CN102855498A - Character recognition method and device - Google Patents
Character recognition method and device Download PDFInfo
- Publication number
- CN102855498A CN102855498A CN2011101928273A CN201110192827A CN102855498A CN 102855498 A CN102855498 A CN 102855498A CN 2011101928273 A CN2011101928273 A CN 2011101928273A CN 201110192827 A CN201110192827 A CN 201110192827A CN 102855498 A CN102855498 A CN 102855498A
- Authority
- CN
- China
- Prior art keywords
- character
- identified
- tuple
- proper vector
- convex closure
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Abstract
The invention relates to a character recognition method and device. The character recognition method comprises the following steps of: selecting a tetrad formed by four points in sequence from a convex hull polygon of a character to be recognized by a cross-ratio; converting the character to be recognized into a perspective invariant coordinate system which is defined by the tetrad; extracting features from the converted character to be recognized so as to obtain feature vectors of the character to be recognized; checking records which are matched with the obtained feature vectors of the character to be recognized in a character class table which is stored in advance, and voting character classes corresponding to the checked records; and repeating the steps on different tetrads on the convex hull polygon of the character to be recognized for preset times, thereby determining the character class having the most votes as a character recognition result.
Description
Technical field
The present invention relates to area of pattern recognition, relate more specifically to a kind of character identifying method and device.
Background technology
It is a very important problem that character under the large perspective distortion is identified, because the character of identification perspective distortion is the basis of a lot of practical applications, and the character of perspective distortion extensively is present in our daily life, for example, the character recognition under the real scene.
In order to address this problem, a kind of basic skills is at first to correct the image of perspective distortion to the front elevation picture, then the image after correcting is carried out traditional OCR (optical character identification) identification.But this method is subject to concrete restriction of using, as requires to exist profile and character lines and some structures etc.Therefore the technician begins to be absorbed in each monocase is identified.
How much hash algorithms (Geometric Hashing (GH)) are general object recognition algorithms based on model, and this algorithm also can be suitable for when object has experienced various conversion and only had partial information to exist.The advantage of how much hash algorithms is simply parallel processing, and the ability that also can work when only having partial information.Therefore, how much hash algorithms are used in the object identification and three-dimensional body identification of affined transformation.
Be similar to a kind of classic method that transmission model is identification perspective distortion character with affine Transform Model.M.Iwamura, T.Tsuji, the people such as A.Horimatsu and K.Kise has improved hash algorithm how much in the article that is entitled as " Realtime camera-based recognition of characters and pictograms " that ICDAR in 2009 delivers, and has proposed a kind of camera to be taken the real time algorithm that character is identified.This algorithm adopts affine model, in order to make up affine invariant geomagnetic coordinates, needs 3 coordinate points (3 tuple) as the base of coordinate system.Adopt affine invariant, namely center of gravity and Area Ratio (area ratio) reduce the degree of freedom of 3 tuples.Yet, since affine model only the size of object than the enough little condition of the distance between object and the camera under, namely perspective distortion hour just can be considered to the approximate of perspective model, and when perspective distortion is larger, above-mentionedly is similar to no longer establishment.So, in order to identify the character under the large perspective distortion, need the new method of research and development.
The existing method of another kind of identification perspective distortion character is for each character shape conclusion of the business comparison (cross ratio spectrum), and the double ratio spectrum by current character relatively and the double ratio of template character compose identification character (article that is entitled as " Character recognition under severe perspective distortion " of delivering at ICPR in 2008 referring to Linlin Li and Chew Lim Tan).The defective of this method be the required time of character recognition along with the growth of the categorical measure of character to be identified linear growth.Therefore the use of this method in the more application of the classification of character to be identified is restricted.
Summary of the invention
According to a first aspect of the invention, provide a kind of character identifying method, having comprised: the four-tuple of utilizing double ratio to select four orderly points to consist of from the convex closure polygon of character to be identified; With character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple; Extract the proper vector that feature obtains character to be identified the character to be identified after conversion; In pre-stored character class table, search the record that mates with the proper vector of the character to be identified that obtains, vote to the corresponding character class of the record that finds; Repeat the above-mentioned steps pre-determined number for the different four-tuple on the convex closure polygon of character to be identified; And the maximum character class of poll that will obtain to vote is defined as character identification result.
According to a second aspect of the invention, provide a kind of character recognition device, having comprised: selected cell is arranged to the four-tuple of utilizing double ratio to select four orderly points to consist of from the convex closure polygon of character to be identified; Converter unit is arranged to character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple; Extraction unit is arranged to and extracts the proper vector that feature obtains character to be identified the character to be identified after conversion; The ballot unit is arranged to and searches the record that mates with the proper vector of the character to be identified that obtains in pre-stored character class table, votes to the corresponding character class of the record that finds; Repeat control module, be arranged to for the different four-tuple on the convex closure polygon of character to be identified and repeat the above-mentioned steps pre-determined number; And determining unit, be arranged to and be defined as character identification result with obtaining the maximum character class of ballot poll.
In addition, embodiments of the invention also provide the computer program that is used for realizing said method.
In addition, embodiments of the invention also provide at least computer program of computer-readable medium form, record on it for the computer program code of realizing said method.
By the present invention, can identify the character under the large perspective distortion, use the method for affine model to compare with tradition, discrimination of the present invention is higher.
In addition, method of the present invention still can be identified when partial information lacks.
The character of all right Division identification different fonts of method of the present invention.
By below in conjunction with the detailed description of accompanying drawing to most preferred embodiment of the present invention, these and other advantage of the present invention will be more obvious.
Description of drawings
With reference to below in conjunction with the explanation of accompanying drawing to the embodiment of the invention, can understand more easily above and other purpose of the present invention, characteristics and advantage.Parts in the accompanying drawing are just in order to illustrate principle of the present invention.In the accompanying drawings, same or similar technical characterictic or parts will adopt same or similar Reference numeral to represent.In the accompanying drawing:
Fig. 1 shows the process flow diagram according to the method that is used for identification character of the embodiment of the invention;
Fig. 2 A shows convex closure polygon and the in-profile of character ' H ';
Fig. 2 B shows the diagram of the intersection point of line segment between on the convex closure polygon of character ' H ' 2 and in-profile;
Fig. 2 C shows the diagram of the character ' H ' that perspective distortion has occured;
Fig. 2 D shows the diagram that the sawtooth on the in-profile of line segment between on the convex closure polygon of character ' H ' 2 and character intersects;
The character ' H ' that Fig. 3 A shows among Fig. 2 B and the 2C transforms to the diagram of having an X-rayed under the invariant geomagnetic coordinates;
The invalid four-tuple that Fig. 3 B shows selection transforms to the diagram of having an X-rayed under the invariant geomagnetic coordinates;
Fig. 4 A shows the diagram that the perspective invariant geomagnetic coordinates is divided into 4*4 square;
Fig. 4 B shows normalized histogram;
Fig. 4 C shows the diagram of a record in the Hash table;
Fig. 5 A-5C shows respectively diagram under the large perspective distortion, that damage and character disappearance;
Fig. 6 shows the composition frame chart according to the device that is used for identification character of the embodiment of the invention;
Fig. 7 shows the composition frame chart of the extraction unit in the identification character device; And
Fig. 8 shows the schematic block diagram that can be used for implementing according to the computing machine of the method and apparatus of the embodiment of the invention.
Embodiment
Embodiments of the invention are described with reference to the accompanying drawings.The element of describing in an accompanying drawing of the present invention or a kind of embodiment and feature can combine with element and the feature shown in one or more other accompanying drawing or the embodiment.Should be noted that for purpose clearly, omitted expression and the description of parts that have nothing to do with the present invention, known to persons of ordinary skill in the art and processing in accompanying drawing and the explanation.
Fig. 1 shows the process flow diagram that is used for according to an embodiment of the invention the method for identification character.
At first, in step S102, the four-tuple of utilizing double ratio to select four orderly points to consist of from the convex closure polygon of character to be identified.
Key concept about double ratio (Cross Radio) is as follows: if P
0, A, B, P
24 conllinear, then double ratio Cr is defined as follows:
Wherein, the distance between D () represents at 2.
Those skilled in the art as can be known, double ratio is the invariant in the perspective distortion, that is, under any perspective distortion, double ratio Cr (P
0, A, B, P
2) remain unchanged.
The below, describes in detail and how to select four orderly points at the convex closure polygon of character to be identified take character ' H ' as example with reference to Fig. 2 A-2D.
In the present invention, convex closure is polygonal to be defined as follows: the convex closure polygon is a polygon, and this polygonal summit is the convex closure of character.For example, in Fig. 2 A, the convex closure polygon of character ' H ' is represented by solid line, and the in-profile of character ' H ' is illustrated by the broken lines.Double ratio value for two intersection point calculation of the in-profile of the line segment between 2 on the convex closure polygon of character ' H ' and these 2 and character ' H ' remains unchanged under perspective distortion, wherein, when intersection point during more than two, select the first two intersection point to calculate the double ratio value.For example, in Fig. 2 B, P
0, P
1, P
2And P
3Be the point on the convex closure polygon of character ' H ', A and B are line segment P
0P
2With the first two intersection point of the in-profile of character ' H ', C and D are line segment P
1P
3The first two intersection point with the in-profile of character ' H '.P on the convex closure polygon of selected character ' H '
0Point can be determined the P on the convex closure polygon
2Point is so that P
0, A, B, P
2The double ratio value Cr (P of point
0, A, B, P
2) near a predetermined value.
Because crenellated phenomena may appear in the in-profile of character, thereby the intersection point of the in-profile of the line segment between on the convex closure polygon 2 and character may be the intersection point of the sawtooth on the in-profile with character, for example B shown in Fig. 2 D
1And B
2In this case, because B
1And B
2Between distance very short, thereby the double ratio value that calculates is very large.The situation that sawtooth on the in-profile of appearance and character intersects, the reconnaissance criterion can not be selected too large double ratio value.According to inventor's observation in practice, in general, the double ratio value should (1,2] scope in, and adopt the double ratio value near 1.5 can obtain preferably effect.Therefore in practice, the predetermined value of double ratio value chooses 1.5 usually.
As mentioned above, select arbitrarily P
0Point can be determined P
2Point is so that P
0, A, B, P
2Satisfy | 1.5-Cr (P
0, A, B, P
2) | minimum; Select again P
1Point, wherein, P
0, P
1, P
2Satisfy counterclockwise and arrange, can determine P
3Point is so that P
1, C, D, P
3Satisfy | 1.5-Cr (P
1, C, D, P
3) | minimum, so just finished from the convex closure polygon of character to be identified and selected four orderly points.It will be understood by those skilled in the art that also and can make P
0, P
1, P
2Satisfy arranged clockwise.
Fig. 2 C is the diagram that perspective distortion has occured the character ' H ' among Fig. 2 B, some P wherein
0', P
1', P
2', P
3' correspond respectively to the some P among Fig. 2 B
0, P
1, P
2And P
3, A ', B ' and C ', D ' are respectively line segment P
0' P
2' and line segment P
1' P
3' with the first two intersection point of the in-profile of ' H ' that perspective distortion occurs, so A ', B ', C ', D ' correspond respectively to the A among Fig. 2 B, B, C, D.Because double ratio is the perspective invariant, so some P on ' H ' of perspective distortion occurs among Fig. 2 C
0', A ', B ', P
2' double ratio Cr (P
0', A ', B ', P
2') and some P
1', C ', D ', P
3' double ratio Cr (P
1', C ', D ', P
3') remain unchanged.That is to say:
Cr(P
0′，A′，B′，P
2′)＝Cr(P
0，A，B，P
2)；
Cr(P
1′，C′，D′，P
3′)＝Cr(P
1，C，D，P
3)。
In the manner described above, given P
0(P
0') point, just can determine P
2(P
2') point, no matter which type of perspective transform occurs; Equally, given P
1(P
1') point, just can determine P
3(P
3') point.By the four-tuple that such mode is selected, only have two degree of freedom, thereby can computation reduction.
Next, in step S104, with character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple.
In the present invention, will have an X-rayed invariant geomagnetic coordinates and be defined as a square, the foursquare length of side is L, and the center is (xc, yc).Four-tuple P on the character ' H ' in selected Fig. 2 B
0, P
1, P
2, P
3Situation under, character ' H ' transforms to the diagram of perspective under the invariant geomagnetic coordinates as shown in Figure 3A.Four some P among Fig. 2 B
0, P
1, P
2, P
3Be mapped to respectively the I in the perspective invariant geomagnetic coordinates among Fig. 3 A
0(xc-l, yc+l), I
1(xc-l, yc-l), I
2(xc+l, yc-l), I
3(xc+l, yc+l), wherein, l=α * L/2, α ∈ [0,1], that is, and I
0, I
1, I
2And I
3Fall in the square.Because Fig. 2 C is the character ' H ' among Fig. 2 B perspective distortion has occured, at selected corresponding four-tuple P
0', P
1', P
2', P
3' situation under, the character among Fig. 2 C ' H ' transforms to the diagram of perspective under the invariant geomagnetic coordinates also as shown in Figure 3A, and the some P among Fig. 2 C
0', P
1', P
2', P
3' be mapped to respectively too the I in the perspective invariant geomagnetic coordinates among Fig. 3 A
0, I
1, I
2And I
3
P
0, P
1, P
2, P
3(P
0', P
1', P
2', P
3') and I
0, I
1, I
2And I
3These four to the reply can determine unique perspective transformation matrix, according to this perspective transformation matrix, can with character transformation to be identified to the perspective invariant geomagnetic coordinates under.
The method of determining perspective transformation matrix is known, pertinent literature has Multiple View Geometry in Computer Vision.Richard Hartley and Andrew Zisserman, Cambridge University Press, 2004, do not do at this and to give unnecessary details.
Preferably, after character transformation to be identified is arrived the perspective invariant geomagnetic coordinates, can be transformed according to the point on the convex closure polygon of character to be identified the validity that the ratio of having an X-rayed under the invariant geomagnetic coordinates is determined selected four-tuple.When the point on the convex closure polygon is transformed the ratio of perspective under the invariant geomagnetic coordinates less than predetermined ratio, can determine that selected four-tuple is invalid.
In one example, if the point on the convex closure polygon of character to be identified be transformed under this perspective invariant geomagnetic coordinates ratio more than or equal 90%, think that then the four-tuple of choosing is effective.
If the point on the convex closure polygon of character to be identified is transformed ratio under this perspective invariant geomagnetic coordinates less than 90%, can determine that then selected four-tuple is invalid, give up this four-tuple, re-start selection.
It will be understood by those skilled in the art that above-mentioned predetermined ratio also can choose other value beyond 90%.
Fig. 3 B shows point on the convex closure polygon of character to be identified and is transformed ratio under the perspective invariant geomagnetic coordinates less than 90% situation, and in this case, selected four-tuple is invalid, will be rejected in identifying.
Next, in step S106, extract the proper vector that feature obtains character to be identified the character to be identified after conversion.
Preferably, can with the perspective invariant geomagnetic coordinates be divided into a plurality of subregions, make up histogram according to the number of pixels of character to be identified in each sub regions, with this histogram as proper vector.For example, the perspective invariant geomagnetic coordinates can be divided into m * m square, wherein m is the integer greater than 1.Like this, sequence number, the ordinate that can to obtain a horizontal ordinate be subregion is the histogram of the number of the character pixels that comprises in the subregion, with this histogram as proper vector fcurrent.Further preferably, can carry out normalization to this histogram, with the histogram after the normalization as proper vector fcurrent.It will be understood by those skilled in the art that also can extract the features such as edge direction, gradient information forms proper vector.
In step S108, in pre-stored character class table, search the record that mates with the proper vector of the character to be identified that obtains, vote to the corresponding character class of the record that finds.
Specifically, comprise character class and characteristic of correspondence vector in every record in the character class table.Wherein, proper vector can be the forms such as histogram, normalized histogram.
In one example, the character class table can be Hash table, and wherein every record in the Hash table also comprises proper vector is wherein carried out that Hash is processed and the index value that obtains.
In this example, proper vector is carried out Hash process, for example, carry out even two-value and quantize, and convert the proper vector that quantizes to an index value bin.It will be understood by those skilled in the art that also and can carry out three grades of multi-stage quantizations such as quantification to proper vector, and convert the proper vector that quantizes to index value bin.
Bin value according to obtaining finds corresponding character class d and proper vector fstored in pre-stored character class table (such as Hash table).
Euclidean distance between the proper vector fstored that stores in the proper vector fcurrent of character more to be identified and the character class table is if ‖ fcurrent-fstored ‖ throws a ticket less than a less predetermined value, then for character class d.
If have many tickets to throw to same character class d for identical bin value, then ignore from this bin value throwing other polls to same character class d, that is, only calculate a ticket.
Fig. 4 A shows will have an X-rayed invariant geomagnetic coordinates and is divided into 4*4 square and obtains proper vector as example.Fig. 4 B shows the character pixels number in each square is carried out the histogram as proper vector that normalization obtains.The even two-value of proper vector is quantized, and convert the proper vector that quantizes to index value bin, namely 40944 in the Hash table shown in Fig. 4 C.According to this value, can find character class d and corresponding proper vector in the Hash table.
For the character class table among the step S108, can before carrying out character recognition, make up in the following manner above-mentioned character class table at learning phase for each the template character that is used for study: at first, extract the proper vector of template character, comprising: the four-tuple of utilizing double ratio from the convex closure polygon of template character, to select four orderly points to consist of; The template character is transformed in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple; Extract the proper vector that feature obtains the template character the template character after conversion.The step of the proper vector of extraction template character is similar with the step of the proper vector that above obtains character to be identified, does not do at this and gives unnecessary details.
Need to prove, after the template character being transformed to the perspective invariant geomagnetic coordinates, be transformed according to the point on the convex closure polygon of template character the validity that the ratio of having an X-rayed under the invariant geomagnetic coordinates is determined selected four-tuple.If selected four-tuple is invalid, then give up this four-tuple, re-start selection.Here the method for determining the method for validity of selected four-tuple and the above validity of determining selected four-tuple for character to be identified is similar, does not do at this and gives unnecessary details.
After the proper vector of having extracted the template character, the character class of template character and the proper vector that obtains are left in the character class table as a record.Then, repeat from the convex closure polygon of template character, to select the four-tuple of four orderly somes formations until obtain a plurality of steps of a record in the character class table, until all four-tuple on the convex closure polygon of traversal template character.Thereby made up the character class table that comprises each template character.
It should be noted that at this, how generate character class table although described in the above, but those skilled in the art is to be understood that, for the method that is used for identification character according to the embodiment of the invention, only needing an aforesaid character class table pre-stored gets final product, and need not to be concerned about how the character class table generates.
Next, in step S110, repeat the above-mentioned steps pre-determined number for the different four-tuple on the convex closure polygon of character to be identified.Here, pre-determined number can be predefined number of times, also can be to carry out reconnaissance every a point or every two points in the point on the convex closure polygon, carries out above-mentioned steps S102-S108,, finishes single ballot that is.
At last, in step S112, be defined as character identification result with obtaining the maximum character class of ballot poll.
Above-mentioned character class also can broad sense use, for example, in the situation that multiple font, font A-Arial and A-Calibri can be regarded as a class, also can be regarded as two classes and come Division identification.
According to character identifying method of the present invention, the character under the large perspective distortion shown in Fig. 5 A, and the character shown in Fig. 5 B and the 5C has in the situation of some damages or disappearance, can carry out character recognition.
Fig. 6 shows the block diagram according to the device that is used for identification character of the embodiment of the invention.Character recognition device 600 comprises selected cell 602, converter unit 604, extraction unit 606, ballot unit 608, repeats control module 610 and determining unit 612.Selected cell 602 is arranged to the four-tuple of utilizing double ratio to select four orderly points to consist of from the convex closure polygon of character to be identified.Converter unit 604 is arranged to character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple.Extraction unit 606 is arranged to and extracts the proper vector that feature obtains character to be identified the character to be identified after conversion.Ballot unit 608 is arranged to and searches the record that mates with the proper vector of the character to be identified that obtains in pre-stored character class table, votes to the corresponding character class of the record that finds.Repeat control module 610, be arranged to for the different four-tuple on the convex closure polygon of character to be identified and repeat the above-mentioned steps pre-determined number.Determining unit 612 is arranged to and is defined as character identification result with obtaining the maximum character class of ballot poll.
Similarly, the character class table can be pre-stored at the device for identification character, also can make up the character class table at learning phase for each the template character that is used for study before carrying out character recognition.Here make up the mode of character class table, similar with above mode referring to figs. 1 through making up the character class table among the character identifying method embodiment of the present invention of Fig. 4 description, do not repeat them here.
Alternatively, character recognition device 600 also comprises: the judging unit (not shown), be arranged to the ratio that is transformed under the perspective invariant geomagnetic coordinates according to the point on the convex closure polygon of character to be identified and whether judge less than predetermined ratio whether selected four-tuple is invalid, if determine that selected four-tuple is invalid, then give up this four-tuple.
Alternatively, the character class table adopts Hash table, and wherein every record in the Hash table also comprises proper vector is wherein carried out that Hash is processed and the index value that obtains.
Alternatively, extraction unit 606 comprises cuts apart subelement 6062 and histogram structure subelement 6064, cutting apart subelement 6062 is arranged to the perspective invariant geomagnetic coordinates is divided into a plurality of subregions, histogram structure subelement 6064 is arranged to according to the number of pixels of character to be identified in each sub regions and makes up histogram, and histogram is used as proper vector.
Can with reference to the embodiments of the invention of describing in conjunction with Fig. 1 to Fig. 4, be not described in detail here about the operation of the various piece of character recognition device 600 and the details of function.
Need to prove at this, Fig. 6 and character recognition device 600 shown in Figure 7 and the structure of component units thereof only are exemplary, and those skilled in the art can make amendment to Fig. 6 and structured flowchart shown in Figure 7 as required.
Ultimate principle of the present invention has below been described in conjunction with specific embodiments, but, it is to be noted, for those of ordinary skill in the art, can understand whole or any steps or the parts of method and apparatus of the present invention, can be in the network of any calculation element (comprising processor, storage medium etc.) or calculation element, realized with hardware, firmware, software or their combination, this is those of ordinary skills in the situation that read the basic programming skill that explanation of the present invention uses them and just can realize.
Therefore, purpose of the present invention can also be by realizing in any program of calculation element operation or batch processing.Described calculation element can be known fexible unit.Therefore, purpose of the present invention also can be only by providing the program product that comprises the program code of realizing described method or device to realize.That is to say, such program product also consists of the present invention, and the storage medium that stores such program product also consists of the present invention.Obviously, described storage medium can be any storage medium that develops in any known storage medium or future.
In the situation that realize embodiments of the invention by software and/or firmware, from storage medium or network to the computing machine with specialized hardware structure, for example multi-purpose computer 800 shown in Figure 8 is installed the program that consists of this software, this computing machine can be carried out various functions etc. when various program is installed.
Fig. 8 shows the schematic block diagram that can be used for implementing according to the computing machine of the method and apparatus of the embodiment of the invention.In Fig. 8, CPU (central processing unit) (CPU) 801 carries out various processing according to the program of storage in the ROM (read-only memory) (ROM) 802 or from the program that storage area 808 is loaded into random access memory (RAM) 803.In RAM 803, also store as required data required when CPU 801 carries out various processing etc.CPU 801, ROM 802 and RAM 803 are connected to each other via bus 804.Input/output interface 805 also is connected to bus 804.
Following parts are connected to input/output interface 805: importation 806 (comprising keyboard, mouse etc.), output 807 (comprise display, such as cathode-ray tube (CRT) (CRT), liquid crystal display (LCD) etc., with loudspeaker etc.), storage area 808 (comprising hard disk etc.), communications portion 809 (comprising that network interface unit is such as LAN card, modulator-demodular unit etc.).Communications portion 809 is processed such as the Internet executive communication via network.As required, driver 810 also can be connected to input/output interface 805.Detachable media 811 can be installed on the driver 810 as required such as disk, CD, magneto-optic disk, semiconductor memory etc., so that the computer program of therefrom reading is installed in the storage area 808 as required.
In the situation that realize above-mentioned series of processes by software, such as detachable media 811 program that consists of software is installed such as the Internet or storage medium from network.
It will be understood by those of skill in the art that this storage medium is not limited to shown in Figure 8 wherein has program stored therein, distributes separately to provide the detachable media 811 of program to the user with equipment.The example of detachable media 811 comprises disk (comprising floppy disk (registered trademark)), CD (comprising compact disc read-only memory (CD-ROM) and digital universal disc (DVD)), magneto-optic disk (comprising mini-disk (MD) (registered trademark)) and semiconductor memory.Perhaps, storage medium can be hard disk that comprises in ROM 802, the storage area 808 etc., computer program stored wherein, and be distributed to the user with the equipment that comprises them.
The present invention also proposes a kind of program product that stores the instruction code that machine readable gets.When described instruction code is read and carried out by machine, can carry out above-mentioned method according to the embodiment of the invention.
Correspondingly, being used for carrying the above-mentioned storage medium that stores the program product of the instruction code that machine readable gets is also included within of the present invention open.Described storage medium includes but not limited to floppy disk, CD, magneto-optic disk, storage card, memory stick etc.
In the above in the description to the specific embodiment of the invention, can in one or more other embodiment, use in same or similar mode for the feature that a kind of embodiment is described and/or illustrated, combined with the feature in other embodiment, or the feature in alternative other embodiment.
Should emphasize, term " comprises/comprise " existence that refers to feature, key element, step or assembly when this paper uses, but does not get rid of the existence of one or more further feature, key element, step or assembly or additional.
In addition, the time sequencing of describing during method of the present invention is not limited to is to specifications carried out, also can according to other time sequencing ground, carry out concurrently or independently.The execution sequence of the method for therefore, describing in this instructions is not construed as limiting technical scope of the present invention.
Although the above discloses the present invention by the description to specific embodiments of the invention,, should be appreciated that, all above-mentioned embodiment and example all are illustrative, and not restrictive.Those skilled in the art can design various modifications of the present invention, improvement or equivalent in the spirit and scope of claims.These modifications, improvement or equivalent also should be believed to comprise in protection scope of the present invention.
Claims (10)
1. character identifying method comprises:
The four-tuple of utilizing double ratio to select four orderly points to consist of from the convex closure polygon of character to be identified;
With described character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple;
Extract the proper vector that feature obtains character to be identified the character described to be identified after conversion;
In pre-stored character class table, search the record that mates with the proper vector of the character to be identified that obtains, vote to the corresponding character class of the record that finds;
Repeat the above-mentioned steps pre-determined number for the different four-tuple on the convex closure polygon of described character to be identified; And
Be defined as character identification result with obtaining the maximum character class of ballot poll.
2. the method for claim 1, wherein said character class indicator makes up in the following manner to each the template character that is used for study:
The four-tuple of utilizing double ratio from the convex closure polygon of template character, to select four orderly points to consist of;
Described template character is transformed in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple;
Extract the proper vector that feature obtains described template character the described template character after conversion;
The character class of described template character and the proper vector of the described template character that obtains are left in the described character class table as a record; And
Repeat above-mentioned processing until travel through all four-tuple on the convex closure polygon of described template character, in order to make up described character class table.
3. method as claimed in claim 1 or 2, wherein after with the step of described character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple, also comprise: whether judge less than predetermined ratio whether selected four-tuple is invalid according to the ratio that the point on the convex closure polygon of described character to be identified is transformed under the described perspective invariant geomagnetic coordinates, if determine that selected four-tuple is invalid, then give up this four-tuple, re-execute the step of utilizing double ratio to select the four-tuple of four orderly somes formations from the convex closure polygon of character to be identified.
4. method as claimed in claim 1 or 2, wherein said character class table adopts Hash table, and every record in the wherein said Hash table also comprises proper vector is wherein carried out that Hash is processed and the index value that obtains.
5. method as claimed in claim 3, wherein extracting the step that feature obtains the proper vector of character to be identified the character described to be identified after conversion comprises: described perspective invariant geomagnetic coordinates is divided into a plurality of subregions, make up histogram according to the number of pixels of described character to be identified in each sub regions, with described histogram as described proper vector.
6. character recognition device comprises:
Selected cell is arranged to the four-tuple of utilizing double ratio to select four orderly points to consist of from the convex closure polygon of character to be identified;
Converter unit is arranged to described character transformation to be identified in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple;
Extraction unit is arranged to and extracts the proper vector that feature obtains character to be identified the character described to be identified after conversion;
The ballot unit is arranged to and searches the record that mates with the proper vector of the character to be identified that obtains in pre-stored character class table, votes to the corresponding character class of the record that finds;
Repeat control module, be arranged to for the different four-tuple on the convex closure polygon of described character to be identified and repeat the above-mentioned steps pre-determined number; And
Determining unit is arranged to and is defined as character identification result with obtaining the maximum character class of ballot poll.
7. device as claimed in claim 6, wherein, described character class indicator makes up in the following manner to each the template character that is used for study:
The four-tuple of utilizing double ratio from the convex closure polygon of template character, to select four orderly points to consist of;
Described template character is transformed in the perspective invariant geomagnetic coordinates of being determined by selected four-tuple;
Extract the proper vector that feature obtains described template character the described template character after conversion;
The character class of described template character and the proper vector of the described template character that obtains are left in the described character class table as a record; And
Repeat above-mentioned processing until travel through all four-tuple on the convex closure polygon of described template character, in order to make up described character class table.
8. such as claim 6 or 7 described devices, also comprise:
Judging unit, be arranged to the ratio that is transformed under the described perspective invariant geomagnetic coordinates according to the point on the convex closure polygon of described character to be identified and whether judge less than predetermined ratio whether selected four-tuple is invalid, if determine that selected four-tuple is invalid, then give up this four-tuple.
9. such as claim 6 or 7 described devices, wherein said character class table adopts Hash table, and every record in the wherein said Hash table also comprises proper vector is wherein carried out that Hash is processed and the index value that obtains.
10. device as claimed in claim 8, wherein said extraction unit comprises cuts apart subelement and histogram structure subelement, the described subelement of cutting apart is arranged to described perspective invariant geomagnetic coordinates is divided into a plurality of subregions, described histogram structure subelement is arranged to according to the number of pixels of described character to be identified in each sub regions and makes up histogram, and described histogram is used as described proper vector.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110192827.3A CN102855498B (en) | 2011-07-01 | 2011-07-01 | Character identifying method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201110192827.3A CN102855498B (en) | 2011-07-01 | 2011-07-01 | Character identifying method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102855498A true CN102855498A (en) | 2013-01-02 |
| CN102855498B CN102855498B (en) | 2016-08-31 |
Family
ID=47402074
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201110192827.3A Active CN102855498B (en) | 2011-07-01 | 2011-07-01 | Character identifying method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102855498B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105184328A (en) * | 2015-08-17 | 2015-12-23 | 浪潮软件集团有限公司 | Method and device for identifying image |
| CN105474235A (en) * | 2013-06-28 | 2016-04-06 | 谷歌公司 | Comparing extracted card data using continuous scanning |
| CN117085969A (en) * | 2023-10-11 | 2023-11-21 | 中国移动紫金(江苏)创新研究院有限公司 | Artificial intelligence industrial vision detection method, device, equipment and storage medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101373519A (en) * | 2007-08-20 | 2009-02-25 | 富士通株式会社 | Device and method for recognizing character |
| CN101536035A (en) * | 2006-08-31 | 2009-09-16 | 公立大学法人大阪府立大学 | Image recognition method, image recognition device, and image recognition program |
-
2011
- 2011-07-01 CN CN201110192827.3A patent/CN102855498B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101536035A (en) * | 2006-08-31 | 2009-09-16 | 公立大学法人大阪府立大学 | Image recognition method, image recognition device, and image recognition program |
| CN101373519A (en) * | 2007-08-20 | 2009-02-25 | 富士通株式会社 | Device and method for recognizing character |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105474235A (en) * | 2013-06-28 | 2016-04-06 | 谷歌公司 | Comparing extracted card data using continuous scanning |
| US10152647B2 (en) | 2013-06-28 | 2018-12-11 | Google Llc | Comparing extracted card data using continuous scanning |
| US10515290B2 (en) | 2013-06-28 | 2019-12-24 | Google Llc | Comparing extracted card data using continuous scanning |
| US10963730B2 (en) | 2013-06-28 | 2021-03-30 | Google Llc | Comparing extracted card data using continuous scanning |
| CN105184328A (en) * | 2015-08-17 | 2015-12-23 | 浪潮软件集团有限公司 | Method and device for identifying image |
| CN105184328B (en) * | 2015-08-17 | 2018-11-27 | 浪潮金融信息技术有限公司 | A kind of method and device identifying image |
| CN117085969A (en) * | 2023-10-11 | 2023-11-21 | 中国移动紫金(江苏)创新研究院有限公司 | Artificial intelligence industrial vision detection method, device, equipment and storage medium |
| CN117085969B (en) * | 2023-10-11 | 2024-02-13 | 中国移动紫金(江苏)创新研究院有限公司 | Artificial intelligence industrial vision detection method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN102855498B (en) | 2016-08-31 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107944020B (en) | Face image searching method and device, computer device and storage medium | |
| WO2020125216A1 (en) | Pedestrian re-identification method, device, electronic device and computer-readable storage medium | |
| WO2018166116A1 (en) | Car damage recognition method, electronic apparatus and computer-readable storage medium | |
| US8607331B2 (en) | Captcha image authentication method and system | |
| CN103403739B (en) | For the method and system of movement images | |
| JP2014232533A (en) | System and method for ocr output verification | |
| US8503777B2 (en) | Geometric feature based image description and fast image retrieval | |
| Türkyılmaz et al. | License plate recognition system using artificial neural networks | |
| CN106126585B (en) | The unmanned plane image search method combined based on quality grading with perceived hash characteristics | |
| CN102592148A (en) | Face identification method based on non-negative matrix factorization and a plurality of distance functions | |
| Lepsøy et al. | Statistical modelling of outliers for fast visual search | |
| CN104850822B (en) | Leaf identification method under simple background based on multi-feature fusion | |
| CN112199462A (en) | Cross-modal data processing method and device, storage medium and electronic device | |
| CN109582813B (en) | Retrieval method, device, equipment and storage medium for cultural relic exhibit | |
| CN107169425A (en) | A kind of recognition methods of item property and device | |
| CN104915673A (en) | Object classification method and system based on bag of visual word model | |
| CN109426828B (en) | Point cloud classification method, device, equipment and storage medium | |
| CN108537223B (en) | License plate detection method, system and equipment and storage medium | |
| CN109726725A (en) | The oil painting writer identification method of heterogeneite Multiple Kernel Learning between a kind of class based on large-spacing | |
| CN111695609A (en) | Target damage degree determination method, target damage degree determination device, electronic device, and storage medium | |
| US20130121598A1 (en) | System and Method for Randomized Point Set Geometry Verification for Image Identification | |
| US8942515B1 (en) | Method and apparatus for image retrieval | |
| CN102855498A (en) | Character recognition method and device | |
| CN113192028B (en) | Quality evaluation method and device for face image, electronic equipment and storage medium | |
| CN105844299B (en) | A kind of image classification method based on bag of words |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |