CN106203417A - A kind of adhesion character alienable RMB crown word number identification method - Google Patents
A kind of adhesion character alienable RMB crown word number identification method Download PDFInfo
- Publication number
- CN106203417A CN106203417A CN201610544250.0A CN201610544250A CN106203417A CN 106203417 A CN106203417 A CN 106203417A CN 201610544250 A CN201610544250 A CN 201610544250A CN 106203417 A CN106203417 A CN 106203417A
- Authority
- CN
- China
- Prior art keywords
- character
- picture
- segmentation
- character picture
- characteristic pattern
- 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
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
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
- G06V10/225—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition based on a marking or identifier characterising the area
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Character Discrimination (AREA)
Abstract
The invention discloses a kind of adhesion character alienable RMB crown word number identification method, split i.e. by horizontal thin: judged more than single character picture Breadth Maximum D by widthmaxFor unusual character figure i.e. adhesion character picture, and by the upright projection curve of unusual character image, in the subinterval searching minimum point set as Character segmentation point, then carry out vertical cutting, if the character picture that cutting processes is more than minimum character picture width Dmin, then retain this character picture, so conglutination segmentation come, improve the Character segmentation accuracy of preprocessing process, reduce serial number misclassification rate.
Description
Technical field
The invention belongs to pattern recognition and optical character recognition (OCR) technical field, particularly, more specifically,
Relate to a kind of adhesion character alienable RMB crown word number identification method, can be applicable to the people of counting currency examination cleaning-sorting machine system
Coin crown word number identification.
Background technology
" RMB-banknote discriminating device general technical specifications " national standard amendment drafting group disclosed new edition in 2009 and tests
Paper money machine standard GB 16999-2010, it is desirable to cash inspecting machine increases paper money number identification function, therefore embeds bank note in paper money counter
Crown word number identification function has become as up-to-date research topic.
It is stained etc. by machine performance or banknote itself and affects, when crown word number on banknote is shot or scans by cash inspecting machine,
Some interference can be produced so that character sticks together, and so causes cash inspecting machine cannot correctly identify the crown word number of RMB so that
It is unable to reach the serial number misclassification rate of the regulation requirement less than 0.03% in " RMB-banknote discriminating device general technical specifications ", its
In, serial number misclassification rate refers to that the bank note quantity occurring serial number by mistake to know and reality identify the ratio of bank note quantity.
Summary of the invention
It is an object of the invention to overcome the deficiencies in the prior art, propose a kind of adhesion character alienable RMB prefix
Number recognition methods, effectively to split adhesion character, reduces serial number misclassification rate.
For achieving the above object, adhesion character of the present invention alienable RMB crown word number identification method, its feature
It is, comprises the following steps:
(1), Character segmentation
1.1), the horizontal coarse segmentation of character
The crown word number gray level image that input is gathered by imageing sensor, and carry out binary conversion treatment, and then hang down
Deliver directly shadow to resolve, be at 0 value at horizontal level projection value, carry out vertical cutting, obtain the coarse segmentation image of character；
1.2), character horizontal thin segmentation
1.2.1), the width of the character picture that detection coarse segmentation obtains: if width is more than the maximum width of single character picture
Degree Dmax, this character picture is considered as unusual character image, goes to step 1.2.2)；Otherwise this character picture is considered as single character
Image, continues detection character late image, until having detected last character image termination character segmentation；
1.2.2), statistics unusual character image upright projection curve, comprising the character picture water of horizontal centre coordinate
Flat interval subinterval [xl+s,xr-s] the interior minimum point finding upright projection curve, and using this point as conglutination segmentation
Point, wherein, xl,xrObtain the left and right coordinate of character picture horizontal direction for coarse segmentation, s is subinterval setup parameter, according to specifically
Situation sets, and its value is the biggest, and subinterval is the least；
1.2.3), with step 1.2.2) unusual character image does vertical cutting, and sentences by the conglutination segmentation point that obtains
The character duration of the left and right character picture after disconnected cutting, if this character picture width is less than minimum character picture width Dmin, then
Give up this character picture, otherwise retain this character picture, be then back to step 1.2), continue detection character late image；
1.3), character vertical segmentation
By step 1.2) character picture that obtains carries out floor projection parsing, is that 0 value character picture is removed i.e. at projection value
Upper and lower for character picture blank parts is excised；
(2), character recognition
Single character picture step (1) Character segmentation obtained is after character picture size normalization, according to level
Cutting order, the convolutional neural networks being sequentially sent to train is identified, and obtains corresponding character；
(3), all of character identified, it is combined according to cutting order, obtains crown word number (code).
The object of the present invention is achieved like this.
Adhesion character of the present invention alienable RMB crown word number identification method, splits i.e. by horizontal thin: pass through width
Judge more than single character picture Breadth Maximum DmaxFor unusual character figure i.e. adhesion character picture, and pass through unusual character
The upright projection curve of image, in the subinterval searching minimum point set as Character segmentation point, then carries out vertical cutting,
If the character picture that cutting processes is more than minimum character picture width Dmin, then this character picture is retained, so by adhesion character
Separated, improve the Character segmentation accuracy of preprocessing process, reduce serial number misclassification rate.
Accompanying drawing explanation
Fig. 1 is adhesion character of the present invention alienable RMB crown word number identification method one detailed description of the invention flow process
Figure；
Fig. 2 is the crown word number gray level image one instantiation figure that there is adhesion character；
Fig. 3 is the upright projection curve chart of crown word number gray level image shown in Fig. 2；
Fig. 4 is the character picture that crown word number gray level image shown in Fig. 2 carries out that horizontal thin segmentation obtains；
Fig. 5 is that convolutional neural networks one shown in Fig. 1 is embodied as structure chart.
Detailed description of the invention
Below in conjunction with the accompanying drawings the detailed description of the invention of the present invention is described, in order to those skilled in the art is preferably
Understand the present invention.Requiring particular attention is that, in the following description, when known function and design detailed description perhaps
When can desalinate the main contents of the present invention, these are described in and will be left in the basket here.
Fig. 1 is adhesion character of the present invention alienable RMB crown word number identification method one detailed description of the invention flow process
Figure.
In the present embodiment, as described in Figure 1, adhesion character of the present invention alienable RMB crown word number identification method includes
Following steps:
One, Character segmentation
8 bit gradation crown word number image (input picture) that S101, input are gathered by imageing sensor, carry out binaryzation
Process, obtain binary image；In the present embodiment, after binary conversion treatment, having carried out inversion operation, i.e. 0 becomes 255, and 255 become
Being 0, such character has just become white, and background has become black；
S102, binary image is carried out upright projection parsing, horizontal level projection value be at 0 value i.e. at this without vertically
Direction be the pixel number of 255 be 0, carry out vertical cutting, obtain the coarse segmentation image of character；
In the present embodiment, step S101, S102 are the horizontal coarse segmentation of character.
The i.e. character horizontal thin of S103, adhesion character machining and segmentation is split；
In the present embodiment, as in figure 2 it is shown, crown word number image acquisition process may cause word due to noise with being stained
Adhesion is there is between symbol, if using traditional upright projection split-run i.e. step S101, the method for S102 directly to split,
U and F can be considered a character and split, it is clear that segmentation result is wrong, and this can cause None-identified or identify mistake.
Meanwhile, last character 0 causes character duration to be more than D due to noisemax, it is also regarded as unusual character.
In the present invention, following methods is used to carry out character horizontal thin segmentation, it may be assumed that
The width of the character picture that S1031, detection coarse segmentation obtain: if width is more than the Breadth Maximum of single character picture
Dmax, this character picture is considered as unusual character image, goes to step S1032；Otherwise this character picture is considered as single character figure
Picture, continues detection character late image, until having detected last character image termination character segmentation；
S1032, the upright projection curve of statistics unusual character image, comprising the character picture level of horizontal centre coordinate
Interval subinterval [xl+s,xr-s] in find upright projection curve minimum point i.e. this vertical direction be the pixel of 255
Number is minimum, and using this point as conglutination segmentation point, wherein, xl,xrA left side for character picture horizontal direction is obtained for coarse segmentation
Right coordinate, s is subinterval setup parameter, sets as the case may be, and its value is the biggest, and subinterval is the least；In the present embodiment, word
Symbol image size uses pixel to represent, parameter s=3, i.e. 3 pixels
Unusual character image is done vertical cutting, and judges by S1033, the conglutination segmentation point obtained with step S1032
The character duration of the left and right character picture after cutting, if this character picture width is less than minimum character picture width Dmin, then give up
Abandon this character picture, otherwise retain this character picture, be then back to step 1.2), continue detection character late image；
After using above-mentioned character horizontal thin dividing method, obtain two adhesion character pictures of Fig. 3 heavy black line mark
Segmentation coordinate, obtains two characters after first unusual character image segmentation, and character duration is both greater than character minimum widith Dmin,
The most all remaining, obtain two characters after second unusual character segmentation, the peak width on the adhesion cut-point left side is more than
Dmin, reserved character, the peak width on the right is less than Dmin, therefore give up.Finally obtain ten characters after correct segmentation, such as figure
Shown in 4.
S104, character vertical segmentation
Character picture step S103 obtained carries out floor projection parsing, is that 0 value character picture is removed i.e. at projection value
Upper and lower for character picture blank parts is excised.
Two, character recognition
S201, single character picture step one Character segmentation obtained are through character picture size normalization.In this reality
Execute in example, normalization a size of 14x14.
Character picture after S202, normalization size, according to horizontal cutting order, is sequentially sent to the convolutional Neural trained
Network is identified, and obtains corresponding character.
Compared with tradition template matching mutually and the artificial recognition methods extracting character feature, convolutional neural networks has identification
Rate is high, automatically extracts the feature of the various profound feature of character picture.In the present embodiment, in order to reduce, resource is disappeared
Consumption, improves recognition speed, takes into account the serial number accuracy (serial number misclassification rate) of identification, to traditional convolutional Neural simultaneously
Network structure is simplified.
In the present embodiment, as it is shown in figure 5, convolutional neural networks structure is based on LeNet5 framework, simplify operation
For:
1, not comprising input layer, the convolutional neural networks of LeNet5 framework has 7 layers, and in the present embodiment, convolution god
Being of five storeys altogether through network, i.e. decrease C5 layer and F6 layer, C5 layer is the convolutional layer of 120 characteristic pattern compositions, and F6 is full articulamentum,
Removing this two-layer and can reduce substantial amounts of training parameter, saved calculating resource and memory resource, S4 layer characteristic pattern is directly adopted
It is connected to output layer, it is not necessary to middle hidden layer by full connected mode；
2, reducing by the characteristic pattern quantity of every layer, the convolutional neural networks C3 layer characteristic pattern of LeNet5 framework contains 16 features
Figure, and C3 layer characteristic pattern of the present invention comprises only 6 characteristic patterns；
3, S2 layer characteristic pattern uses the full mode connected rather than the convolutional neural networks of LeNet5 framework to C3 characteristic pattern
In part connected mode, add a part of Connecting quantity in this way although adopting, but be reduction of FPGA and realize difficulty.
4, in the present embodiment, down-sampling layer uses average pond, it is not necessary to training parameter, only need to train one for every layer
Offset parameter.
Specifically, in the present embodiment, described convolutional neural networks be identified into:
S2021, single for the 14x14 of size normalization character picture is carried out convolution, obtain comprising the C1 layer of 6 characteristic patterns
Characteristic pattern, the size of each characteristic pattern is 12x12；
S2022, C1 layer characteristic pattern is carried out down-sampling, obtain the S2 layer characteristic pattern comprising 6 characteristic patterns, each characteristic pattern
Size be 6x6；
S2023, S2 layer characteristic pattern is carried out convolution, obtain the C3 layer characteristic pattern comprising 6 characteristic patterns, each characteristic pattern
Size is 4x4；
S2024, C3 layer characteristic pattern is carried out down-sampling, obtain the S4 layer characteristic pattern comprising 6 characteristic patterns, each characteristic pattern
Size be 2x2；
S2025, S4 layer characteristic pattern directly uses full connected mode to be connected to output layer, exports corresponding character.
Three, all of character identified, it is combined according to cutting order, obtains crown word number (code).Crown word number.
In the present embodiment, first build and training convolutional neural networks on PC Host Computer Software Platform, then will training
The character picture that good convolutional neural networks downloads in the fpga chip of cash inspecting machine obtain collection segmentation is identified.
Although detailed description of the invention illustrative to the present invention is described above, in order to the technology of the art
Personnel understand the present invention, the common skill it should be apparent that the invention is not restricted to the scope of detailed description of the invention, to the art
From the point of view of art personnel, as long as various change limits and in the spirit and scope of the present invention that determine in appended claim, these
Change is apparent from, and all utilize the innovation and creation of present inventive concept all at the row of protection.
Claims (2)
1. an adhesion character alienable RMB crown word number identification method, it is characterised in that comprise the following steps:
(1), Character segmentation
1.1), the horizontal coarse segmentation of character
The crown word number gray level image that input is gathered by imageing sensor, and carry out binary conversion treatment, and the most vertically throw
Shadow resolves, and is at 0 value at horizontal level projection value, carries out vertical cutting, obtains the coarse segmentation image of character；
1.2), character horizontal thin segmentation
1.2.1), the width of the character picture that detection coarse segmentation obtains: if width is more than the Breadth Maximum of single character picture
Dmax, this character picture is considered as unusual character image, goes to step 1.2.2)；Otherwise this character picture is considered as single character figure
Picture, continues detection character late image, until having detected last character image termination character segmentation；
1.2.2), statistics unusual character image upright projection curve, in the character picture horizontal zone comprising horizontal centre coordinate
Between subinterval [xl+s,xr-s] the interior minimum point finding upright projection curve, and using this point as conglutination segmentation point,
Wherein, xl,xrObtain the left and right coordinate of character picture horizontal direction for coarse segmentation, s is subinterval setup parameter, according to concrete feelings
Condition sets, and its value is the biggest, and subinterval is the least；
1.2.3), with step 1.2.2) unusual character image does vertical cutting, and judges to cut by the conglutination segmentation point that obtains
The character duration of the left and right character picture after Fen, if this character picture width is less than minimum character picture width Dmin, then give up
This character picture, on the contrary retain this character picture, it is then back to step 1.2), continue detection character late image；
1.3), character vertical segmentation
By step 1.2) character picture that obtains carries out floor projection parsing, is that 0 value character picture removes will word at projection value
Symbol image upper and lower blank parts excision；
(2), character recognition
Single character picture step (1) Character segmentation obtained is after character picture size normalization, according to horizontal cutting
Sequentially, the convolutional neural networks being sequentially sent to train is identified, and obtains corresponding character；
(3), all of character identified, it is combined according to cutting order, obtains crown word number (code).
Recognition methods the most according to claim 1, it is characterised in that in step (2), described convolutional neural networks enters
Row is identified as:
2.1), single for the 14x14 of size normalization character picture is carried out convolution, obtain the C1 layer feature comprising 6 characteristic patterns
Figure, the size of each characteristic pattern is 12x12；
2.2), C1 layer characteristic pattern is carried out down-sampling, obtain the S2 layer characteristic pattern comprising 6 characteristic patterns, each characteristic pattern big
Little for 6x6；
2.3), S2 layer characteristic pattern is carried out convolution, obtain the C3 layer characteristic pattern comprising 6 characteristic patterns, the size of each characteristic pattern
For 4x4；
2.4), C3 layer characteristic pattern is carried out down-sampling, obtain the S4 layer characteristic pattern comprising 6 characteristic patterns, each characteristic pattern big
Little for 2x2；
2.5), S4 layer characteristic pattern directly use full connected mode to be connected to output layer, export corresponding character.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610544250.0A CN106203417A (en) | 2016-07-12 | 2016-07-12 | A kind of adhesion character alienable RMB crown word number identification method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610544250.0A CN106203417A (en) | 2016-07-12 | 2016-07-12 | A kind of adhesion character alienable RMB crown word number identification method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106203417A true CN106203417A (en) | 2016-12-07 |
Family
ID=57477244
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610544250.0A Pending CN106203417A (en) | 2016-07-12 | 2016-07-12 | A kind of adhesion character alienable RMB crown word number identification method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106203417A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874909A (en) * | 2017-01-18 | 2017-06-20 | 深圳怡化电脑股份有限公司 | A kind of recognition methods of image character and its device |
| CN108491845A (en) * | 2018-03-02 | 2018-09-04 | 深圳怡化电脑股份有限公司 | Determination, character segmentation method, device and the equipment of Character segmentation position |
| CN108596166A (en) * | 2018-04-13 | 2018-09-28 | 华南师范大学 | A kind of container number identification method based on convolutional neural networks classification |
| CN108664978A (en) * | 2017-03-30 | 2018-10-16 | 北京君正集成电路股份有限公司 | The character segmentation method and device of fuzzy license plate |
| CN108830380A (en) * | 2018-04-11 | 2018-11-16 | 开放智能机器（上海）有限公司 | A kind of training pattern generation method and system based on cloud service |
| CN110135563A (en) * | 2019-05-13 | 2019-08-16 | 北京航空航天大学 | A kind of convolutional neural networks binarization method and computing circuit |
| CN110472505A (en) * | 2019-07-11 | 2019-11-19 | 深圳怡化电脑股份有限公司 | Recognition methods, identification device and the terminal of bill serial number |
| CN112257715A (en) * | 2020-11-18 | 2021-01-22 | 西南交通大学 | Method and system for identifying adhesive characters |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2008250754A (en) * | 2007-03-30 | 2008-10-16 | Nidec Sankyo Corp | Character string recognition method and device |
| CN104408814A (en) * | 2014-12-13 | 2015-03-11 | 天津远目科技有限公司 | Method for identifying RMB code |
| CN105046252A (en) * | 2014-11-21 | 2015-11-11 | 华中科技大学 | Method for recognizing Renminbi (Chinese currency yuan) crown codes |
-
2016
- 2016-07-12 CN CN201610544250.0A patent/CN106203417A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2008250754A (en) * | 2007-03-30 | 2008-10-16 | Nidec Sankyo Corp | Character string recognition method and device |
| CN105046252A (en) * | 2014-11-21 | 2015-11-11 | 华中科技大学 | Method for recognizing Renminbi (Chinese currency yuan) crown codes |
| CN104408814A (en) * | 2014-12-13 | 2015-03-11 | 天津远目科技有限公司 | Method for identifying RMB code |
Non-Patent Citations (2)
| Title |
|---|
| 冯博远: "复杂背景下人民币冠字号码识别技术研究", 《中国博士学位论文全文数据库 信息科技辑》 * |
| 张晴晴: "人民币冠字号码的图像识别技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874909A (en) * | 2017-01-18 | 2017-06-20 | 深圳怡化电脑股份有限公司 | A kind of recognition methods of image character and its device |
| CN106874909B (en) * | 2017-01-18 | 2019-10-11 | 深圳怡化电脑股份有限公司 | A kind of recognition methods of image character and its device |
| CN108664978A (en) * | 2017-03-30 | 2018-10-16 | 北京君正集成电路股份有限公司 | The character segmentation method and device of fuzzy license plate |
| CN108664978B (en) * | 2017-03-30 | 2021-10-26 | 北京君正集成电路股份有限公司 | Character segmentation method and device for fuzzy license plate |
| CN108491845A (en) * | 2018-03-02 | 2018-09-04 | 深圳怡化电脑股份有限公司 | Determination, character segmentation method, device and the equipment of Character segmentation position |
| CN108491845B (en) * | 2018-03-02 | 2022-05-31 | 深圳怡化电脑股份有限公司 | Character segmentation position determination method, character segmentation method, device and equipment |
| CN108830380A (en) * | 2018-04-11 | 2018-11-16 | 开放智能机器（上海）有限公司 | A kind of training pattern generation method and system based on cloud service |
| CN108596166A (en) * | 2018-04-13 | 2018-09-28 | 华南师范大学 | A kind of container number identification method based on convolutional neural networks classification |
| CN108596166B (en) * | 2018-04-13 | 2021-10-26 | 华南师范大学 | Container number identification method based on convolutional neural network classification |
| CN110135563A (en) * | 2019-05-13 | 2019-08-16 | 北京航空航天大学 | A kind of convolutional neural networks binarization method and computing circuit |
| CN110135563B (en) * | 2019-05-13 | 2022-07-26 | 北京航空航天大学 | Convolution neural network binarization method and operation circuit |
| CN110472505A (en) * | 2019-07-11 | 2019-11-19 | 深圳怡化电脑股份有限公司 | Recognition methods, identification device and the terminal of bill serial number |
| CN110472505B (en) * | 2019-07-11 | 2022-03-08 | 深圳怡化电脑股份有限公司 | Bill serial number identification method, bill serial number identification device and terminal |
| CN112257715A (en) * | 2020-11-18 | 2021-01-22 | 西南交通大学 | Method and system for identifying adhesive characters |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106203417A (en) | A kind of adhesion character alienable RMB crown word number identification method | |
| CN110033000B (en) | Text detection and identification method for bill image | |
| CN110009057B (en) | Graphic verification code identification method based on deep learning | |
| CN104298982B (en) | A kind of character recognition method and device | |
| CN102542660B (en) | Bill anti-counterfeiting identification method based on bill watermark distribution characteristics | |
| CN105303676B (en) | A kind of version recognition methods of bank note and system | |
| CN104063706B (en) | Video fingerprint extraction method based on SURF algorithm | |
| CN110008680A (en) | System and method is generated based on the identifying code to resisting sample | |
| CN104408449B (en) | Intelligent mobile terminal scene literal processing method | |
| CN102194012B (en) | Microblog topic detecting method and system | |
| CN102737242A (en) | Automatic bill recognition method and system applied to mobile terminal | |
| CN106056751A (en) | Prefix number identification method and system | |
| CN107808358A (en) | Image watermark automatic testing method | |
| CN106295502A (en) | A kind of method for detecting human face and device | |
| CN101329727A (en) | Fingerprint identification method combining point with line | |
| CN101980211A (en) | Machine learning model and establishing method thereof | |
| CN102194275A (en) | Automatic ticket checking method for train tickets | |
| CN114842524B (en) | Face false distinguishing method based on irregular significant pixel cluster | |
| CN106778717A (en) | A kind of test and appraisal table recognition methods based on image recognition and k nearest neighbor | |
| CN106683259A (en) | Method and device for identifying paper money | |
| CN107885723A (en) | Conversational character differentiating method and system | |
| CN110414505A (en) | Processing method, processing system and the computer readable storage medium of image | |
| CN108038495A (en) | A kind of incompleteness Chinese characters recognition method | |
| CN109815268A (en) | A kind of transaction sanction list matching system | |
| CN108509950A (en) | Railway contact line pillar number plate based on probability characteristics Weighted Fusion detects method of identification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into 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: 20161207 |