CN115497097A - Inclined Chinese character click verification code identification method - Google Patents

Inclined Chinese character click verification code identification method Download PDF

Info

Publication number
CN115497097A
CN115497097A CN202211046062.7A CN202211046062A CN115497097A CN 115497097 A CN115497097 A CN 115497097A CN 202211046062 A CN202211046062 A CN 202211046062A CN 115497097 A CN115497097 A CN 115497097A
Authority
CN
China
Prior art keywords
foreground
chinese character
verification code
character
sub
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
Application number
CN202211046062.7A
Other languages
Chinese (zh)
Inventor
池凯凯
金鑫豪
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Zhejiang University of Technology ZJUT
Original Assignee
Zhejiang University of Technology ZJUT
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Zhejiang University of Technology ZJUT filed Critical Zhejiang University of Technology ZJUT
Priority to CN202211046062.7A priority Critical patent/CN115497097A/en
Publication of CN115497097A publication Critical patent/CN115497097A/en
Pending legal-status Critical Current

Links

Images

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
    • G06V30/146—Aligning or centring of the image pick-up or image-field
    • G06V30/1475—Inclination or skew detection or correction of characters or of image to be recognised
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/30—Authentication, i.e. establishing the identity or authorisation of security principals
    • G06F21/31—User authentication
    • G06F21/36—User authentication by graphic or iconic representation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using 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
    • G06V30/14—Image acquisition
    • G06V30/1444—Selective acquisition, locating or processing of specific regions, e.g. highlighted text, fiducial marks or predetermined fields
    • 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
    • G06V30/16—Image preprocessing
    • G06V30/162—Quantising the image signal
    • 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
    • G06V30/18—Extraction of features or characteristics of the image
    • G06V30/1801—Detecting partial patterns, e.g. edges or contours, or configurations, e.g. loops, corners, strokes or intersections
    • 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
    • G06V30/19—Recognition using electronic means
    • G06V30/19007—Matching; Proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2221/00—Indexing scheme relating to security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F2221/21—Indexing scheme relating to G06F21/00 and subgroups addressing additional information or applications relating to security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F2221/2133—Verifying human interaction, e.g., Captcha
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/07—Target detection

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Multimedia (AREA)
  • Software Systems (AREA)
  • Computer Security & Cryptography (AREA)
  • Evolutionary Computation (AREA)
  • Medical Informatics (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Computing Systems (AREA)
  • General Health & Medical Sciences (AREA)
  • Databases & Information Systems (AREA)
  • Computer Hardware Design (AREA)
  • General Engineering & Computer Science (AREA)
  • Geometry (AREA)
  • Character Input (AREA)

Abstract

The invention discloses a recognition method of a click-selection verification code of inclined Chinese characters, which predicts a character position frame through a target detection model YOLOX, the character position frame is used as an input parameter for foreground segmentation, chinese character foreground is segmented, a segmented foreground image is binarized, a minimum circumscribed rectangle frame is drawn, the slope and the angle of the circumscribed rectangle frame are calculated, 3 optimal correction angles are given, an original image is rotated to obtain 3 correction images with different angles, the correction images are input into a character recognition model, recognition probabilities in a candidate Chinese character range are compared, a correction image corresponding to the maximum probability is a final correction result, and the maximum value is searched in a finally generated probability matrix in a circulating mode and the row and column are deleted to obtain the optimal matching result of the character position frame and the candidate Chinese characters. The invention corrects the inclined Chinese characters in the verification code and improves the identification accuracy of the Chinese character click verification code.

Description

Inclined Chinese character click verification code identification method
Technical Field
The application belongs to the technical field of identifying codes, and particularly relates to a method for identifying inclined Chinese character click identifying codes.
Background
The verification code is an important means for preventing malicious attacks of a website, the verification code is various in forms, the Chinese character click verification code is an excellent design, and compared with the traditional input character type verification code, the Chinese character click verification code is high in safety and very friendly to users.
The Chinese character click verification code effectively prevents malicious attack, and meanwhile, the automatic program without malicious intent is greatly influenced, so that the originally automatic flow is forced to be separated, and the working efficiency is reduced. Meanwhile, in order to evaluate the security of Chinese character click verification codes, a corresponding verification code identification method is often required to be designed for countermeasure testing.
The existing click verification code identification technology is only suitable for conventional non-inclined characters, and an effective solution is not provided for identifying the Chinese character click verification code with an inclined angle.
Therefore, the method for identifying the inclined Chinese character clicking verification code is provided to improve the identification accuracy of the Chinese character clicking verification code and provide a reference solution for verification code identification test and malicious-free automatic programs.
Disclosure of Invention
The method can effectively correct the Chinese character inclination within a certain angle range, and effectively improve the accuracy of character recognition by utilizing the known conditions in the title of the verification code to the maximum extent.
In order to achieve the purpose, the technical scheme adopted by the invention is as follows:
a method for identifying a tilted Chinese character click verification code comprises the following steps:
acquiring a verification code picture, and performing target detection to obtain position rectangular frames of all Chinese characters in the picture;
performing foreground and background segmentation by taking the obtained rectangular frame as a basis for self-adaptive foreground and background segmentation, and adding the foreground containing the Chinese characters obtained by the foreground and background segmentation to obtain a verification code picture only containing the Chinese character foreground;
cutting the verification code picture only containing the Chinese character foreground according to the position of the rectangular frame, cutting out a sub-picture corresponding to each Chinese character, converting the sub-picture into a binary image, detecting the outline of the binary image and obtaining a corresponding minimum circumscribed rectangular frame;
calculating the slope of two adjacent edges of the circumscribed rectangular frame of each subgraph, converting the slope into an inclination angle, respectively correcting the angle of the subgraph according to the obtained inclination angle, and taking the subgraph original graph and the two corrected subgraphs as subgraphs to be recognized;
identifying the three sub-images to be identified of each Chinese character by respectively adopting a character identification model to obtain the probability of identifying each sub-image to be identified as each candidate Chinese character, selecting the sub-image to be identified corresponding to the maximum probability from the sub-images to be identified, and identifying the sub-images to be identified selected from all the Chinese characters in the verification code picture and the probability of identifying the sub-images to be identified as each candidate Chinese character to form a probability matrix;
and circularly using a mode of searching the maximum value and deleting the row and the column in the probability matrix to obtain the best matching result of the character position rectangular frame and the candidate Chinese characters.
Further, the target detection is performed to obtain the position rectangular frames of all the Chinese characters in the picture, and the target detection is performed by using YOLOX.
Further, the obtained rectangular frame is used as a basis for self-adaptive foreground and background segmentation to perform foreground and background segmentation, and a GrabCut foreground and background segmentation algorithm in OpenCV is adopted.
Further, the subgraph is converted into a binary image, the outline of the binary image is detected, a corresponding circumscribed rectangle frame is obtained, and an OpenCV outline detection method is adopted.
Further, the character recognition models are respectively adopted for recognition, and the adopted character recognition model is paddleOCR.
Further, the foreground and background segmentation is performed by using the obtained rectangular frame as a basis for adaptive foreground and background segmentation, and the foregrounds containing the Chinese characters obtained by the foreground and background segmentation are added to obtain the verification code picture only containing the Chinese character foregrounds, and the method further comprises the following steps:
and if the area ratio of the segmented foreground is less than 10% of the optimal probability, judging that the foreground of the Chinese character is mistakenly taken as the background and is removed, and in this case, giving up the segmentation result and adopting the original image as the verification code picture containing the foreground of the Chinese character.
Compared with the existing click verification code identification method, the method corrects the inclined Chinese characters in the verification code and improves the identification accuracy of the Chinese character click verification code.
Drawings
FIG. 1 is a flow chart of the identification method for Chinese character clicking verification codes of the present application.
FIG. 2 is a foreground diagram of a segmented Chinese character according to the present application.
Fig. 3 is a diagram of a minimum bounding rectangle framing a binary image outline according to the present application.
FIG. 4 is a schematic diagram of a Chinese character tilt angle according to the present application.
FIG. 5 is a probability list of candidate syndrome recognition results.
Fig. 6 is a list of final recognition results.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
In one embodiment, as shown in fig. 1, a method for identifying a tilted chinese character click verification code is provided, including:
s1, obtaining a verification code picture, and carrying out target detection to obtain position rectangular frames of all Chinese characters in the picture.
When verification code verification is needed, an inclined Chinese character clicking interface is popped up, a user is prompted to click the Chinese characters on the interface according to a prompting sequence, and a clicked verification code picture is submitted after the user clicks.
Then, target detection is performed on the verification code picture, for example, by using YOLOX, and the position information of the rectangular frame of all the tilted Chinese characters in the verification code picture is obtained.
In this embodiment, N is used to represent the number of all the chinese characters in the verification code picture, so that the rectangular frame position vectors of N chinese characters are obtained through target detection: box = [ ] 1 ,box 2 ,...,box N ]Wherein the ith rectangular box is represented as: box i ＝[x i1 ,y i1 ,x i2 ,y i2 ]，(x i1 ,y i1 ) Denotes the coordinates of the vertex at the upper left corner of the rectangular box, (x) i2 ,y i2 ) Representing the coordinates of the vertex of the lower right corner of the rectangular box.
And S2, performing foreground and background segmentation by taking the obtained rectangular frame as a basis for self-adaptive foreground and background segmentation, and adding the foreground containing the Chinese characters obtained by the foreground and background segmentation to obtain a verification code picture only containing the Chinese character foreground.
In the embodiment, a GrabCut foreground and background segmentation algorithm in OpenCV is adopted, N times are circulated, and the positions box of each Chinese character rectangular frame are sequentially subjected to i Performing foreground and background segmentation, i.e. rectangular frame position box i The foreground and background styles are carried out as the input parameters of the algorithm, the Chinese character foreground is reserved after the segmentation, and the Chinese character foreground is recorded as grabcutImg i Adding the N divided foregrounds to obtain a verification code picture grabcutImg with only N Chinese character foregrounds:
Figure BDA0003822432000000041
in a specific embodiment, the obtained verification code picture containing the foreground of the tilted chinese character is shown in fig. 2.
And S3, cutting the verification code picture only containing the Chinese character foreground according to the position of the rectangular frame, cutting out a sub-image corresponding to each Chinese character, converting the sub-image into a binary image, detecting the outline of the binary image and obtaining a corresponding minimum circumscribed rectangular frame.
In the embodiment, each Chinese character is cut out from the verification code picture containing the inclined Chinese character foregroundSub-graph boxImg = [ boxImg ] corresponding to character rectangular frame 1 ,boxImg 2 ,...,boxImg N ]And converting the image into a binary image, and then obtaining the outline of the binary image by using an OpenCV outline detection method.
Then, the contour points are framed by the minimum circumscribed rectangle, and as shown in fig. 3, four point coordinates rect = [ rect ] of the circumscribed rectangle are obtained 1 ,rect 2 ,...,rect N ]Wherein the ith bounding rectangle is represented as: rect i ＝[[x i1 ,y i1 ],[x i2 ,y i2 ],[x i3 ,y i3 ],[x i4 ,y i4 ]],[x i1 ,y i1 ],[x i2 ,y i2 ],[x i3 ,y i3 ],[x i4 ,y i4 ]Respectively representing the coordinates of the four vertices.
And S4, calculating the slopes of two adjacent edges of the circumscribed rectangular frame of each sub-image, converting the slopes into inclination angles, respectively correcting the angles of the sub-images according to the obtained inclination angles, and taking the sub-image original image and the corrected two sub-images as sub-images to be recognized.
In the present embodiment, three different angles angle = [ angle ] are used 1 ,angle 2 ,angle 3 ]The subgraph of (a) is taken as a subgraph to be recognized.
Specifically, as shown in fig. 4, the left and the next two points of the four points are found, and the slope k = -1/((y 1-y 2)/(x 1-x 2)) of the straight line where the points are located is calculated, and then the angle is converted to an angle 1 = arctan (k) × 180/pi, and likewise, the angle corresponding to the straight line of the right and the next two points is angle 2 And there is a relation angle 1 ＝angle 2 And +90 degrees, and then respectively rotating the boxImg by a specified angle to obtain two subgraphs to be recognized.
The third sub-image to be recognized is the sub-image original image, i.e. angle 3 ＝0°。
And S5, respectively identifying the three sub-images to be identified of each Chinese character by adopting a character identification model to obtain the probability of identifying each sub-image to be identified as each candidate Chinese character, selecting the sub-image to be identified corresponding to the maximum probability from the sub-images to be identified, and forming a probability matrix by using the sub-images to be identified selected from all Chinese characters in the verification code picture and the probability of identifying the sub-images to be identified as each candidate Chinese character.
In this embodiment, assume that the verification code hint includes M (M)<Word = N) Chinese characters word = [ word = 1 ,word 2 ,...,word M ]As candidate Chinese characters.
Aiming at each rectangular frame, obtaining sub-images to be recognized at three different angles, respectively inputting the sub-images to be recognized into a character recognition model paddleOCR for recognition, and obtaining probabilities respectively corresponding to M characters of the candidate Chinese characters from model output:
answer＝[answer 1 ,answer 2 ,answer 3 ](answer i ＝[probability 1 ,probability 2 ,...,probability M ])。
for example, there are three candidate Chinese characters, which are meat, sandwiched and pie. The answer containing the maximum probability is selected from the probability tables of three candidate Chinese characters:
bestAnswer＝answer i ,max[max(answer),max(answer 2 ),max(answer 3 )]∈answer i 。
because its maximum probability can correspond to a word in the candidate Chinese characters, as shown in FIG. 5, the maximum probability is marked in bold and the probability array bestAnswer is marked by underlining, so that each boxImg corresponds to an optimal correction angle and the probability array bestAnswer of a candidate Chinese character, and N boxImgs correspond to a probability matrix mat of N M.
Selecting a probability maximum value from mat, and determining the corresponding small graph boxImg in the row and column (i, j) i And word of candidate Chinese characters j Then, the ith row and the jth column are deleted from the mat, and the best matching result is obtained by repeating M times.
As shown in fig. 6, the first round has a maximum probability of 0.94733 in the second row and the second column, which means that the second graph corresponds to the second candidate chinese character, and then the second row and the second column are deleted, and the second round has a maximum probability of 0.85409 in the third row and the first column, which means that the third graph corresponds to the first candidate chinese character, and then the third row and the first column are deleted, and the third round has only 0.00072 in the first row and the second column, which means that the first graph corresponds to the second candidate chinese character.
And determining the sequence of the positions box of the rectangular frames of the Chinese characters required to be clicked according to the obtained matching result and the sequence of the Chinese characters word in the candidate Chinese characters.
In this embodiment, "GrabCut" is an interactive image segmentation algorithm, and the foreground position is marked by a rectangular frame, so that automatic segmentation can be performed. For a few scenes with complex backgrounds, the segmentation effect is not as expected, the foreground of the Chinese character is mistakenly taken as the background and is removed, whether the situation happens or not can be judged according to the area ratio of the segmented foreground, and in this case, grabCut segmentation is abandoned and the original image is reserved.
For example, if the area ratio of the divided foreground is less than 10% of the preferred probability, the foreground of the Chinese character is judged to be mistaken for the background and removed, in this case, the division result is abandoned, and the original image is used as the verification code image containing the foreground of the Chinese character.
In this embodiment, the term "binary image" refers to that a three-channel image is converted into a single-channel gray scale image, and then all pixels which are not 0 are set to 255.
In this embodiment, "contour detection" and "minimum bounding rectangle" are existing methods of OpenCV, but the connection of the two methods requires manual summarization of multi-level contour points.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent application shall be subject to the appended claims.

Claims (6)

1. A method for identifying a tilted Chinese character click verification code is characterized by comprising the following steps:
acquiring a verification code picture, and performing target detection to obtain position rectangular frames of all Chinese characters in the picture;
taking the obtained rectangular frame as a basis for self-adaptive foreground and background segmentation to perform foreground and background segmentation, and adding the foreground containing the Chinese characters obtained by the foreground and background segmentation to obtain a verification code picture only containing the Chinese character foreground;
cutting the verification code picture only containing the Chinese character foreground according to the position of the rectangular frame, cutting out a sub-picture corresponding to each Chinese character, converting the sub-picture into a binary image, detecting the outline of the binary image and obtaining a corresponding minimum circumscribed rectangular frame;
calculating the slope of two adjacent edges of the circumscribed rectangular frame of each subgraph, converting the slope into an inclination angle, respectively correcting the angle of the subgraph according to the obtained inclination angle, and taking the subgraph original graph and the two corrected subgraphs as subgraphs to be recognized;
identifying the three sub-images to be identified of each Chinese character by respectively adopting a character identification model to obtain the probability of identifying each sub-image to be identified as each candidate Chinese character, selecting the sub-image to be identified corresponding to the maximum probability from the sub-images to be identified, and identifying the sub-images to be identified selected from all the Chinese characters in the verification code picture and the probability of identifying the sub-images to be identified as each candidate Chinese character to form a probability matrix;
and circularly using the mode of searching the maximum value and deleting the row and column in the probability matrix to obtain the best matching result of the character position rectangular frame and the candidate Chinese characters.
2. The method for identifying inclined Chinese character click verification codes according to claim 1, wherein the position rectangular frames of all Chinese characters in the picture are obtained by performing target detection, and YOLOX is adopted to perform target detection.
3. The method for identifying tilted Chinese character click verification codes according to claim 1, wherein the foreground and background segmentation is performed by using the obtained rectangular box as a basis for adaptive foreground and background segmentation, and a GrabCut foreground and background segmentation algorithm in OpenCV is adopted.
4. The method for identifying tilted Chinese character click verification codes according to claim 1, wherein the subgraph is converted into a binary image, the outline of the binary image is detected, a corresponding circumscribed rectangular frame is obtained, and an OpenCV outline detection method is adopted.
5. The method for recognizing a tilted Chinese character click verification code according to claim 1, wherein said recognition is performed by using a character recognition model, and the character recognition model used is paddleOCR.
6. The method for identifying an inclined chinese character click verification code according to claim 1, wherein the foreground and background segmentation is performed using the obtained rectangular frame as a basis for adaptive foreground and background segmentation, and the foregrounds containing the chinese characters obtained by the foreground and background segmentation are added to obtain a verification code picture containing only the chinese character foregrounds, further comprising:
and if the area ratio of the divided foreground is less than 10% of the optimal probability, judging that the foreground of the Chinese character is mistakenly taken as the background and is removed, in this case, giving up the division result, and adopting the original image as the verification code image containing the foreground of the Chinese character.
CN202211046062.7A 2022-08-30 2022-08-30 Inclined Chinese character click verification code identification method Pending CN115497097A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202211046062.7A CN115497097A (en) 2022-08-30 2022-08-30 Inclined Chinese character click verification code identification method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202211046062.7A CN115497097A (en) 2022-08-30 2022-08-30 Inclined Chinese character click verification code identification method

Publications (1)

Publication Number Publication Date
CN115497097A true CN115497097A (en) 2022-12-20

Family

ID=84466554

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202211046062.7A Pending CN115497097A (en) 2022-08-30 2022-08-30 Inclined Chinese character click verification code identification method

Country Status (1)

Country Link
CN (1) CN115497097A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118072324A (en) * 2024-04-23 2024-05-24 浙江保融科技股份有限公司 Verification code identification method based on semantic point character

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118072324A (en) * 2024-04-23 2024-05-24 浙江保融科技股份有限公司 Verification code identification method based on semantic point character

Similar Documents

Publication Publication Date Title
CN107609549B (en) Text detection method for certificate image in natural scene
CN108416279B (en) Table analysis method and device in document image
Korus et al. Multi-scale fusion for improved localization of malicious tampering in digital images
CN107967475B (en) Verification code identification method based on window sliding and convolutional neural network
CN110032998B (en) Method, system, device and storage medium for detecting characters of natural scene picture
CN110569830A (en) Multi-language text recognition method and device, computer equipment and storage medium
US5410611A (en) Method for identifying word bounding boxes in text
Korus et al. Evaluation of random field models in multi-modal unsupervised tampering localization
US8325998B2 (en) Multidirectional face detection method
CN110276351B (en) Multi-language scene text detection and identification method
CN109343920B (en) Image processing method and device, equipment and storage medium thereof
CN110647882A (en) Image correction method, device, equipment and storage medium
WO2009094446A1 (en) Method and apparatus for cropping images
RU2648638C2 (en) Methods and systems of effective automatic recognition of symbols using a multiple clusters of symbol standards
CN110598581B (en) Optical music score recognition method based on convolutional neural network
CN109948621B (en) Image processing and character segmentation method based on picture verification code
WO2021159802A1 (en) Graphical captcha recognition method, apparatus, computer device, and storage medium
WO2021128851A1 (en) Picture classification method and apparatus based on digital slice, and computer device
CN113989604A (en) Tire DOT information identification method based on end-to-end deep learning
CN115497097A (en) Inclined Chinese character click verification code identification method
CN111626145A (en) Simple and effective incomplete form identification and page-crossing splicing method
CN112949649B (en) Text image identification method and device and computing equipment
CN113887375A (en) Text recognition method, device, equipment and storage medium
CN112434640B (en) Method, device and storage medium for determining rotation angle of document image
CN109740607A (en) The incomplete region detection of character picture and incomplete character picture restoration methods

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