CN107292311A - A kind of recognition methods of the Characters Stuck identifying code based on neutral net - Google Patents
A kind of recognition methods of the Characters Stuck identifying code based on neutral net Download PDFInfo
- Publication number
- CN107292311A CN107292311A CN201710682566.0A CN201710682566A CN107292311A CN 107292311 A CN107292311 A CN 107292311A CN 201710682566 A CN201710682566 A CN 201710682566A CN 107292311 A CN107292311 A CN 107292311A
- Authority
- CN
- China
- Prior art keywords
- image
- character
- identifying code
- picture
- noise
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 230000007935 neutral effect Effects 0.000 title claims abstract description 16
- 230000011218 segmentation Effects 0.000 claims abstract description 20
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000013528 artificial neural network Methods 0.000 claims abstract description 9
- 238000012545 processing Methods 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 8
- 230000007797 corrosion Effects 0.000 claims description 7
- 238000005260 corrosion Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 6
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 13
- 238000011946 reduction process Methods 0.000 abstract description 4
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 238000013461 design Methods 0.000 description 7
- 238000000605 extraction Methods 0.000 description 4
- 238000012015 optical character recognition Methods 0.000 description 4
- 230000008859 change Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 239000002245 particle Substances 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000008030 elimination Effects 0.000 description 1
- 238000003379 elimination reaction Methods 0.000 description 1
- 230000003628 erosive effect Effects 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 210000002569 neuron Anatomy 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- 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
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
- G06V10/507—Summing image-intensity values; Histogram projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20036—Morphological image processing
-
- 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)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Character Input (AREA)
- Character Discrimination (AREA)
Abstract
A kind of recognition methods of the Characters Stuck identifying code based on neutral net, comprises the following steps:Intercept a number of identifying code picture automatically from website standby, basic pretreatment is carried out to the image of the identifying code picture of deposit, original identifying code image is set to be easier to Character segmentation, by the Character segmentation in pretreated image into single character, the character picture segmented is normalized, training is identified to normalized character picture by artificial neural network, recognition accuracy is calculated.Beneficial effect of the present invention:Pass through different degrees of noise reduction process, the complicated ambient noise of identifying code can be removed well, substantially zero noise Character segmentation of progress can be reached, also there is preferable effect simultaneously for the segmentation of the character of adhesion, have higher discrimination for the complex background identifying code with adhesion character by the training of neutral net.
Description
Technical field
The present invention relates to intelligent information processing technology field, specifically a kind of Characters Stuck based on neutral net is tested
Demonstrate,prove the recognition methods of code.
Background technology
The invention of identifying code strengthens the protection to information, prevents some illegal with ensureing the safety of network using being
Molecule passes through the various means such as high performance hardware device, the rogue program of specific function, web crawlers and website design leak
Crack account password, steal the automatic registration of user profile, malice.Identifying code is included compared with other safety verification modes
Data volume is smaller, and effectively raises web portal security performance and anti-attack ability.Identifying code is to discriminate between computer and the mankind
Full-automatic turing test program.
But due to continuing to develop for network, identifying code is also reformed constantly, identification identifying code will be seen that identifying code is set
The rule and principle of meter, contribute to find identifying code design defect, so as to design it is safer, more ripe, more conform to use
The identifying code of family usage experience.The identification of identifying code simultaneously is also a large amount of big data analyses that data are captured from network of needs
Worker provides facility, promotes the development of correlative study.
Checking code type on network has many kinds, mainly have it is based on Text Mode, based on image model, based on language
Sound pattern and video mode is at least partly based on, it is most widely used with the identifying code based on Text Mode.Text based is tested
Card code initially only includes simple numeral and alphabetical, optical character identification (Optical Character Recognition,
OCR) text in image can directly be extracted identification and have good recognition effect by technology.The safety of network, which is promoted, to be verified
The identification difficulty of code is continuously increased, and starts to change character color, increases image background, increase image noise interference etc., for
This kind of identifying code first can carry out reusing OCR extractions after simply pre-processing to image, also there is more good effect.
With continuing to develop for identifying code, begin through the text in identifying code image is distorted, the method such as adhesion
To increase the difficulty of its identification.The difficulty of such identifying code is to be split the character of adhesion, and researcher carries
Go out many character extraction algorithms, such as skeletal extraction algorithm, color filling method, Drop fall algorithm etc., passing through these algorithms will
Character in identifying code split or profile extraction, then by OCR, naive Bayesian, SVMs etc. is trained.
But segmentation effect is often unsatisfactory, so as to cause final recognition correct rate not high.
The content of the invention
The technical problems to be solved by the invention are to provide a kind of identification of the Characters Stuck identifying code based on neutral net
Method, solves the problem of current identifying code Characters Stuck, background complexity are difficult to.
The present invention is for the technical scheme that is used of solution above-mentioned technical problem:A kind of Characters Stuck based on neutral net
The recognition methods of identifying code, comprises the following steps:
Step 1: intercepting a number of identifying code picture automatically from website, deposit file is standby;
Step 2: carrying out basic pretreatment to the image of the identifying code picture of deposit, original identifying code image is set to be easier to word
Symbol is cut, and specific method is:
(1)By the coloured image gray processing in the identifying code picture of collection;
(2)Gray level image goes dry processing:The noise in corrosion treatment, downscaled images, usage threshold algorithm, selection are carried out to image
Appropriate threshold value, binary conversion treatment is carried out by image, and the image after binary conversion treatment is negated, finally using median filtering method
Remove most of noise in image;
(3)The size of locating verification code, to step(2)Image after denoising removes frame;
(4)To going the image after frame further to remove companies all in noise, i.e. selection image by connected region noise reduction hair
Logical region, selectes a suitable threshold value, removes the less noise of connected region, retains the larger character of connected region;
Step 3: by the Character segmentation in the pretreated image of step 2 into single character, specific method is:
(1)The gray value of image after step 2 is handled carries out the projection of vertical direction, obtains the projection of image vertical direction
Histogram;
(2)By Character segmentation of the position of trough in the projection histogram of obtained vertical direction, character picture is divided into list by point
Individual character picture;
(3)Single character picture gray value is subjected to the projection in horizontal direction, the throwing of single character picture horizontal direction is obtained
Shadow histogram, determines the up-and-down boundary of character by the projection histogram of horizontal direction and is split;
Step 4: the character picture segmented is normalized;
Step 5: training is identified to normalized character picture by artificial neural network;
Step 6: calculating recognition accuracy.
In step 5 of the present invention, it is to the specific method that training is identified in normalized character picture:Selected BP
Neutral net is as training tool, from three-layer neural network structure, i.e. input layer, single hidden layer and output layer, using with
Training is identified in the training type of machine gradient method and adaptive regularized learning algorithm rate.
The beneficial effects of the invention are as follows:The present invention is by selecting the order of suitable denoising mode and denoising method to checking
Code image is pre-processed, by different degrees of noise reduction process, well can be removed the complicated ambient noise of identifying code, right
Also there is preferable effect in the segmentation of the character of adhesion.During corrosion treatment is applied into noise reduction, background not remove only
In molecule noise, while also been removed the noise for being attached to character edge, become apparent from character outline;On basis
The method that connected region noise reduction is utilized after the completion of noise reduction process, the noise that medium filtering is difficult to remove is removed, can be reached substantially
To zero noise Character segmentation of progress;Verify that this method is for the complex background with adhesion character using the training of neutral net
Identifying code has higher discrimination.
Brief description of the drawings
Fig. 1 is the flow chart of method for recognizing verification code involved in the present invention;
Fig. 2 is identifying code original image of the embodiment of the present invention;
Fig. 3 is the later image of identifying code image gray processing of the embodiment of the present invention;
Fig. 4 is the later design sketch of identifying code Image erosion of the embodiment of the present invention;
Fig. 5 is identifying code of embodiment of the present invention binaryzation and the image negated;
Fig. 6 is the design sketch after identifying code medium filtering of the embodiment of the present invention;
Fig. 7 is the design sketch after identifying code connected region noise reduction of the embodiment of the present invention;
Fig. 8 is identifying code image projection histogram divion explanation of the embodiment of the present invention;
Fig. 9 is identifying code Character segmentation design sketch of the embodiment of the present invention;
Figure 10 is autoadapted learning rate of the embodiment of the present invention and error change curve map.
Embodiment
With reference to embodiment of the Figure of description to the present invention(Embodiment)It is described, makes the skill of this area
Art personnel better understood when the present invention.
A kind of recognition methods of the Characters Stuck identifying code based on neutral net, comprises the following steps:
Step 1, by the way that C++ codings are automatic refresh Website page and intercept identifying code picture automatically, a certain amount of picture of acquisition is simultaneously
It is stored in standby in specific file;
Step 2, the image to the identifying code picture of interception are pre-processed, and original identifying code image is become suitable for segmentation
Image, specific method is:
2.1st, it is first according to formula for the image of identifying code to be identified
Coloured image gray processing processing is carried out, wherein,R、G、BThe pixel value of pixel red, green, blue in coloured image is represented respectively,GrayFor the gray value of required point, effect is as shown in Figure 3;
2.2nd, the noise in gray level image can influence the result that image is split, therefore carry out denoising to gray level image, due to
Identifying code image background is complicated, it is necessary to from suitable denoising method and suitable noise reduction order so that image to be split is
The effect of noise reduction is reached, while remaining the clear profile of character.Specific method is:
2.2.1, corrosion is a kind of elimination boundary point, and the process for making border internally shrink, expansion can regard pair of corrosion as
Even computing.Use expansion formulaReverse operating to gray level image carry out corrosion treatment, wherein A is to treat
The image of processing, B is used structural element,Represent that B hits A, i.e., in the presence of a point being both the element and A in B
Element,It is to be obtained after structural element B is translated z, ifA is hit, this z point is write down, all z points for meeting condition
Set is referred to as the result that A is expanded by B.The region relative decrease of character inside image can be made by corrosion treatment, and eroded
Some noises of image character edge adhesion, become apparent from character outline, at the same can also by particle in background image compared with
Small noise is easily removed, and effect is as shown in Figure 4;
2.2.2, usage threshold algorithm, selects appropriate threshold value, and the image that step 2.2.1 is treated carries out binary conversion treatment,
And the image after binaryzation is negated, the image of such black matrix wrongly written or mispronounced character can more clearly observe the noise in background, understand
The convenient denoising work for carrying out next step of the feature of remaining noise, effect is as shown in Figure 5;
2.2.3, because noise particle is smaller and position isolated, therefore it is filtered, is removed using 3 × 3 medium filtering template
Most of noise in image, if starting directly to use medium filtering, can cause background to be obscured with character, character outline is not clear enough
It is clear, and increase the difficulty of later stage character recognition, effect is as shown in Figure 6;
2.3rd, locating verification code image size, the frame for hindering processing is removed to the image after denoising;
2.4th, the image treated to step 2.3 further removes noise by connected region Method of Noise, i.e., will own in image
Connected region be marked, because most of noise has been removed in image, so noise larger and isolated in image is just
It can clearly mark out and, due to a required character necessarily big connected region, so the size of contrast connected region
Appropriate threshold is set, less connected region is deleted, retains the big connected region for including character, to reach the purpose of denoising.
If starting just to use connected region noise reduction, the noise sticked in around character can be made together to calculate the connected region for belonging to character
In, cause character outline smudgy, segmentation difficulty is increased, and effect is as shown in Figure 7;
Step 3, the image pretreated to step 2 are divided into single character picture, and specific method is:
The 3.1st, the gray value of pretreated image is carried out to the projection of vertical direction, the projection Nogata of image vertical direction is obtained
Figure, because background color is black, and the gray value of black is 0, and for the character of non-adhesion, gray value is two for 0 position
Space between character, judges the relative width of single character;
3.2nd, for adhesion character, Characters Stuck part is smaller, and gray value is also relatively small, and resulting projection histogram will
Corresponding trough is formed, the position of trough is selected with reference to the relative width of character, Character segmentation point, projection histogram segmentation is determined
Illustrate as shown in Figure 8;
3.3rd, the character picture of pre-segmentation is divided into single character picture according to cut point, Character segmentation effect is as shown in Figure 9;
3.4th, single character picture gray value is subjected to the projection in horizontal direction, by grey in the projection histogram of horizontal direction
Angle value determines the up-and-down boundary of character and split for 0 peak width;
Step 4, the character picture that step 3 is segmented is normalized, the purpose is to the dimension of the input due to neutral net
Degree needs unified, and normalized size is;
Step 5, by BP neural network training is identified in the character picture normalized, specific method is:
5.1st, from three-layer neural network structure, i.e. input layer, single hidden layer and output layer, the pixel letter of single character is obtained
Breath, and as the input layer of BP neural network;It is 126 to set node in hidden layer, if setting excessive node easily to lead
Cause over-fitting and slow down training speed;Set output layer and include 37 as one(10 numerals, 26 letters and one are not
Know)The linear neuron of target output, wherein activation primitive used in hidden layer is hyperbolic tangent function in neutral net
(tanh) sigmoid functions, are used in output layer;
5.2nd, using the training type of Stochastic gradient method, there is faster training speed and just when to set maximum exercise wheel number be 256
True rate;
5.3rd, using adaptive regularized learning algorithm rate, when training error declines, learning rate is raised, and makes instruction to increase learning process
Practice result more accurate;But then directly terminate training when reaching target error, therefore target error numerical value should not be set
Greatly, learning rate and error change curve are as shown in Figure 9.Other specification sets as shown in table 1 in neutral net.
The parameter setting of table 1
Parameter name | Initial learning rate | Momentum | Target error |
Parameter value | 0.0001 | Learning rate * 0.9 | 0.002 |
Step 6, calculating recognition accuracy, using the method for nine folding cross validations, wherein recognition correct rate is that all characters are correct
RatePAverage, and calculate nine times accuracy average, to reach an accurate result of fair relatively.Wherein accuracyP's
Calculation formula is:, whereinTPThe quantity being correctly validated for a certain character,FNIt is erroneously identified for the character
Into other character quantities.
By different degrees of noise reduction process, the complicated ambient noise of identifying code can be removed well, can reached substantially
To zero noise Character segmentation of progress.Also there is preferable effect simultaneously for the segmentation of the character of adhesion, pass through the instruction of neutral net
White silk has higher discrimination for the complex background identifying code with adhesion character.
Claims (2)
1. a kind of recognition methods of the Characters Stuck identifying code based on neutral net, it is characterised in that:Comprise the following steps:
Step 1: intercepting a number of identifying code picture automatically from website, deposit file is standby;
Step 2: carrying out basic pretreatment to the image of the identifying code picture of deposit, original identifying code image is set to be easier to word
Symbol is cut, and specific method is:
By the coloured image gray processing in the identifying code picture of collection;
Gray level image goes dry processing:The noise in corrosion treatment, downscaled images is carried out to image, usage threshold algorithm, selection is suitable
When threshold value, image is subjected to binary conversion treatment, and the image after binary conversion treatment is negated, finally gone using median filtering method
Except most of noise in image;
The size of locating verification code, to step(2)Go it is dry after image remove frame;
To going the image after frame further to remove connected regions all in noise, i.e. selection image by connected region noise reduction hair
Domain, selectes a suitable threshold value, removes the less noise of connected region, retains the larger character of connected region;
Step 3: by the Character segmentation in the pretreated image of step 2 into single character, specific method is:
The gray value of image after step 2 is handled carries out the projection of vertical direction, obtains the projection Nogata of image vertical direction
Figure;
By Character segmentation of the position of trough in the projection histogram of obtained vertical direction, character picture is divided into single by point
Character picture;
Single character picture gray value is subjected to the projection in horizontal direction, the projection for obtaining single character picture horizontal direction is straight
Fang Tu, determines the up-and-down boundary of character by the projection histogram of horizontal direction and is split;
Step 4: the character picture segmented is normalized;
Step 5: training is identified to normalized character picture by artificial neural network;
Step 6: calculating recognition accuracy.
2. a kind of recognition methods of Characters Stuck identifying code based on neutral net according to claim 1, its feature exists
In:In the step 5, it is to the specific method that training is identified in normalized character picture:Selected BP neural network conduct
Training tool, from three-layer neural network structure, i.e. input layer, single hidden layer and output layer, using the instruction of Stochastic gradient method
Practice type and training is identified in adaptive regularized learning algorithm rate.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710682566.0A CN107292311A (en) | 2017-08-10 | 2017-08-10 | A kind of recognition methods of the Characters Stuck identifying code based on neutral net |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710682566.0A CN107292311A (en) | 2017-08-10 | 2017-08-10 | A kind of recognition methods of the Characters Stuck identifying code based on neutral net |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107292311A true CN107292311A (en) | 2017-10-24 |
Family
ID=60104858
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710682566.0A Pending CN107292311A (en) | 2017-08-10 | 2017-08-10 | A kind of recognition methods of the Characters Stuck identifying code based on neutral net |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107292311A (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108446704A (en) * | 2018-03-29 | 2018-08-24 | 哈尔滨理工大学 | A kind of segmentation of adhesion character identifying code and recognition methods |
CN108647683A (en) * | 2018-04-28 | 2018-10-12 | 浙江大学 | It is a kind of that the character made an uproar is added to fight verification code generation method and system based on frequency domain |
CN108874756A (en) * | 2018-06-29 | 2018-11-23 | 广东智媒云图科技股份有限公司 | A kind of identifying code optimization method |
CN109086772A (en) * | 2018-08-16 | 2018-12-25 | 成都市映潮科技股份有限公司 | A kind of recognition methods and system distorting adhesion character picture validation code |
CN109101810A (en) * | 2018-08-14 | 2018-12-28 | 电子科技大学 | A kind of text method for recognizing verification code based on OCR technique |
CN109933969A (en) * | 2017-12-15 | 2019-06-25 | 腾讯科技(深圳)有限公司 | Method for recognizing verification code, device, electronic equipment and readable storage medium storing program for executing |
CN109948621A (en) * | 2019-03-20 | 2019-06-28 | 南京工业大学 | A kind of image procossing and character segmentation method based on picture validation code |
CN110069915A (en) * | 2019-03-12 | 2019-07-30 | 杭州电子科技大学 | A kind of nine grids figure method for recognizing verification code based on contours extract |
CN110222753A (en) * | 2019-05-28 | 2019-09-10 | 北京金山数字娱乐科技有限公司 | Character recognition verification method, system, computer equipment, storage medium and chip |
CN110232375A (en) * | 2018-03-05 | 2019-09-13 | 重庆邮电大学 | A kind of method for recognizing verification code hollow end to end |
CN110298811A (en) * | 2018-03-21 | 2019-10-01 | 北京大学 | Preprocess method, device, terminal and the computer readable storage medium of image |
CN110348449A (en) * | 2019-07-10 | 2019-10-18 | 电子科技大学 | A kind of identity card character recognition method neural network based |
CN110348450A (en) * | 2019-07-15 | 2019-10-18 | 中国工商银行股份有限公司 | Safety evaluation method, device and computer system for image authentication code |
CN110728687A (en) * | 2019-10-15 | 2020-01-24 | 卓尔智联(武汉)研究院有限公司 | File image segmentation method and device, computer equipment and storage medium |
CN111178352A (en) * | 2019-12-13 | 2020-05-19 | 中国建设银行股份有限公司 | Method and device for identifying verification code characters |
CN111274957A (en) * | 2020-01-20 | 2020-06-12 | 阳光人寿保险股份有限公司 | Webpage verification code identification method, device, terminal and computer storage medium |
CN111968091A (en) * | 2020-08-19 | 2020-11-20 | 南京图格医疗科技有限公司 | Method for detecting and classifying lesion areas in clinical image |
CN112257715A (en) * | 2020-11-18 | 2021-01-22 | 西南交通大学 | Method and system for identifying adhesive characters |
CN112364871A (en) * | 2020-10-27 | 2021-02-12 | 重庆大学 | Part code spraying character segmentation method based on improved projection algorithm |
CN114440727A (en) * | 2022-01-28 | 2022-05-06 | 南京航空航天大学 | Vernier caliper zero calibration and automatic calibration device and method |
CN117132989A (en) * | 2023-10-23 | 2023-11-28 | 山东大学 | Character verification code identification method, system and equipment based on convolutional neural network |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070201745A1 (en) * | 2006-01-31 | 2007-08-30 | The Penn State Research Foundation | Image-based captcha generation system |
CN104252620A (en) * | 2014-09-25 | 2014-12-31 | 同程网络科技股份有限公司 | Character-touching graph verification code recognition method |
CN105989377A (en) * | 2015-06-30 | 2016-10-05 | 乐视致新电子科技(天津)有限公司 | Character identification method and apparatus |
CN106446897A (en) * | 2016-09-09 | 2017-02-22 | 浪潮软件股份有限公司 | Identification method of hollow verification code |
-
2017
- 2017-08-10 CN CN201710682566.0A patent/CN107292311A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20070201745A1 (en) * | 2006-01-31 | 2007-08-30 | The Penn State Research Foundation | Image-based captcha generation system |
CN104252620A (en) * | 2014-09-25 | 2014-12-31 | 同程网络科技股份有限公司 | Character-touching graph verification code recognition method |
CN105989377A (en) * | 2015-06-30 | 2016-10-05 | 乐视致新电子科技(天津)有限公司 | Character identification method and apparatus |
CN106446897A (en) * | 2016-09-09 | 2017-02-22 | 浪潮软件股份有限公司 | Identification method of hollow verification code |
Non-Patent Citations (2)
Title |
---|
王璐等: "粘连字符的图片验证码识别", 《计算机工程与应用》 * |
田怀川: "基于神经网络的图形验证码识别及防识别的研究与应用", 《中国优秀硕士论文全文数据库》 * |
Cited By (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109933969B (en) * | 2017-12-15 | 2021-05-07 | 腾讯科技(深圳)有限公司 | Verification code identification method and device, electronic equipment and readable storage medium |
CN109933969A (en) * | 2017-12-15 | 2019-06-25 | 腾讯科技(深圳)有限公司 | Method for recognizing verification code, device, electronic equipment and readable storage medium storing program for executing |
CN110232375A (en) * | 2018-03-05 | 2019-09-13 | 重庆邮电大学 | A kind of method for recognizing verification code hollow end to end |
CN110298811B (en) * | 2018-03-21 | 2021-10-26 | 北京大学 | Image preprocessing method, device, terminal and computer readable storage medium |
CN110298811A (en) * | 2018-03-21 | 2019-10-01 | 北京大学 | Preprocess method, device, terminal and the computer readable storage medium of image |
CN108446704A (en) * | 2018-03-29 | 2018-08-24 | 哈尔滨理工大学 | A kind of segmentation of adhesion character identifying code and recognition methods |
CN108647683A (en) * | 2018-04-28 | 2018-10-12 | 浙江大学 | It is a kind of that the character made an uproar is added to fight verification code generation method and system based on frequency domain |
CN108874756A (en) * | 2018-06-29 | 2018-11-23 | 广东智媒云图科技股份有限公司 | A kind of identifying code optimization method |
CN109101810A (en) * | 2018-08-14 | 2018-12-28 | 电子科技大学 | A kind of text method for recognizing verification code based on OCR technique |
CN109101810B (en) * | 2018-08-14 | 2021-07-06 | 电子科技大学 | Character verification code recognition method based on OCR technology |
CN109086772A (en) * | 2018-08-16 | 2018-12-25 | 成都市映潮科技股份有限公司 | A kind of recognition methods and system distorting adhesion character picture validation code |
CN110069915A (en) * | 2019-03-12 | 2019-07-30 | 杭州电子科技大学 | A kind of nine grids figure method for recognizing verification code based on contours extract |
CN110069915B (en) * | 2019-03-12 | 2021-04-13 | 杭州电子科技大学 | Sudoku graphic verification code identification method based on contour extraction |
CN109948621A (en) * | 2019-03-20 | 2019-06-28 | 南京工业大学 | A kind of image procossing and character segmentation method based on picture validation code |
CN110222753A (en) * | 2019-05-28 | 2019-09-10 | 北京金山数字娱乐科技有限公司 | Character recognition verification method, system, computer equipment, storage medium and chip |
CN110348449A (en) * | 2019-07-10 | 2019-10-18 | 电子科技大学 | A kind of identity card character recognition method neural network based |
CN110348450A (en) * | 2019-07-15 | 2019-10-18 | 中国工商银行股份有限公司 | Safety evaluation method, device and computer system for image authentication code |
CN110728687A (en) * | 2019-10-15 | 2020-01-24 | 卓尔智联(武汉)研究院有限公司 | File image segmentation method and device, computer equipment and storage medium |
CN110728687B (en) * | 2019-10-15 | 2022-08-02 | 卓尔智联(武汉)研究院有限公司 | File image segmentation method and device, computer equipment and storage medium |
CN111178352A (en) * | 2019-12-13 | 2020-05-19 | 中国建设银行股份有限公司 | Method and device for identifying verification code characters |
CN111274957A (en) * | 2020-01-20 | 2020-06-12 | 阳光人寿保险股份有限公司 | Webpage verification code identification method, device, terminal and computer storage medium |
CN111968091A (en) * | 2020-08-19 | 2020-11-20 | 南京图格医疗科技有限公司 | Method for detecting and classifying lesion areas in clinical image |
CN111968091B (en) * | 2020-08-19 | 2022-04-01 | 南京图格医疗科技有限公司 | Method for detecting and classifying lesion areas in clinical image |
CN112364871A (en) * | 2020-10-27 | 2021-02-12 | 重庆大学 | Part code spraying character segmentation method based on improved projection algorithm |
CN112257715A (en) * | 2020-11-18 | 2021-01-22 | 西南交通大学 | Method and system for identifying adhesive characters |
CN114440727A (en) * | 2022-01-28 | 2022-05-06 | 南京航空航天大学 | Vernier caliper zero calibration and automatic calibration device and method |
CN117132989A (en) * | 2023-10-23 | 2023-11-28 | 山东大学 | Character verification code identification method, system and equipment based on convolutional neural network |
CN117132989B (en) * | 2023-10-23 | 2024-01-26 | 山东大学 | Character verification code identification method, system and equipment based on convolutional neural network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107292311A (en) | A kind of recognition methods of the Characters Stuck identifying code based on neutral net | |
US10817741B2 (en) | Word segmentation system, method and device | |
CN107305630B (en) | Text sequence identification method and device | |
CN104751142B (en) | A kind of natural scene Method for text detection based on stroke feature | |
CN107679531A (en) | Licence plate recognition method, device, equipment and storage medium based on deep learning | |
CN104299008B (en) | Vehicle type classification method based on multi-feature fusion | |
CN110598686B (en) | Invoice identification method, system, electronic equipment and medium | |
CN106874909A (en) | A kind of recognition methods of image character and its device | |
CN107871125A (en) | Architecture against regulations recognition methods, device and electronic equipment | |
CN105760891A (en) | Chinese character verification code recognition method | |
CN111145209A (en) | Medical image segmentation method, device, equipment and storage medium | |
CN111783757A (en) | OCR technology-based identification card recognition method in complex scene | |
CN112966685B (en) | Attack network training method and device for scene text recognition and related equipment | |
CN111680690A (en) | Character recognition method and device | |
CN105447508A (en) | Identification method and system for character image verification codes | |
CN107992807A (en) | A kind of face identification method and device based on CNN models | |
CN113159045A (en) | Verification code identification method combining image preprocessing and convolutional neural network | |
US20210390667A1 (en) | Model generation | |
CN109886978A (en) | A kind of end-to-end warning information recognition methods based on deep learning | |
CN106503694A (en) | Digit recognition method based on eight neighborhood feature | |
CN113673384A (en) | Oracle character detection method for guiding texture feature autonomous learning by LM filter bank | |
CN105404885A (en) | Two-dimensional character graphic verification code complex background noise interference removal method | |
CN110427909B (en) | Mobile terminal driving license detection method and system, electronic equipment and storage medium | |
CN116597466A (en) | Engineering drawing text detection and recognition method and system based on improved YOLOv5s | |
CN108171229A (en) | A kind of recognition methods of hollow adhesion identifying code and system |
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: 20171024 |