CN109753967A - A kind of picture character recognition methods - Google Patents
A kind of picture character recognition methods Download PDFInfo
- Publication number
- CN109753967A CN109753967A CN201811632008.4A CN201811632008A CN109753967A CN 109753967 A CN109753967 A CN 109753967A CN 201811632008 A CN201811632008 A CN 201811632008A CN 109753967 A CN109753967 A CN 109753967A
- Authority
- CN
- China
- Prior art keywords
- region
- picture
- characteristic value
- character
- character recognition
- 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.)
- Pending
Links
Abstract
The present invention is a kind of picture character recognition methods, extracts comprising steps of carrying out 10 regional characteristic values to the Chinese character in multiple fonts, constructs matrix magazine；The recognition accuracy for calculating Figure 10 characteristic area of word of multiple fonts, plurality of specifications, each characteristic area is sorted according to recognition accuracy, final to choose 7 features for combining identification；Feature extraction is carried out to text on picture to be identified, calculates characteristic value；The characteristic value obtained will be calculated to be compared in matrix magazine, export the highest Chinese character of similarity.When the present invention carries out Text region, only with the monochrome pixels feature of picture character, principle is simple, and calculation amount is small, and accuracy rate is high.
Description
Technical field
The present invention relates to field of character recognition more particularly to a kind of picture character recognition methods.
Background technique
Printed Chinese character, which knows method for distinguishing, at present mainly structure model recognition method and statistical pattern recognition method.Structure
The stroke feature of Chinese character, structure point feature, projection properties, wheel are mainly utilized in mode identification method and statistical pattern recognition method
The features such as wide feature, histogram feature.
But the tactic pattern and statistical model existing characteristics used in currently available technology extracts difficult, some algorithm complexity
The problems such as degree is high, computationally intensive.
Summary of the invention
Technical problem to be solved by the present invention lies in provide a kind of picture character recognition methods, utilize the overall situation of Chinese character
Printed Chinese character is identified with the assemblage characteristics of local monochrome pixels, this method principle is simple, and calculation amount is small, and identification is quasi-
True rate is high.
The present invention provides a kind of picture character recognition methods, comprising steps of
S1, characteristics extraction is carried out to the Chinese character in multiple fonts, constructs matrix magazine；
S2, feature extraction is carried out to text on picture to be identified, calculates characteristic value；
S3, the characteristic value for calculating acquisition is compared in matrix magazine, exports the highest type matrix Chinese character of similarity.
Further, the step 1 specifically includes:
S101, region division is carried out to a Chinese character, which is divided into N number of different region, N is whole greater than 2
Number；
S102, the characteristic value for calculating each region, the characteristic value in N number of region is stored；
S103, the n-quadrant according to corresponding to the multiple fonts of Chinese characters all in character library and a variety of font sizes characteristic value structure
Build matrix magazine.
Further, step 1 further include:
S104, the recognition accuracy that single region is calculated according to the characteristic value in each region, choose it is therein M as than
Compared with region, M≤N；
Further, N is 10 in the step 101, and region specifically includes entirety, middle section, top half, lower half
Divide, left-half, right half part, upper left hand corner section, top-right part, lower left corner part, lower right corner part.
Further, the step 2 specifically includes:
S201, the text in picture to be identified is extracted with graphic form；
S202, region division is carried out to the text picture extracted, text picture is divided into different regions；
S203, selection M region corresponding with comparison domain；
S204, the characteristic value for calculating the M region chosen.
Further, the form in the step 202 when progress region division using rectangle frame, in addition to middle section,
The wire of the rectangle frame in remaining region at least with one of the top of picture, left end, bottom, right end phase
It cuts.
Further, M is 6 to 8.
Further, in the step 3, the Chinese character of output is editable form.The present invention is beneficial compared with prior art
Effect is, when carrying out Text region, acts on strong with calculating difference only with the monochrome pixels feature of picture character, and only
6-8 region is combined calculating, and principle is simple, and calculation amount is small, and accuracy rate is high.
Detailed description of the invention
Fig. 1 is picture character recognition methods flow chart in embodiment of the present invention；
Fig. 2 is the schematic diagram that Chinese character carries out region division in embodiment of the present invention.
Specific embodiment
With reference to embodiments, the present invention is further described in detail, but not limited to this.
A kind of picture character recognition methods of the invention, as shown in Figure 1, comprising steps of
S1, characteristics extraction is carried out to the Chinese character in multiple fonts, constructs matrix magazine；
S2, feature extraction is carried out to text on picture to be identified, calculates characteristic value；
S3, the characteristic value for calculating acquisition is compared in matrix magazine, exports the highest type matrix Chinese character of similarity.
On the basis of above scheme, further, the step 1 is specifically included:
S101, region division is carried out to a Chinese character, which is divided into N number of different region, N is whole greater than 2
Number.Preferably, 10 N, region division is as shown in Fig. 2, specifically include the whole F of Chinese character1, middle section F2, top half F3、
Lower half portion F4, left-half F5, right half part F6, upper left hand corner section F7, top-right part F8, lower left corner part F9, the lower right corner
Part F10.In other embodiments, the region divided to Chinese character is not limited to these parts, may be used also to improve accuracy
To increase more regions.
S102, the characteristic value for calculating each region, the characteristic value in N number of region is stored.
X-Y coordinate is specifically established with plane locating for Chinese character, is laterally X-axis, longitudinal is Y-axis, and (x, y) is pixel
Abscissa and ordinate, then enable
Wherein 0 < x < i, 0 < y < j, formula (1)
Wherein, i is the picture traverse in the region, and j is the picture altitude in the region,
The collection of the black and white values in the region is enabled to be combined into T,
T={ t (x, y) }, formula (2).
Then the characteristic value η in the region is,
Wherein η1For in T 1 number, η2=i × j is the number of element in T.
S103, the n-quadrant according to corresponding to the multiple fonts of Chinese characters all in character library and a variety of font sizes characteristic value structure
Build matrix magazine.
Preferably, construct matrix magazine font include the Song typeface, regular script, lishu, imitation Song-Dynasty-style typeface, etc. lines, the refined multiple fonts such as black, word
Body can also be expanded as needed.Font size include first number (42pt), it is small just (36pt), No.1 (26pt), small by one (24pt),
No. two (22pt), small by two (18pt), No. three (16pt), small by three (15pt), No. four (14pt), small by four (12pt), No. five
(10.5pt), small by five (9pt), No. six (7.5pt), common 14 kinds of font sizes such as small by six (6.5pt).
On the basis of above scheme, further, S104, the knowledge that single region is calculated according to the characteristic value in each region
Other accuracy rate chooses M therein as comparison domain, M≤N.Preferably, M is 6 to 8.
Specifically, choosing one or more Chinese characters from character library as accuracy rate tests Chinese character, by N number of area of step 101
Domain divides accuracy rate test Chinese character, successively calculates the recognition accuracy of each region.
Calculating process is as follows:
Δ η=(η 'n-ηn)2, 0 < n≤N, formula (4),
Wherein, Δ η is the characteristic value η ' that accuracy rate tests n-th of region of Chinese characternWith pair of candidate word a certain in template library
Answer characteristic value ηnDifference.
Δ η is ranked up, the corresponding Chinese character of Δ η minimum value in matrix magazine is exported, if the Chinese character and accuracy rate of output
It is identical to test Chinese character, then identifies success, thinks failure if different.
By taking the Song typeface as an example, if PnFor the recognition accuracy of n-th of region n, then each characteristic area recognition accuracy such as 1 institute of table
Show, wherein unit is percentage in table:
The recognition accuracy (unit %) of each characteristic area of 1 Song typeface of table
| P1 | P2 | P3 | P4 | P5 | P6 | P7 | P8 | P9 | P10 | |
| No. six | 1.83 | 0.69 | 0.89 | 1.14 | 0.96 | 1.28 | 0.35 | 0.52 | 0.51 | 0.62 |
| It is small by five | 2.76 | 1.34 | 1.34 | 1.54 | 1.39 | 1.68 | 0.54 | 0.65 | 0.54 | 0.69 |
| No. five | 3.44 | 1.09 | 1.60 | 2.08 | 1.88 | 2.20 | 0.60 | 0.91 | 0.82 | 1.00 |
| It is small by four | 4.89 | 1.97 | 2.48 | 2.76 | 2.39 | 3.11 | 1.05 | 1.29 | 1.03 | 1.19 |
| No. four | 7.03 | 1.71 | 3.21 | 3.68 | 3.50 | 4.22 | 1.51 | 1.82 | 1.53 | 1.65 |
| It is small by three | 7.03 | 2.80 | 3.61 | 5.09 | 3.99 | 4.62 | 1.45 | 2.07 | 2.02 | 2.17 |
| No. three | 7.60 | 3.45 | 4.59 | 4.45 | 4.67 | 4.50 | 1.96 | 2.10 | 1.97 | 1.82 |
| It is small by two | 12.53 | 4.78 | 6.32 | 7.87 | 7.44 | 7.58 | 2.94 | 3.16 | 3.34 | 3.34 |
| No. two | 18.65 | 7.43 | 11.02 | 10.86 | 10.68 | 11.16 | 4.61 | 5.10 | 5.21 | 4.65 |
| It is small by one | 18.23 | 6.41 | 11.77 | 11.85 | 11.22 | 11.99 | 4.99 | 5.75 | 5.66 | 5.09 |
| No.1 | 22.38 | 9.40 | 13.58 | 14.95 | 14.67 | 14.53 | 5.86 | 6.86 | 7.41 | 6.36 |
| It is small first | 41.87 | 15.29 | 26.24 | 30.24 | 28.40 | 29.79 | 12.48 | 14.55 | 15.52 | 14.15 |
| First number | 53.38 | 23.30 | 35.12 | 40.58 | 37.99 | 37.22 | 16.40 | 18.72 | 22.02 | 19.08 |
| Mean value | 15.51 | 6.13 | 9.37 | 10.55 | 9.94 | 10.30 | 4.21 | 4.88 | 5.20 | 4.75 |
From table 1 we can see that the recognition accuracy in each region is increased with the increase of font size.By taking region 1 as an example,
The recognition accuracy in region 1 is 1.83% under No. six fonts, and the recognition accuracy in region 1 is 53.38% under first font.This says
Bright picture quality is higher, and retrievable feature is more accurate.
The single region recognition rate of the Song typeface is from high to low successively are as follows: F1、F4、F6、F5、F3、F2、F9、F8、F10、F7.If according to
The size in region carries out the grouping in region, and region can be divided into three groups: F1It (account for whole word identification region 100%) is the 1st group,
F3、F4、F5And F6It (account for about whole word identification region 50%) is the 2nd group, F2、F7、F8、F9With region F10(account for about whole word cog region
The 25% of domain) it is the 3rd group.The experimental data of observation multiple fonts shows that the discrimination for being grouped 1 is correct higher than the identification of grouping 2
Rate, be grouped 2 recognition correct rate be higher than grouping 3 recognition correct rate.Under different fonts, the discrimination for being grouped interior each region is accurate
Rate ranking will be different, but the accuracy rate ranking between grouping is consistent.
Since under common font size, the recognition accuracy of single area is insufficient for the needs of practical application.So needing
Multiple regions are chosen as comparison domain.
We successively increase feature quantity according to feature recognition accuracy height, step up recognition accuracy.Make simultaneously
When with n region, the calculation formula of the Δ η of a certain candidate word in Chinese character to be identified and template library are as follows:
Inventor has found that the region chosen recognition accuracy at 6 to 8 is suitble to, higher in accuracy rate at this time and participation meter
The region of calculation is less.By taking Song typeface as an example, the recognition accuracy of multiple regions combination is as shown in table 2, and wherein unit is percentage.
The recognition accuracy (unit %) of 2 Song typeface combination zone of table
| Feature | P 123456 | P1234569 | P 12345689 | P12345678910 |
| No. six | 49.78 | 62.09 | 62.09 | 62.09 |
| It is small by five | 74.06 | 86.11 | 86.13 | 86.13 |
| No. five | 86.24 | 94.24 | 94.24 | 94.24 |
| It is small by four | 96.59 | 98.83 | 98.83 | 98.83 |
| No. four | 98.49 | 99.71 | 99.71 | 99.71 |
| It is small by three | 98.98 | 99.74 | 99.74 | 99.74 |
| No. three | 99.58 | 99.98 | 99.98 | 99.98 |
| It is small by two | 99.89 | 100.00 | 100.00 | 100.00 |
| No. two | 99.97 | 99.98 | 99.98 | 99.98 |
| It is small by one | 99.98 | 99.98 | 99.98 | 99.98 |
| No.1 | 99.98 | 100.00 | 100.00 | 100.00 |
| It is small first | 100.00 | 100.00 | 100.00 | 100.00 |
| First number | 100.00 | 100.00 | 100.00 | 100.00 |
Further progress verifying, when the region of selection is F1、F4、F6、F5、F3、F2、F10At this 7, accuracy rate tests Chinese character
As shown in table 3 in the recognition accuracy of five kinds of different font sizes of seven kinds of fonts, wherein unit is percentage.
The recognition accuracy (unit %) of 3 seven kinds of common fonts of table
| No. six | It is small by five | No. five | It is small by four | No. four | |
| Lishu | 54.54 | 85.93 | 93.10 | 97.57 | 99.31 |
| The Song typeface | 62.09 | 86.13 | 94.24 | 98.83 | 99.71 |
| Imitation Song-Dynasty-style typeface | 69.18 | 82.23 | 92.08 | 97.95 | 99.28 |
| Regular script | 72.92 | 87.27 | 96.87 | 99.06 | 99.77 |
| Black matrix | 80.58 | 94.88 | 96.16 | 99.51 | 99.78 |
| Children's circle | 86.61 | 92.83 | 97.57 | 99.29 | 99.88 |
| Microsoft is refined black | 89.86 | 95.92 | 98.47 | 99.80 | 99.95 |
Recognition accuracy moves closer to 100% after font size is small No. five words as can be seen from Table 3.
On the basis of above scheme, further, the step 2 is specifically included:
S201, the text in picture to be identified is extracted with graphic form；
S202, region division is carried out to the text picture extracted, text picture is divided into different regions；
S203, selection M region corresponding with comparison domain；
S204, the characteristic value for calculating the M region chosen.
On the basis of above scheme, further, the shape in the step 202 when progress region division using rectangle frame
Formula removes middle section F2In addition, the wire of the rectangle frame in remaining region at least with the top of picture, left end, most lower
One of end, right end are tangent.
On the basis of above scheme, further, in the step 3, the Chinese character of output is editable form.
Embodiment one
Below using " Ah " is illustrated the process of character recognition method as example, wherein as shown in Fig. 2, in picture
" Ah "'s word is the Song typeface, No. four words.
(1) matrix magazine is established:
Region division is carried out in the case where different fonts, different font sizes to each of character library Chinese character.
Font includes seven kinds of the most frequently used fonts such as the Song typeface, imitation Song-Dynasty-style typeface, black matrix, regular script, the refined black, lishu of Microsoft, children's circle.Font size packet
Include just number (42pt), small just (36pt), No.1 (26pt), small by one (24pt), No. two (22pt), small by two (18pt), No. three
(16pt), small by three (15pt), No. four (14pt), small by four (12pt), No. five (10.5pt), small by five (9pt), No. six (7.5pt), it is small
Common 14 kinds of font sizes such as six (6.5pt).
Region includes entirety F1, middle section F2, top half F3, lower half portion F4, left-half F5, right half part F6、
Upper left hand corner section F7, top-right part F8, lower left corner part F9, lower right corner part F10This 10 regions.
Black and white values conversion is carried out using formula (1) to these regions.The characteristic value of each region is calculated using formula (3),
And characteristic value is stored.
The characteristic value collection of different fonts and all Chinese characters of font size is stored, matrix magazine is constructed.
The case where taking into account accuracy rate and computational efficiency, chosen area F1、F4、F6、F5、F3、F2、F9This 7 region conducts
Comparison domain.
(2) picture to be identified is read, such as " Ah " calculates its tangent region, according to tangent region, selects size, font most
Close matrix magazine.
" Ah "'s word feature: to " Ah "'s word progress region division, including whole F on picture is calculated in picture1, middle section
F2, top half F3, lower half portion F4, left-half F5, right half part F6, upper left hand corner section F7, top-right part F8, the lower left corner
Part F9, lower right corner part F10。
Choose F therein1、F4、F6、F5、F3、F2、F9This 7 regions carry out black and white two using formula (1) to this 7 regions
Value figure conversion, the characteristic value in each region is obtained using formula (3).
(3) characteristic value obtained will be calculated to be compared in matrix magazine, exports the highest type matrix Chinese character of similarity.Utilize public affairs
Formula (5) searches in matrix magazine and " the smallest Chinese character of Ah "'s characteristic value difference exports Chinese character in the form of editable.
Above embodiment and embodiment, which are intended to illustrate the present invention, to be realized or make for professional and technical personnel in the field
With modifying to above embodiment will be readily apparent to those skilled in the art, therefore packet of the present invention
Above embodiment is included but is not limited to, it is any to meet the claims or specification description, meet and original disclosed herein
Reason and novelty, the method for inventive features, technique, product, fall within the scope of protection of the present invention.
Claims (8)
1. a kind of picture character recognition methods, which is characterized in that comprising steps of
S1, characteristics extraction is carried out to the Chinese character in multiple fonts, constructs matrix magazine；
S2, feature extraction is carried out to text on picture to be identified, calculates characteristic value；
S3, the characteristic value for calculating acquisition is compared in matrix magazine, exports the highest type matrix Chinese character of similarity.
2. character recognition method as described in claim 1, which is characterized in that the step 1 specifically includes:
S101, region division is carried out to a Chinese character, which is divided into N number of different region, N is the integer greater than 2；
S102, the characteristic value for calculating each region, the characteristic value in N number of region is stored；
S103, the n-quadrant according to corresponding to the multiple fonts of Chinese characters all in character library and a variety of font sizes characteristic value construct word
Mould library.
3. character recognition method as claimed in claim 2, which is characterized in that the step 1 further include:
S104, the recognition accuracy that single region is calculated according to the characteristic value in each region choose M therein and are used as and compare area
Domain, M≤N.
4. character recognition method as claimed in claim 2, which is characterized in that N is 10 in the step 101, and region is specifically wrapped
Include entirety, middle section, top half, lower half portion, the left-half, right half part, upper left hand corner section, upper right corner of Chinese character
Divide, lower left corner part, lower right corner part.
5. character recognition method as claimed in claim 2, which is characterized in that the step 2 specifically includes:
S201, the text in picture to be identified is extracted with graphic form；
S202, region division is carried out to the text picture extracted, text picture is divided into different regions；
S203, selection M region corresponding with comparison domain；
S204, the characteristic value for calculating the M region chosen.
6. character recognition method as claimed in claim 5, which is characterized in that carry out adopting when region division in the step 202
With the form of rectangle frame, in addition to middle section, the wire of the rectangle frame in remaining region at least with the top of picture, most
One of left end, bottom, right end are tangent.
7. character recognition method as claimed in claim 5, which is characterized in that M is 6 to 8.
8. character recognition method as described in claim 1, which is characterized in that in the step 3, the Chinese character of output is editable
Form.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811632008.4A CN109753967A (en) | 2018-12-29 | 2018-12-29 | A kind of picture character recognition methods |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811632008.4A CN109753967A (en) | 2018-12-29 | 2018-12-29 | A kind of picture character recognition methods |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109753967A true CN109753967A (en) | 2019-05-14 |
Family
ID=66404287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811632008.4A Pending CN109753967A (en) | 2018-12-29 | 2018-12-29 | A kind of picture character recognition methods |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109753967A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766893A (en) * | 2019-01-09 | 2019-05-17 | 北京数衍科技有限公司 | Picture character recognition methods suitable for receipt of doing shopping |
| CN112784932A (en) * | 2021-03-01 | 2021-05-11 | 北京百炼智能科技有限公司 | Font identification method and device and storage medium |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS6275785A (en) * | 1985-09-28 | 1987-04-07 | Toshiba Corp | Character recognition device |
| US5119441A (en) * | 1989-03-28 | 1992-06-02 | Ricoh Company, Ltd. | Optical character recognition apparatus and method using masks operation |
| CN1221927A (en) * | 1997-12-19 | 1999-07-07 | 松下电器产业株式会社 | Character recognizor and its method, and recording medium for computer reading out |
| CN101364267A (en) * | 2007-08-09 | 2009-02-11 | 富士通株式会社 | Character recognition apparatus, character recognition method, and computer product |
| CN101561866A (en) * | 2009-05-27 | 2009-10-21 | 上海交通大学 | Character recognition method based on SIFT feature and gray scale difference value histogram feature |
| CN101923741A (en) * | 2010-08-11 | 2010-12-22 | 西安理工大学 | Paper currency number identification method based on currency detector |
| CN102184412A (en) * | 2011-05-09 | 2011-09-14 | 东南大学 | Licence plate number and letter identification method based on minimum-error-rate Bayes classifier |
| CN102184383A (en) * | 2011-04-18 | 2011-09-14 | 哈尔滨工业大学 | Automatic generation method of image sample of printed character |
| CN102663382A (en) * | 2012-04-25 | 2012-09-12 | 重庆邮电大学 | Video image character recognition method based on submesh characteristic adaptive weighting |
| CN102831416A (en) * | 2012-08-15 | 2012-12-19 | 广州广电运通金融电子股份有限公司 | Character identification method and relevant device |
| CN103400127A (en) * | 2013-08-05 | 2013-11-20 | 苏州鼎富软件科技有限公司 | Picture and text identifying method |
| CN104050450A (en) * | 2014-06-16 | 2014-09-17 | 西安通瑞新材料开发有限公司 | Vehicle license plate recognition method based on video |
| CN104182732A (en) * | 2014-08-12 | 2014-12-03 | 南京师范大学 | Handwritten Chinese character stroke confirmation method for carrying out similarity matching on the basis of characteristic matrix |
| CN106778727A (en) * | 2016-12-16 | 2017-05-31 | 高格（天津）信息科技发展有限公司 | Picture character recognition method |
| CN106951890A (en) * | 2017-02-16 | 2017-07-14 | 广东小天才科技有限公司 | A kind of character recognition method and device of dictionary pen |
-
2018
- 2018-12-29 CN CN201811632008.4A patent/CN109753967A/en active Pending
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPS6275785A (en) * | 1985-09-28 | 1987-04-07 | Toshiba Corp | Character recognition device |
| US5119441A (en) * | 1989-03-28 | 1992-06-02 | Ricoh Company, Ltd. | Optical character recognition apparatus and method using masks operation |
| CN1221927A (en) * | 1997-12-19 | 1999-07-07 | 松下电器产业株式会社 | Character recognizor and its method, and recording medium for computer reading out |
| CN101364267A (en) * | 2007-08-09 | 2009-02-11 | 富士通株式会社 | Character recognition apparatus, character recognition method, and computer product |
| CN101561866A (en) * | 2009-05-27 | 2009-10-21 | 上海交通大学 | Character recognition method based on SIFT feature and gray scale difference value histogram feature |
| CN101923741A (en) * | 2010-08-11 | 2010-12-22 | 西安理工大学 | Paper currency number identification method based on currency detector |
| CN102184383A (en) * | 2011-04-18 | 2011-09-14 | 哈尔滨工业大学 | Automatic generation method of image sample of printed character |
| CN102184412A (en) * | 2011-05-09 | 2011-09-14 | 东南大学 | Licence plate number and letter identification method based on minimum-error-rate Bayes classifier |
| CN102663382A (en) * | 2012-04-25 | 2012-09-12 | 重庆邮电大学 | Video image character recognition method based on submesh characteristic adaptive weighting |
| CN102831416A (en) * | 2012-08-15 | 2012-12-19 | 广州广电运通金融电子股份有限公司 | Character identification method and relevant device |
| CN103400127A (en) * | 2013-08-05 | 2013-11-20 | 苏州鼎富软件科技有限公司 | Picture and text identifying method |
| CN104050450A (en) * | 2014-06-16 | 2014-09-17 | 西安通瑞新材料开发有限公司 | Vehicle license plate recognition method based on video |
| CN104182732A (en) * | 2014-08-12 | 2014-12-03 | 南京师范大学 | Handwritten Chinese character stroke confirmation method for carrying out similarity matching on the basis of characteristic matrix |
| CN106778727A (en) * | 2016-12-16 | 2017-05-31 | 高格（天津）信息科技发展有限公司 | Picture character recognition method |
| CN106951890A (en) * | 2017-02-16 | 2017-07-14 | 广东小天才科技有限公司 | A kind of character recognition method and device of dictionary pen |
Non-Patent Citations (3)
| Title |
|---|
| 吴炜 等: "一种基于模糊模板匹配的车牌汉字识别方法", 《微型机与应用》 * |
| 瞿中 等: "改进的车牌相似字符分级分类识别算法研究", 《计算机工程与设计》 * |
| 马飞 等: "基于最短欧氏距离匹配的印刷体汉字识别", 《平顶山学院学报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766893A (en) * | 2019-01-09 | 2019-05-17 | 北京数衍科技有限公司 | Picture character recognition methods suitable for receipt of doing shopping |
| CN112784932A (en) * | 2021-03-01 | 2021-05-11 | 北京百炼智能科技有限公司 | Font identification method and device and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10489682B1 (en) | Optical character recognition employing deep learning with machine generated training data | |
| CN110738207A (en) | character detection method for fusing character area edge information in character image | |
| CN108009222B (en) | Three-dimensional model retrieval method based on better view and deep convolutional neural network | |
| CN110909801B (en) | Data classification method, system, medium and device based on convolutional neural network | |
| CN113239954B (en) | Attention mechanism-based image semantic segmentation feature fusion method | |
| CN107844740A (en) | A kind of offline handwriting, printing Chinese character recognition methods and system | |
| KR20170137170A (en) | Method and apparatus for text image processing | |
| CN106709872B (en) | A kind of rapid image ultra-resolution ratio reconstructing method | |
| Retsinas et al. | Keyword spotting in handwritten documents using projections of oriented gradients | |
| JP2001325564A (en) | Retrieval of handwritten comment in chinese character in cursive style based upon radical model | |
| CN111652273B (en) | Deep learning-based RGB-D image classification method | |
| CN110852326B (en) | Handwriting layout analysis and multi-style ancient book background fusion method | |
| CN109753967A (en) | A kind of picture character recognition methods | |
| CN112784531A (en) | Chinese font and word stock generation method based on deep learning and part splicing | |
| CN111368695A (en) | Table structure extraction method | |
| CN106156798A (en) | Scene image classification method based on annular space pyramid and Multiple Kernel Learning | |
| CN110837808A (en) | Hyperspectral image classification method based on improved capsule network model | |
| Garg et al. | An algorithm for text line segmentation in handwritten skewed and overlapped Devanagari script | |
| Gillies et al. | Arabic text recognition system | |
| CN107729918B (en) | Classification method for cellular automaton emerging phenomenon based on cost-sensitive support vector machine | |
| CN109858609A (en) | A kind of method and system in blockette pond | |
| CN102842046A (en) | Calligraphic style identification method based on overall feature extracting and training | |
| Liu et al. | Language identification of character images using machine learning techniques | |
| CN113723410A (en) | Digital tube digital identification method and device | |
| Desai et al. | Adversarial Network for Photographic Image Synthesis from Fine-grained Captions |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190514 |