is used for comparing with signatures of known patterns in order to determine a degree of equality between the compared signatures.
- the method can be modified such that it either does not take into account certain features of the pattern to be classified and recognized, or only takes into account, for example, that to scale Differences between the pattern to be classified and recognized and the pattern of a known object are disregarded. This means that the method can be suitably modified for every conceivable application.
- the invention is explained in more detail below on the basis of the description of exemplary embodiments with reference to the drawing.
- FIG. 2 shows a graphic representation of a change in length as a function of repetition steps of smoothing according to the first exemplary embodiment of the present invention
- FIGS. 3 (a) and 3 (b) show the signature of the images shown in FIGS. 3 (a) and 3 (b) to illustrate the function of the first exemplary embodiment of the present invention
- FigL 5 is a polygon to illustrate the
- FIG. 6 shows a first possibility for detecting a property of a respective point on the polygon course according to the second th embodiment of the present invention
- Fig. 7 shows a second way of detecting a property of each
- Fig. 9 shows the image of a when smoothing the
- the first exemplary embodiment of the present invention explains the functioning of the method according to the invention on the basis of a two-dimensional polyline in the plane.
- the process is not limited to this; rather, an advantage of the method according to the invention is that it can be used with any m-dimensional polygonal lines (1 ⁇ m ⁇ oo).
- a m-dimensional structure is made from any m-dimensional structure using a method known per se.
- the two-dimensional polygon course in this exemplary embodiment consists of the letter "M".
- a desired trait of the traverse is detected.
- the desired property to be detected with respect to the letter "M” can be, for example, a measure of length. Consequently, a length dimension is defined for each point of the polyline so that it clearly reflects a relationship between the point that precedes the respective point, the respective point and the point that follows the respective point.
- the length dimension of the respective point is determined in such a way that each point is assigned half the distance to its predecessor and half the distance to its successor, that is to say as the Euclidean length. This is carried out for each point of the polygon.
- the measures of length recorded in this way are linked to form a total measure of length that characterizes the polygon.
- the total length dimension is determined from a summation of the recorded length dimensions for each point.
- a so-called smoothing of the polygon is then carried out.
- the new polyline Pl can be determined, for example, using the following equation:
- r ⁇ (k) ⁇ r 0 (kl) + 2r 0 (k) + r 0 (k + l)> / 4 (1)
- r ⁇ (k) denotes the coordinate of the point of the new polyline P] _
- rn (kl) denotes the coordinate of the point preceding the respective point on the preceding polyline PQ
- rn (k + l) denotes the coordinate of the point following the respective point on the previous polygon PQ.
- FIG. 2 An overall property thus results for each of the runs, which in this case represents a length.
- the 4 overall properties obtained from FIGS. 1 (a) to 1 (d) are plotted graphically with regard to the number of steps.
- the illustration shown in FIG. 2 represents the signature for the letter “M”, which in this case represents the object or pattern to be classified and recognized.
- This signature is compared with signatures of known patterns, in order to achieve a degree of equality between the compared signatures. The signature can thus be used to determine what the original object or pattern was.
- an object of the shape of the letter "M" to be classified and recognized is recognized as being equal to a known object of the shape of the letter "M", that is larger or smaller by an arbitrary factor the object to be classified and recognized is, that is, the two objects have a different scale.
- a length measure was used as a property. However, this is not absolutely necessary. Any other dimension suitable for the object or any combination of suitable dimensions can be used.
- An outstanding property of the method according to the invention is that it differentiates according to the fractality of the polygon present, which characterizes an object or a pattern.
- this classification of the object not only a single measure for the fractality but an entire set of numbers is used, that is to say the aforementioned vector with k components.
- FIGS. 4 (a) and 4 (b) show two different patterns to be classified.
- the pattern 3 (a) "fractal" to the pattern in Fig. 3 (b).
- FIGS. 4 (a) and 4 (b) show the result of performing the method described in the first exemplary embodiment on these patterns.
- the signature in FIG. 4 (a) being the pattern in FIG. 3 ( a) is associated and the signature in Fig. 4 (b) is associated with the pattern in Fig. 3 (b). It is thus clear from FIGS. 4 (a) and 4 (b) that the previous method distinguishes very well between regular or smooth and irregular or fractal structures.
- a two-dimensional pattern is again used for simplification.
- a two-dimensional polygon is generated for this two-dimensional pattern.
- 5 shows the polygon used in the second embodiment.
- This polyline has the number n of points x (0) to x (n-l).
- 6 shows a first possibility for detecting a property of a respective point of the polygon.
- the Euclidean length is assigned to each respective point of the polygon. This means that each point x (i) is assigned half the distance to its predecessor x (i-l) and half the distance to its successor x (i + l). This assignment is expressed by the following equation (2):
- ⁇ (i, x (i)) Jj ⁇
- ⁇ (i, x (i)) represents the property or the dimension of the point x (i) (0 ⁇ i ⁇ n).
- Fig. 7 The square root of the area of the triangle is used here, which is spanned by the point x (i) and its neighbors x (il) and x ⁇ i + l) in a suitable manner. This can be expressed by the following equation (3):
- the property or the dimension ⁇ (i, x (i)) of each respective point of the polygon can be clearly calculated from the point x (i) and its neighbors x (i-l) and x (i + l).
- the properties or dimensions ⁇ (i, x (i)) are positive for nonidentical points x (il), x (i) and x ⁇ i + l), as is the case, for example, in FIGS. 6 and 7 see is.
- the method is furthermore to be able to recognize as identical patterns which differ in features other than the size of the property, appropriate conditions must be derived for this.
- This condition can be taken into account by using the condition expressed in equation (2). In general, this means that a condition is used that the property ⁇ (i, x (i)) only of
- any conceivable combination of conditions can be used in this method, as a result of which the method according to the invention can be used for any type of pattern which can be classified and recognized, and each pattern by means of a any number of features of each point can be classified.
- ⁇ (a, b) represents the measure of a distance between points a and b on the polygon.
- ⁇ ( ⁇ M ⁇ ) ⁇ (0, n-1) (5)
- the polygon is smoothed.
- the smoothing is carried out in such a way that an averaging process is carried out which is dependent on a scale parameter ⁇ ( ⁇ > 0).
- This scale parameter represents the area of the averaging on the path through the ordered point set M.
- the averaging method depends on the previously recorded dimensions or properties ⁇ (0), ..., ⁇ (nl) ⁇ on the polygon.
- the averaging method used in this second exemplary embodiment is shown in the following equation (7): ⁇ x (j) * p ( ⁇ (iJ), ⁇ ) + ⁇ x (j) * p ( ⁇ (i, j) J ⁇ )
- x '(i) represents the point on the new polyline derived from point x (i) on the original polyline, where i in equation (7) is from 0 to n-1 runs and n represents the number of points of the polygon.
- the expression p ( ⁇ (i, j), ⁇ ) in equation (7) represents a weight function.
- the ordered point set M ' ⁇ x' (0), ..., x '(i ), ..., x '(nl) ⁇ , i.e.
- the new smoothed polyline with each point x' (i) of the new smoothed polyline from the coordinates x (il), x (i), x (i + l) the point of the original polyline is obtained.
- the point x '(i) of the point set M' ultimately represents a normalized and averaged sum of the point x (i) on the original polygon course and the one it represents ⁇ neighboring points x (il), x (i + l).
- the process steps are repeated with the new polyline (k1) from the acquisition of a property or a measure for the polyline, in order to thus obtain k overall properties or to achieve dimensions.
- the total dimension of the points of the new polygon train depends on the scale parameter ⁇ used, the scale parameter ⁇ being different for each run of the method. This means that the pattern is ultimately characterized by a number of k total dimensions, that is to say a vector with k components, the k total dimensions or components of the vector thus representing scale-dependent overall dimensions of the respective averaged point set M '.
- An advantageous embodiment of the method according to the invention further consists in that instead of “sharp” sizes for the vectors of known patterns, “unsharp” quantities are used, the distribution of which can be learned from the pattern recognition method. This makes it possible to significantly increase the probability of detection.
- a practical application of the method according to the invention in which a pattern recognition with low computing power (also for fractal patterns) is desirable, consists, for example, in creating a cost-saving but extremely secure access control in which only a commercially available data processing device (personal computer ) and a graphics tablet for handwritten input of symbols are required.
- these entered symbols define patterns (polygons in m-dimensional space) by means of which the user can be clearly identified.
- the generalization of the traditional password to these patterns has the decisive advantage that no access can be gained simply by knowing the form.
- this method is not bound to country-specific letters, since only patterns and non-semantic contents are checked.
- the abundance of patterns in time and space makes it possible to dispense with other access control mechanisms (e.g. chip or magnetic cards), since these are difficult to protect against loss.
Abstract
Description
Claims
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE59706842T DE59706842D1 (de) | 1996-04-25 | 1997-04-24 | Verfahren zum klassifizieren und wiedererkennen von mustern, wobei eine signatur durch das glätten eines polygonzugs erzeugt wird |
US08/973,748 US6229920B1 (en) | 1996-04-25 | 1997-04-24 | Method of classifying and recognizing patterns by additive contour extraction and averaging |
AU30877/97A AU3087797A (en) | 1996-04-25 | 1997-04-24 | Method of classification and recognition of patterns according to which a signature is produced by smoothing polygon outline |
EP97925829A EP0843864B1 (de) | 1996-04-25 | 1997-04-24 | Verfahren zum klassifizieren und wiedererkennen von mustern, wobei eine signatur durch das glätten eines polygonzugs erzeugt wird |
DE19780359T DE19780359D2 (de) | 1996-04-25 | 1997-04-24 | Verfahren zum Klassifizieren und Wiedererkennen von Mustern, wobei eine Signatur durch das Glätten eines Polygonzugs erzeugt wird |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
DE19616565 | 1996-04-25 | ||
DE19616565.2 | 1996-04-25 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO1997041529A1 true WO1997041529A1 (de) | 1997-11-06 |
Family
ID=7792439
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/DE1997/000836 WO1997041529A1 (de) | 1996-04-25 | 1997-04-24 | Verfahren zum klassifizieren und wiedererkennen von mustern, wobei eine signatur durch das glätten eines polygonzugs erzeugt wird |
Country Status (5)
Country | Link |
---|---|
US (1) | US6229920B1 (de) |
EP (1) | EP0843864B1 (de) |
AU (1) | AU3087797A (de) |
DE (2) | DE59706842D1 (de) |
WO (1) | WO1997041529A1 (de) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108304862A (zh) * | 2018-01-08 | 2018-07-20 | 武汉大学 | 一种基于小波变换的地图建筑物多边形模式识别方法 |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE10047189C1 (de) * | 2000-09-23 | 2002-02-21 | Bosch Gmbh Robert | Verfahren zur Insassenklassifikation mit einer Sitzmatte im Fahrzeugsitz |
DE10328322A1 (de) * | 2003-06-24 | 2005-01-27 | Massen Machine Vision Systems Gmbh | Überwachung des Farbeindrucks von mehrfarbig gemusterten Produkten |
US8107710B2 (en) | 2008-05-23 | 2012-01-31 | University Of Rochester | Automated placental measurement |
Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19507564A1 (de) * | 1995-03-03 | 1996-09-05 | Delphi Systemsimulation Gmbh | Verfahren zur Mustererkennung und Verfahren zum Erstellen eines n-dimensionalen Objektes |
Family Cites Families (19)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5619657B2 (de) * | 1973-09-10 | 1981-05-08 | ||
JPS5827551B2 (ja) * | 1979-05-18 | 1983-06-10 | 日本電信電話株式会社 | オンライン手書き文字認識方式 |
US4365235A (en) * | 1980-12-31 | 1982-12-21 | International Business Machines Corporation | Chinese/Kanji on-line recognition system |
US4542412A (en) * | 1982-02-04 | 1985-09-17 | Shaken Co., Ltd. | Method for compressing character or pictorial image data |
JPS60253368A (ja) * | 1983-11-10 | 1985-12-14 | Dainippon Screen Mfg Co Ltd | 複製画像記録表示等に於けるjag除去方法 |
JPS60136892A (ja) * | 1983-12-26 | 1985-07-20 | Hitachi Ltd | オンライン手書き図形認識装置 |
KR900001696B1 (ko) * | 1984-11-09 | 1990-03-19 | 가부시기가이샤 히다찌세이사꾸쇼 | 화상처리장치의 제어방법 |
GB2183118B (en) * | 1985-11-19 | 1989-10-04 | Sony Corp | Image signal processing |
US4791679A (en) * | 1987-12-26 | 1988-12-13 | Eastman Kodak Company | Image character enhancement using a stroke strengthening kernal |
JPH01231183A (ja) * | 1988-03-10 | 1989-09-14 | Mitsubishi Electric Corp | 画像処理装置における直線性判定装置 |
JPH02162475A (ja) * | 1988-12-15 | 1990-06-22 | Dainippon Screen Mfg Co Ltd | 画像輪郭修正方法 |
JP2833092B2 (ja) * | 1990-01-26 | 1998-12-09 | ブラザー工業株式会社 | 輪郭線データの圧縮装置および復元装置 |
US5218649A (en) * | 1990-05-04 | 1993-06-08 | U S West Advanced Technologies, Inc. | Image enhancement system |
JPH056431A (ja) * | 1991-06-27 | 1993-01-14 | Matsushita Electric Ind Co Ltd | 輪郭線特徴点検出装置 |
JPH0535806A (ja) * | 1991-07-29 | 1993-02-12 | Toshiba Corp | 線分解釈方式 |
EP0625853B1 (de) * | 1993-05-21 | 1999-03-03 | Nippon Telegraph And Telephone Corporation | Bewegtbildkoder und -dekoder |
WO1995009416A1 (en) * | 1993-09-30 | 1995-04-06 | Apple Computer, Inc. | Continuous reference adaptation in a pattern recognition system |
JP3347508B2 (ja) * | 1995-02-24 | 2002-11-20 | キヤノン株式会社 | 撮像画像処理装置および撮像画像処理方法 |
TW397951B (en) * | 1995-06-05 | 2000-07-11 | Motorola Inc | Method and microprocessor for preprocessing handwriting having characters composed of a preponderance of straight line segments |
-
1997
- 1997-04-24 DE DE59706842T patent/DE59706842D1/de not_active Expired - Lifetime
- 1997-04-24 WO PCT/DE1997/000836 patent/WO1997041529A1/de active IP Right Grant
- 1997-04-24 EP EP97925829A patent/EP0843864B1/de not_active Expired - Lifetime
- 1997-04-24 DE DE19780359T patent/DE19780359D2/de not_active Expired - Fee Related
- 1997-04-24 US US08/973,748 patent/US6229920B1/en not_active Expired - Lifetime
- 1997-04-24 AU AU30877/97A patent/AU3087797A/en not_active Withdrawn
Patent Citations (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
DE19507564A1 (de) * | 1995-03-03 | 1996-09-05 | Delphi Systemsimulation Gmbh | Verfahren zur Mustererkennung und Verfahren zum Erstellen eines n-dimensionalen Objektes |
Non-Patent Citations (1)
Title |
---|
LEU J -G ET AL: "Polygonal approximation of 2-D shapes through boundary merging", PATTERN RECOGNITION LETTERS, APRIL 1988, NETHERLANDS, vol. 7, no. 4, ISSN 0167-8655, pages 231 - 238, XP002040927 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108304862A (zh) * | 2018-01-08 | 2018-07-20 | 武汉大学 | 一种基于小波变换的地图建筑物多边形模式识别方法 |
Also Published As
Publication number | Publication date |
---|---|
US6229920B1 (en) | 2001-05-08 |
DE59706842D1 (de) | 2002-05-08 |
DE19780359D2 (de) | 1998-11-05 |
EP0843864A1 (de) | 1998-05-27 |
AU3087797A (en) | 1997-11-19 |
EP0843864B1 (de) | 2002-04-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
EP0780002B1 (de) | Verfahren und vorrichtung zur rekonstruktion von in rasterform vorliegenden linienstrukturen | |
DE102006059663B4 (de) | Vorrichtung, Verfahren und Computerprogramm zum Identifizieren eines Verkehrszeichens in einem Bild | |
DE69927738T2 (de) | Verfahren und Vorrichtung zum Editieren von skizzierten Kurven | |
EP1316057B1 (de) | Auswerten von kantenrichtungsinformation | |
WO2000010119A1 (de) | Verfahren zum erkennen von objekten in digitalisierten abbildungen | |
EP1994499B1 (de) | Herstellung einer linienartigen oder flächenartigen form | |
WO2019057402A1 (de) | Verfahren zur erkennung einer objektinstanz und/oder orientierung eines objekts | |
DE3515159A1 (de) | Bildverarbeitungseinrichtung | |
DE19531392C1 (de) | Verfahren zur Erzeugung einer Graphrepräsentation von Bildvorlagen | |
EP2082357B1 (de) | Vorrichtung, verfahren und computerprogramm zur erkennung von schriftzeichen in einem bild | |
WO2008034599A2 (de) | Verfahren und vorrichtung zur bildverarbeitung | |
WO1997041529A1 (de) | Verfahren zum klassifizieren und wiedererkennen von mustern, wobei eine signatur durch das glätten eines polygonzugs erzeugt wird | |
DE1916109A1 (de) | Anordnung zur Bestimmung von Eckpunkten aus einer Vielzahl von eine empirische Form beschreibenden Datenpunkten | |
DE3417407C2 (de) | Kreisgenerator für eine graphische Anzeigeeinheit | |
EP0870275B1 (de) | Verfahren zur mustererkennung und verfahren zum erstellen eines n-dimensionalen objektes | |
DE60008201T2 (de) | Übersetzung von Daten mit elektronischen Bildern | |
WO2002023492A2 (de) | Verfahren zur klassifizierung von dokumenten | |
EP2893510B1 (de) | Verfahren und bildverarbeitungsanlage zum entfernen eines visuellen objektes aus einem bild | |
EP0829809A1 (de) | Verfahren zur Prozessvisualisierung | |
DE19649692C1 (de) | Verfahren zur Verifikation eines Musterschriftzuges mit Hilfe eines Referenzschriftzugs | |
DE2113264C3 (de) | Verfahren zum automatischen Erkennen zweidimensionaler Muster mit Hilfe von Momenten zweiter Ordnung | |
EP2128797B1 (de) | Verfahren und Vorrichtung zur Musterverarbeitung | |
AT413248B (de) | Verfahren zum vergleich eines anfragefingerprints mit einem referenzfingerprint | |
DE69824546T2 (de) | Ermittlung von Kartekoordinaten | |
Brechner | Automatic coccolith classification and extraction of morphological features in SEM images |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
AK | Designated states |
Kind code of ref document: A1 Designated state(s): AL AM AT AU AZ BA BB BG BR BY CA CH CN CU CZ DE DK EE ES FI GB GE HU IL IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MD MG MK MN MW MX NO NZ PL PT RO RU SD SE SG SI SK TJ TM TR TT UA UG US UZ VN AM AZ BY KG KZ MD RU TJ TM |
|
AL | Designated countries for regional patents |
Kind code of ref document: A1 Designated state(s): GH KE LS MW SD SZ UG AT BE CH DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF |
|
WWE | Wipo information: entry into national phase |
Ref document number: 1997925829 Country of ref document: EP |
|
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: 08973748 Country of ref document: US |
|
WWP | Wipo information: published in national office |
Ref document number: 1997925829 Country of ref document: EP |
|
REF | Corresponds to |
Ref document number: 19780359 Country of ref document: DE Date of ref document: 19981105 |
|
WWE | Wipo information: entry into national phase |
Ref document number: 19780359 Country of ref document: DE |
|
NENP | Non-entry into the national phase |
Ref document number: 97538457 Country of ref document: JP |
|
NENP | Non-entry into the national phase |
Ref country code: CA |
|
WWG | Wipo information: grant in national office |
Ref document number: 1997925829 Country of ref document: EP |