CN107016417A - A kind of method and device of character recognition - Google Patents
A kind of method and device of character recognition Download PDFInfo
- Publication number
- CN107016417A CN107016417A CN201710193761.7A CN201710193761A CN107016417A CN 107016417 A CN107016417 A CN 107016417A CN 201710193761 A CN201710193761 A CN 201710193761A CN 107016417 A CN107016417 A CN 107016417A
- Authority
- CN
- China
- Prior art keywords
- images
- text box
- recognized
- target image
- pixel
- 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
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/24—Character recognition characterised by the processing or recognition method
- G06V30/242—Division of the character sequences into groups prior to recognition; Selection of dictionaries
- G06V30/244—Division of the character sequences into groups prior to recognition; Selection of dictionaries using graphical properties, e.g. alphabet type or font
- G06V30/2455—Discrimination between machine-print, hand-print and cursive writing
Abstract
The present invention provides a kind of method and device of character recognition, and methods described is applied to the identification to character in image, and described image includes the text box of undefined position, and the character is filled in text box, and this method includes：Determine text box location in images to be recognized；According to the position of text box, the corresponding target image of interception images to be recognized Chinese version frame；Calculate the target direction histogram of gradients HOG features corresponding to target image and describe operator；According to the good support vector machines grader of target HOG Feature Descriptors and training in advance, the character to be identified of target image is obtained.Using the embodiment of the present invention, by the position that the text box is determined in images to be recognized, and according to the text box position, intercept the corresponding target image of the images to be recognized Chinese version frame, realize the positive location to target image and interception, avoid and intercept the deviation that object region occurs according to default fixed coordinates, improve recognition accuracy.
Description
Technical field
The application is related to the method and device in Computer Recognition Technology field, more particularly to character recognition.
Background technology
At present, many scenes can all be applied to character recognition technologies, for example：During going over examination papers, for scanning to computer
Paper, examination question, answer part are read and appraised mainly by being accomplished manually, for the part of student number, name, fraction record and count then
It can be completed by the identification of computer.This requires identification of the computer to numeral and word to reach certain accuracy rate.
In to paper exemplified by the identification of character, in the prior art, due to requiring the image file lattice as identification object
Formula is unified, and size is fixed, so can be first image file by the paper scanning of paper, and intercepted according to default fixed coordinates
The object regions such as student number, name and fraction.Then HOG (Histogram of Oriented Gradient, side are passed through
To histogram of gradients) gradient orientation histogram of target image is calculated and counts to generate HOG Feature Descriptors, finally will generation
Good support vector machines (the Support Vector Machine) grader of HOG Feature Descriptors input training in advance enter
Row identification.But, because the scanning of paper is by being accomplished manually, there is certain deviation in its image file ultimately produced, if pressed
Target image is intercepted according to default fixed coordinates, then the target image of interception can be made deviation also occur, non-targeted is even truncated to
Image-region, ultimately results in identification mistake or failure.It can be seen that, the accuracy rate of existing character recognition technologies identification is relatively low.
The content of the invention
The method and apparatus that the embodiment of the present invention provides character recognition, the accuracy rate for solving prior art identification is relatively low
The problem of.
There is provided a kind of method of character recognition for first aspect according to embodiments of the present invention, it is characterised in that methods described
Applied to the identification to character in image, described image includes the text box of undefined position, and the character fills in described
In text box, methods described includes：
Determine text box location in images to be recognized；
According to the position of the text box, the corresponding target image of the images to be recognized Chinese version frame is intercepted；
Calculate the target direction histogram of gradients HOG features corresponding to the target image and describe operator；
According to the good support vector machines grader of the target HOG Feature Descriptors and training in advance, obtain described
HOG Feature Descriptors pass corresponding with character is preserved in the character to be identified of target image, the SVM classifier trained
System.
There is provided a kind of device of character recognition for second aspect according to embodiments of the present invention, it is characterised in that described device
Applied to the identification to character in image, described image includes the text box of undefined position, and the character fills in described
In text box, described device includes：
Determining unit, for determining text box location in images to be recognized；
Interception unit, for the position according to the text box, intercepts the corresponding mesh of the images to be recognized Chinese version frame
Logo image；
Computing unit, describes to calculate for calculating the target direction histogram of gradients HOG features corresponding to the target image
Son；
Acquiring unit, for according to the target HOG Feature Descriptors and training in advance good support vector machines point
Class device, obtains in the character to be identified of the target image, the SVM classifier trained and preserves HOG Feature Descriptors
With the corresponding relation of character.
From above technical scheme, the embodiment of the present invention is by the images to be recognized, determining the text box
Position, and according to the text box position, intercept the corresponding target image of the images to be recognized Chinese version frame, realize pair
The positive location of target image and interception, it is to avoid according to default fixed coordinates intercept that object region occurs it is inclined
Difference, improves recognition accuracy.
Brief description of the drawings
Fig. 1 is one embodiment flow chart of the method for character recognition of the present invention；
Fig. 2 is another embodiment flow chart of the method for character recognition of the present invention；
A kind of hardware structure diagram of Fig. 3 equipment where the device of character recognition of the present invention；
Fig. 4 is one embodiment block diagram of the device of character recognition of the present invention.
Embodiment
In order that those skilled in the art are better understood from the technical scheme in the embodiment of the present invention, and make of the invention real
Applying the above-mentioned purpose of example, feature and advantage can be more obvious understandable, below in conjunction with the accompanying drawings to the technology in the embodiment of the present invention
Scheme is described in further detail.
Fig. 1 is one embodiment flow chart of the method for character recognition of the present invention, and methods described is applied to word in image
The identification of symbol, described image includes the text box of undefined position, and the character is filled in the text box, methods described
Comprise the following steps：
Step 101：Determine text box location in images to be recognized.
In an optional mode, above-mentioned images to be recognized can first be done binary conversion treatment, then in the two-value
The coordinate of the text box is determined in images to be recognized after change.Wherein, can be with to the binary conversion treatment of above-mentioned images to be recognized
It is global binary conversion treatment or local auto-adaptive binary conversion treatment.
Step 102：According to the position of the text box, the corresponding target figure of the images to be recognized Chinese version frame is intercepted
Picture.
In an optional example, if finding the seat of the different text box of multigroup area in above-mentioned images to be recognized
The text box coordinate of mark, then Retention area minimum.Reason for this is that the straight line of images to be recognized Chinese version frame is general
A plurality of adjacent pixel column or row can be taken to represent the width of its lines, OpenCV (Open Source are being used
Computer Vision Library, computer vision of increasing income storehouse) when text box is identified, because straight line takes
A plurality of pixel column (OK), so can be identified as a plurality of straight line, such a text box will be identified as area from small to large
Multiple text boxes.
In another optional example, interception target image after, can according to character in target image area and
Height is screened to multiple target images of interception, and an optional screening mode is as follows：
After interception target image, by all target image binaryzations, and the target that Pixel Dimensions are 28*28 is normalized to
Image；
In target image after any above-mentioned normalization, the first picture that longitudinal coordinate is maximum and gray value is minimum is filtered out
Vegetarian refreshments, and the second pixel that longitudinal coordinate is minimum and gray value is minimum, and calculate first pixel and described second
The longitudinal coordinate of pixel is poor；
In target image after any above-mentioned normalization, the 3rd picture that lateral coordinates are maximum and gray value is minimum is filtered out
Vegetarian refreshments, and the 4th pixel that lateral coordinates are minimum and gray value is minimum, and calculate the 3rd pixel and the described 4th
The lateral coordinates of pixel are poor；
If the longitudinal coordinate difference is less than default minimum height values or the longitudinal coordinate is poor poor with the lateral coordinates
Product be less than default minimal face product value, then delete the target image after any normalization.
Step 103：Calculate the target direction histogram of gradients HOG features corresponding to the target image and describe operator.
In an optional mode, the target image of 28*28 Pixel Dimensions can will be normalized in step 102, point
4 14*14 block is cut into, the unit for being then divided into 4 7*7 by each piece again is carrying out HOG features and described the calculating of operator
When, first 7*7 unit can be calculated, the HOG features of 4 7*7 units are then described into the HOG that operator is composed in series block
Feature describes operator, the HOG features of 4 blocks finally are described into operator is composed in series the HOG features of the target image to describe operator.
Step 104：According to the good support vector machines grader of the target HOG Feature Descriptors and training in advance,
Obtain in the character to be identified of the target image, the SVM classifier trained and preserve HOG Feature Descriptors and character
Corresponding relation.
From above technical scheme, the embodiment of the present invention is by the images to be recognized, determining the text box
Position, and according to the text box position, intercept the corresponding target image of the images to be recognized Chinese version frame, realize pair
The positive location of target image and interception, it is to avoid according to default fixed coordinates intercept that object region occurs it is inclined
Difference, improves recognition accuracy.
Fig. 2 is another embodiment flow chart of the method for character recognition of the present invention, and methods described is applied to in image
The identification of character, described image includes the text box of undefined position, and the character is filled in the text box, the side
Method comprises the following steps：
Step 201：Above-mentioned images to be recognized is done into binary conversion treatment.
Can be global binary conversion treatment or office to the binary conversion treatment of above-mentioned images to be recognized in this step
Portion's self-adaption binaryzation processing.
Step 202：The coordinate of the text box is determined in images to be recognized after above-mentioned binaryzation.
In this step, OpenCV can be used to recognize above-mentioned text box, and determine the coordinate of above-mentioned text box.
Step 203：The area of each text box is determined according to above-mentioned text box coordinate, the minimum text box of Retention area is sat
Mark.
It is in this step, the reason for Retention area minimum text box coordinate：One of images to be recognized Chinese version frame
Straight line can typically take a plurality of adjacent pixel column or row to represent the width of its lines, and text box is carried out using OpenCV
During identification, because straight line takes a plurality of pixel column (OK), so can be identified as a plurality of straight line, such a text box is just
The multiple text boxes of area from small to large can be identified as.
, can also be and right according to the number of text box to be determined in images to be recognized in an optional example
The statistics and ranking results of the text box coordinate of above-mentioned reservation, prediction may leak the text box coordinate of choosing, for example：
After being screened by area, two groups of text box coordinates are remained with, wherein the apex coordinate of the first text box is (0,0)
(0,1) (1,1) (1,0), wherein the apex coordinate of the second text box is (2,0) (2,1) (3,1) (3,0).By counting and arranging
Sequence, calculates between above-mentioned two groups of coordinates one, interval just and the text box of the one the second text box homalographics, and its coordinate is：
(1,0) (1,1) (2,1) (2,0).Number in view of text box to be determined in images to be recognized is 3, so prediction (1,0)
The text box coordinate that (1,1) (2,1) (2,0) are selected for leakage.
Step 204：According to the coordinate of the text box retained, above-mentioned target image is intercepted.
Step 205：All target images are normalized to the target image of 28*28 Pixel Dimensions.
Step 206：In target image after any above-mentioned normalization, filter out longitudinal coordinate maximum and gray value is minimum
The first pixel, and the second pixel that longitudinal coordinate is minimum and gray value is minimum, and calculate first pixel with
The longitudinal coordinate of second pixel is poor.
Step 207：In target image after any above-mentioned normalization, filter out lateral coordinates maximum and gray value is minimum
The 3rd pixel, and the 4th pixel that lateral coordinates are minimum and gray value is minimum, and calculate the 3rd pixel with
The lateral coordinates of 4th pixel are poor.
Step 208：It is determined that above-mentioned longitudinal coordinate difference be less than default minimum height values or the longitudinal coordinate it is poor with it is described
The product of lateral coordinates difference is less than after default minimal face product value, deletes the target image after any of the above-described normalization.
Step 209：Calculate the target HOG features corresponding to above-mentioned target image and describe operator.
In an optional mode, the target image of 28*28 Pixel Dimensions can will be normalized to, 4 14* are divided into
14 block, the unit for being then divided into 4 7*7 by each piece again, when carrying out HOG features and describing the calculating of operator, Ke Yixian
7*7 unit is calculated, the HOG features of 4 7*7 units are then described into the HOG features description that operator is composed in series block
Operator, finally describes operator by the HOG features of 4 blocks and is composed in series the HOG features of the target image to describe operator.
Step 210：According to the good support vector machines grader of above-mentioned target HOG Feature Descriptors and training in advance,
Obtain the character to be identified of above-mentioned target image.
In this step, the corresponding relation of HOG Feature Descriptors and character is preserved in the above-mentioned SVM classifier trained.
From above technical scheme, on the one hand, the embodiment of the present invention is by the images to be recognized, it is determined that described
The position of text box, and according to the text box position, the corresponding target image of the images to be recognized Chinese version frame is intercepted, it is real
The positive location to target image and interception are showed, it is to avoid occurred according to default fixed coordinates interception object region
Deviation, improve recognition accuracy.On the other hand, the embodiment of the present invention determines the area of text box, screening by statistics
Go out the minimum text box of area, realize the further screening to recognizing text box, improve to the accurate of target image interception
Property, further increase recognition accuracy.Another further aspect, the embodiment of the present invention is after interception target image, by counting target
The height and area of hand-written character in image, realize and further go wrong and screening to target image, further increase knowledge
Other accuracy rate.
Embodiment with the method for foregoing character recognition is corresponding, and present invention also provides the implementation of the device of character recognition
Example.
The embodiment of the device of the application character recognition can be realized by software, can also pass through hardware or software and hardware
With reference to mode realize.It is the processing by equipment where it as the device on a logical meaning exemplified by implemented in software
Corresponding computer program instructions in nonvolatile memory are read what operation in internal memory was formed by device.From hardware view
Speech, as shown in figure 3, a kind of hardware structure diagram of the device place equipment for the application character recognition, except the processing shown in Fig. 3
Outside device, internal memory, network interface and nonvolatile memory, the equipment in embodiment where device is generally according to the equipment
Actual functional capability, can also include other hardware, this is repeated no more.
Fig. 4 is refer to, is one embodiment block diagram of the device of character recognition of the present invention, described device is applied to image
The identification of middle character, described image includes the text box of undefined position, and the character is filled in the text box, described
Device includes：Determining unit 410, interception unit 420, computing unit 430, acquiring unit 440.
Wherein it is determined that unit 410, for determining text box location in images to be recognized；
Interception unit 420, for the position according to the text box, intercepts the images to be recognized Chinese version frame corresponding
Target image；
Computing unit 430, for calculating the target direction histogram of gradients HOG features description corresponding to the target image
Operator；
Acquiring unit 440, for according to the target HOG Feature Descriptors and the good SVMs of training in advance
HOG features are preserved in SVM classifier, the character to be identified of the acquisition target image, the SVM classifier trained to retouch
State the corresponding relation of son and character.
From above technical scheme, the embodiment of the present invention is by the images to be recognized, determining the text box
Position, and according to the text box position, intercept the corresponding target image of the images to be recognized Chinese version frame, realize pair
The positive location of target image and interception, it is to avoid according to default fixed coordinates intercept that object region occurs it is inclined
Difference, improves recognition accuracy.
In an optional example, stating determining unit 410 is included (not shown in Fig. 4)：Binary conversion treatment subelement, sits
Mark determination subelement.
Wherein, binary conversion treatment subelement, for the images to be recognized to be done into binary conversion treatment；
Coordinate determination subelement, the position for determining the text box in the images to be recognized after the binaryzation；
The interception unit 420, is additionally operable to the position according to the text box, the images to be recognized after the binaryzation
In, intercept the corresponding target image of the images to be recognized Chinese version frame.
In another optional example, the coordinate determination subelement is additionally operable to：
When finding the different text box of multiple areas, the position of the minimum text box of Retention area.
In another optional example, described device also includes (not shown in Fig. 4)：Normalization unit, screening unit,
Delete unit.
Wherein, normalization unit, for after interception target image, all target images to be normalized into same pixel chi
Very little target image；
Screening unit, in the target image after any normalization, filtering out longitudinal coordinate maximum and gray scale
It is worth the first minimum pixel, and the second pixel that longitudinal coordinate is minimum and gray value is minimum, and calculates first picture
The longitudinal coordinate of vegetarian refreshments and second pixel is poor；
It is additionally operable to, in the target image after any normalization, filters out lateral coordinates maximum and gray value is minimum
The 3rd pixel, and the 4th pixel that lateral coordinates are minimum and gray value is minimum, and calculate the 3rd pixel with
The lateral coordinates of 4th pixel are poor；
Delete unit, for when longitudinal coordinate difference be less than default minimum height values or the longitudinal coordinate it is poor with it is described
When the product of lateral coordinates difference is less than default minimal face product value, the target image after any normalization is deleted.
In another optional example, the binary conversion treatment subelement, including：Global binary conversion treatment subelement,
Self-adaption binaryzation handles subelement.
Global binary conversion treatment subelement, for the images to be recognized to be done into global binary conversion treatment；
Self-adaption binaryzation handles subelement, is handled for the images to be recognized to be done into local self-adaption binaryzation.
From above technical scheme, on the one hand, the embodiment of the present invention is by the images to be recognized, it is determined that described
The position of text box, and according to the text box position, the corresponding target image of the images to be recognized Chinese version frame is intercepted, it is real
The positive location to target image and interception are showed, it is to avoid occurred according to default fixed coordinates interception object region
Deviation, improve recognition accuracy.On the other hand, the embodiment of the present invention determines the area of text box, screening by statistics
Go out the minimum text box of area, realize the further screening to recognizing text box, improve to the accurate of target image interception
Property, further increase recognition accuracy.Another further aspect, the embodiment of the present invention is after interception target image, by counting target
The height and area of hand-written character in image, realize and further go wrong and screening to target image, further increase knowledge
Other accuracy rate.
The function of unit and the implementation process of effect specifically refer to correspondence step in the above method in said apparatus
Implementation process, will not be repeated here.
For device embodiment, because it corresponds essentially to embodiment of the method, so related part is real referring to method
Apply the part explanation of example.Device embodiment described above is only schematical, wherein described be used as separating component
The unit of explanation can be or may not be physically separate, and the part shown as unit can be or can also
It is not physical location, you can with positioned at a place, or can also be distributed on multiple NEs.Can be according to reality
Selection some or all of module therein is needed to realize the purpose of application scheme.Those of ordinary skill in the art are not paying
In the case of going out creative work, you can to understand and implement.
The preferred embodiment of the application is the foregoing is only, not to limit the application, all essences in the application
God is with principle, and any modification, equivalent substitution and improvements done etc. should be included within the scope of the application protection.
Claims (10)
1. a kind of method of character recognition, it is characterised in that methods described is applied to the identification to character in image, described image
Include the text box of undefined position, the character is filled in the text box, and methods described includes：
Determine text box location in images to be recognized；
According to the position of the text box, the corresponding target image of the images to be recognized Chinese version frame is intercepted；
Calculate the target direction histogram of gradients HOG features corresponding to the target image and describe operator；
According to the good support vector machines grader of the target HOG Feature Descriptors and training in advance, the target is obtained
The corresponding relation of HOG Feature Descriptors and character is preserved in the character to be identified of image, the SVM classifier trained.
2. according to the method described in claim 1, it is characterised in that the position for determining that text box is residing in images to be recognized
Put, including：
The images to be recognized is done into binary conversion treatment；
The position of the text box is determined in images to be recognized after the binaryzation；
The position according to the text box, intercepts the corresponding target image of the images to be recognized Chinese version frame, including：
According to the position of the text box, in the images to be recognized after the binaryzation, the images to be recognized Chinese is intercepted
The corresponding target image of this frame.
3. method according to claim 2, it is characterised in that also include：
When finding the different text box of multiple areas, the position of the minimum text box of Retention area.
4. method according to claim 2, it is characterised in that also include：
After interception target image, all target images are normalized to the target image of same Pixel Dimensions；
In target image after any normalization, the first pixel that longitudinal coordinate is maximum and gray value is minimum is filtered out
Point, and the second pixel that longitudinal coordinate is minimum and gray value is minimum, and calculate first pixel and second picture
The longitudinal coordinate of vegetarian refreshments is poor；
In target image after any normalization, the 3rd pixel that lateral coordinates are maximum and gray value is minimum is filtered out
Point, and the 4th pixel that lateral coordinates are minimum and gray value is minimum, and calculate the 3rd pixel and the 4th picture
The lateral coordinates of vegetarian refreshments are poor；
If longitudinal coordinate difference be less than default minimum height values or the longitudinal coordinate it is poor with the lateral coordinates are poor multiplies
Product is less than default minimal face product value, then deletes the target image after any normalization.
5. method according to claim 2, it is characterised in that described that the images to be recognized is done into binary conversion treatment, bag
Include：
The images to be recognized is done into global binary conversion treatment, or
The images to be recognized is done into local self-adaption binaryzation processing.
6. a kind of device of character recognition, it is characterised in that described device is applied to the identification to character in image, described image
Include the text box of undefined position, the character is filled in the text box, and described device includes：
Determining unit, for determining text box location in images to be recognized；
Interception unit, for the position according to the text box, intercepts the corresponding target figure of the images to be recognized Chinese version frame
Picture；
Computing unit, operator is described for calculating the target direction histogram of gradients HOG features corresponding to the target image；
Acquiring unit, for being classified according to the good support vector machines of the target HOG Feature Descriptors and training in advance
Device, obtain preserved in the character to be identified of the target image, the SVM classifier trained HOG Feature Descriptors with
The corresponding relation of character.
7. device according to claim 6, it is characterised in that the determining unit, including：
Binary conversion treatment subelement, for the images to be recognized to be done into binary conversion treatment；
Coordinate determination subelement, the position for determining the text box in the images to be recognized after the binaryzation；
The interception unit, is additionally operable to the position according to the text box, in the images to be recognized after the binaryzation, interception
The corresponding target image of the images to be recognized Chinese version frame.
8. device according to claim 7, it is characterised in that the coordinate determination subelement, is additionally operable to：
When finding the different text box of multiple areas, the position of the minimum text box of Retention area.
9. device according to claim 7, it is characterised in that also include：
Normalization unit, for after interception target image, all target images to be normalized to the target of same Pixel Dimensions
Image；
Screening unit, in the target image after any normalization, filtering out longitudinal coordinate maximum and gray value most
The first small pixel, and the second pixel that longitudinal coordinate is minimum and gray value is minimum, and calculate first pixel
Longitudinal coordinate with second pixel is poor；
It is additionally operable to, in the target image after any normalization, filters out that lateral coordinates are maximum and gray value is minimum the
Three pixels, and the 4th pixel that lateral coordinates are minimum and gray value is minimum, and calculate the 3rd pixel with it is described
The lateral coordinates of 4th pixel are poor；
Unit is deleted, for when longitudinal coordinate difference is poor less than default minimum height values or the longitudinal coordinate and the transverse direction
When the product of coordinate difference is less than default minimal face product value, the target image after any normalization is deleted.
10. device according to claim 7, it is characterised in that the binary conversion treatment subelement, including：
Global binary conversion treatment subelement, for the images to be recognized to be done into global binary conversion treatment；
Self-adaption binaryzation handles subelement, is handled for the images to be recognized to be done into local self-adaption binaryzation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710193761.7A CN107016417A (en) | 2017-03-28 | 2017-03-28 | A kind of method and device of character recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710193761.7A CN107016417A (en) | 2017-03-28 | 2017-03-28 | A kind of method and device of character recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107016417A true CN107016417A (en) | 2017-08-04 |
Family
ID=59445919
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710193761.7A Pending CN107016417A (en) | 2017-03-28 | 2017-03-28 | A kind of method and device of character recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107016417A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108335199A (en) * | 2018-03-13 | 2018-07-27 | 平安普惠企业管理有限公司 | Loan requests processing method, device, computer equipment and storage medium |
| CN108509919A (en) * | 2018-04-03 | 2018-09-07 | 哈尔滨哈船智控科技有限责任公司 | A kind of detection and recognition methods based on deep learning to waterline in video or picture |
| CN109344815A (en) * | 2018-12-13 | 2019-02-15 | 深源恒际科技有限公司 | A kind of file and picture classification method |
| CN109492562A (en) * | 2018-10-30 | 2019-03-19 | 河南科技大学 | A kind of contactless domestic animal identifying system |
| CN109993040A (en) * | 2018-01-03 | 2019-07-09 | 北京世纪好未来教育科技有限公司 | Text recognition method and device |
| CN110298353A (en) * | 2019-07-03 | 2019-10-01 | 中国农业银行股份有限公司 | A kind of character identifying method and system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101038626A (en) * | 2007-04-25 | 2007-09-19 | 上海大学 | Method and device for recognizing test paper score |
| CN103500216A (en) * | 2013-09-30 | 2014-01-08 | 北京林业大学 | Method for extracting file information |
| JP2015032088A (en) * | 2013-08-01 | 2015-02-16 | 株式会社デンソー | Character recognition device and character recognition method |
| CN104809461A (en) * | 2015-05-08 | 2015-07-29 | 内蒙古科技大学 | License plate recognition method and system combining sequence image super-resolution reconstruction |
| CN105373978A (en) * | 2015-08-12 | 2016-03-02 | 高学 | Artificial test paper judgment processing device and artificial test paper judgment processing method based on OCR |
| CN105373794A (en) * | 2015-12-14 | 2016-03-02 | 河北工业大学 | Vehicle license plate recognition method |
| CN106355176A (en) * | 2016-11-11 | 2017-01-25 | 赵耕伯 | Test question checking system |
-
2017
- 2017-03-28 CN CN201710193761.7A patent/CN107016417A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101038626A (en) * | 2007-04-25 | 2007-09-19 | 上海大学 | Method and device for recognizing test paper score |
| JP2015032088A (en) * | 2013-08-01 | 2015-02-16 | 株式会社デンソー | Character recognition device and character recognition method |
| CN103500216A (en) * | 2013-09-30 | 2014-01-08 | 北京林业大学 | Method for extracting file information |
| CN104809461A (en) * | 2015-05-08 | 2015-07-29 | 内蒙古科技大学 | License plate recognition method and system combining sequence image super-resolution reconstruction |
| CN105373978A (en) * | 2015-08-12 | 2016-03-02 | 高学 | Artificial test paper judgment processing device and artificial test paper judgment processing method based on OCR |
| CN105373794A (en) * | 2015-12-14 | 2016-03-02 | 河北工业大学 | Vehicle license plate recognition method |
| CN106355176A (en) * | 2016-11-11 | 2017-01-25 | 赵耕伯 | Test question checking system |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109993040A (en) * | 2018-01-03 | 2019-07-09 | 北京世纪好未来教育科技有限公司 | Text recognition method and device |
| CN109993040B (en) * | 2018-01-03 | 2021-07-30 | 北京世纪好未来教育科技有限公司 | Text recognition method and device |
| CN108335199A (en) * | 2018-03-13 | 2018-07-27 | 平安普惠企业管理有限公司 | Loan requests processing method, device, computer equipment and storage medium |
| CN108509919A (en) * | 2018-04-03 | 2018-09-07 | 哈尔滨哈船智控科技有限责任公司 | A kind of detection and recognition methods based on deep learning to waterline in video or picture |
| CN108509919B (en) * | 2018-04-03 | 2022-04-29 | 哈尔滨哈船智控科技有限责任公司 | Method for detecting and identifying waterline in video or picture based on deep learning |
| CN109492562A (en) * | 2018-10-30 | 2019-03-19 | 河南科技大学 | A kind of contactless domestic animal identifying system |
| CN109492562B (en) * | 2018-10-30 | 2021-05-25 | 河南科技大学 | Non-contact livestock identification system |
| CN109344815A (en) * | 2018-12-13 | 2019-02-15 | 深源恒际科技有限公司 | A kind of file and picture classification method |
| CN109344815B (en) * | 2018-12-13 | 2021-08-13 | 深源恒际科技有限公司 | Document image classification method |
| CN110298353A (en) * | 2019-07-03 | 2019-10-01 | 中国农业银行股份有限公司 | A kind of character identifying method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107016417A (en) | A kind of method and device of character recognition | |
| CN110569721A (en) | Recognition model training method, image recognition method, device, equipment and medium | |
| CA3100642A1 (en) | Multi-sample whole slide image processing in digital pathology via multi-resolution registration and machine learning | |
| CN104484643A (en) | Intelligent identification method and system for hand-written table | |
| CN103824373B (en) | A kind of bill images amount of money sorting technique and system | |
| CN108197644A (en) | A kind of image-recognizing method and device | |
| US9785850B2 (en) | Real time object measurement | |
| CN105868708A (en) | Image object identifying method and apparatus | |
| JP7026165B2 (en) | Text recognition method and text recognition device, electronic equipment, storage medium | |
| GB2517674A (en) | Image capture using client device | |
| CN105160343A (en) | Information identification method and device applied to film on-demand-printing system | |
| CN105869122A (en) | Image processing method and apparatus | |
| CN108762740B (en) | Page data generation method and device and electronic equipment | |
| CN106331746A (en) | Method and device for identifying position of watermark in video file | |
| CN106156691A (en) | The processing method of complex background image and device thereof | |
| CN107657251A (en) | Determine the device and method of identity document display surface, image-recognizing method | |
| CN109635729B (en) | Form identification method and terminal | |
| CN107145888A (en) | Video caption real time translating method | |
| EP3798906A1 (en) | System and method for masking text within images | |
| CN111178192B (en) | Method and device for identifying position of target object in image | |
| Ganchimeg | History document image background noise and removal methods | |
| US20170352170A1 (en) | Nearsighted camera object detection | |
| CN106683257A (en) | Serial number location method and device | |
| US20030210818A1 (en) | Knowledge-based hierarchical method for detecting regions of interest | |
| CN113450355A (en) | Method for extracting image features based on multi-membrane CT image and 3DCNN network |
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: 20170804 |