CN110503101A - Font evaluation method, device, equipment and computer readable storage medium - Google Patents
Font evaluation method, device, equipment and computer readable storage medium Download PDFInfo
- Publication number
- CN110503101A CN110503101A CN201910782326.7A CN201910782326A CN110503101A CN 110503101 A CN110503101 A CN 110503101A CN 201910782326 A CN201910782326 A CN 201910782326A CN 110503101 A CN110503101 A CN 110503101A
- Authority
- CN
- China
- Prior art keywords
- image
- evaluation
- feature
- training
- writing words
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/28—Character recognition specially adapted to the type of the alphabet, e.g. Latin alphabet
- G06V30/287—Character recognition specially adapted to the type of the alphabet, e.g. Latin alphabet of Kanji, Hiragana or Katakana characters
Landscapes
- Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Controls And Circuits For Display Device (AREA)
Abstract
The disclosure provides a kind of font evaluation method, device, equipment and computer readable storage medium, implementation: receive include writing words original image；The original image is pre-processed to obtain processing image；The writing words for including in the processing image are split, at least one partitioning portion is obtained；The corresponding feature of the partitioning portion is extracted, the evaluation information of the writing words is determined according to the feature, and feeds back the evaluation information.Method, apparatus, equipment and the computer readable storage medium that the disclosure provides, it can be by analyzing the original image for including writing words, determine the evaluation information of text, the problem of writing is determined by way of comparing copybook without user, the method that the disclosure provides can be accurately to the problems in user feedback its writing words, and then more accurately user is instructed to practice calligraphy.
Description
Technical field
This disclosure relates to image recognition technology more particularly to a kind of font evaluation method, device, equipment and computer-readable
Storage medium.
Background technique
In life, it is often necessary to be write using soft pen.For example, can be practiced calligraphy using writing brush.User makes
When carrying out writing practising with soft pen, the text of the writing to oneself is needed to be corrected.
In the prior art, when user is write using the soft pen such as writing brush, oneself observation copybook and oneself can only be passed through
The difference between text that oneself writes, come determine the text of writing there are the problem of, the mode of oneself this artificial comparison, analysis
As a result it is inaccurate.
Therefore, a kind of method for capableing of the soft pen writing font of automatic Evaluation is needed, in order to for can accurately determine
The soft pen calligraphy of writing there are the problem of.
Summary of the invention
The disclosure provides a kind of font evaluation method, device, equipment and computer readable storage medium, to solve existing skill
It is compared existing for the problems in font by copybook than inaccurate technical problem in art
The first aspect of the disclosure is to provide a kind of font evaluation method, comprising:
Receive the original image including writing words；
The original image is pre-processed to obtain processing image；
The writing words for including in the processing image are split, at least one partitioning portion is obtained；
The feature that the partitioning portion includes is extracted, the evaluation information of the writing words is determined according to the feature, and
Feed back the evaluation information.
Another aspect of the disclosure is to provide a kind of font evaluating apparatus, comprising:
Receiving module, for receiving the original image including writing words；
Preprocessing module obtains processing image for being pre-processed to the original image；
Divide module, for being split to the writing words for including in the processing image, obtains at least one
Partitioning portion；
Extraction module, the feature for including for extracting the partitioning portion；
Evaluation module for determining the evaluation information of the writing words according to the feature, and feeds back the evaluation letter
Breath.
The another aspect of the disclosure is to provide a kind of font valuator device, comprising:
Memory；
Processor；And
Computer program；
Wherein, the computer program stores in the memory, and is configured to be executed by the processor to realize
Font evaluation method as described in above-mentioned first aspect.
The another aspect of the disclosure is to provide a kind of computer readable storage medium, is stored thereon with computer program,
The computer program is executed by processor to realize the font evaluation method as described in above-mentioned first aspect.
Font evaluation method, device, equipment and the computer readable storage medium that the disclosure provides have the technical effect that
Font evaluation method, device, equipment and the computer readable storage medium that the disclosure provides, comprising: reception includes
The original image of writing words；The original image is pre-processed to obtain processing image；To including in the processing image
The writing words be split, obtain at least one partitioning portion；The corresponding feature of the partitioning portion is extracted, according to institute
It states feature and determines the evaluation information of the writing words, and feed back the evaluation information.The method, apparatus of disclosure offer is set
Standby and computer readable storage medium can determine text by analyzing the original image for including writing words
Evaluation information determines the problem of writing, the side that the disclosure provides without user by way of comparing copybook
Method can be accurately to the problems in user feedback its writing words, and then more accurately user is instructed to practice calligraphy.
Detailed description of the invention
Fig. 1 is the system architecture diagram shown in an exemplary embodiment of the invention；
Fig. 2 is the flow chart of the font evaluation method shown in an exemplary embodiment of the invention；
Fig. 2A is the font segmentation schematic diagram shown in an exemplary embodiment of the invention；
Fig. 3 is the flow chart of the font evaluation method shown in another exemplary embodiment of the present invention；
Fig. 4 is the structure chart of the font evaluating apparatus shown in an exemplary embodiment of the invention；
Fig. 5 is the structure chart of the font evaluating apparatus shown in another exemplary embodiment of the present invention；
Fig. 6 is the structure chart of the font valuator device shown in an exemplary embodiment of the invention.
Specific embodiment
Fig. 1 is the system architecture diagram shown in an exemplary embodiment of the invention.
As shown in Figure 1, system architecture of the present invention includes terminal 11, server 12.Terminal for example can be specially
For the equipment of writing, the equipment with camera function, such as mobile phone etc. can also be.
Server 12 can be distributed server, Cloud Server etc., and the present embodiment is limited not to this.
User can be with writing words, and are taken pictures by terminal 11 to text, on the image that terminal 11 can will acquire
Server 12 is reached, so that server 12 exports the evaluation information of writing words to terminal 11.
Wherein, if terminal 11 is the equipment such as mobile phone, scheme provided by the embodiments of the present application be can be set in mobile phone.If
Terminal 11 is used exclusively for the equipment write, then camera has can be set in the equipment, for acquiring the text figure of user's writing
Picture, the equipment can also have network savvy, so that the image of acquisition is uploaded to server 12.
The scheme of the application can be used for evaluating the soft word that user writes, such as Brush calligraphy.
Fig. 2 is the flow chart of the font evaluation method shown in an exemplary embodiment of the invention.
As shown in Fig. 2, font evaluation method provided in this embodiment includes:
Step 201, the original image including writing words is received.
Wherein, user can take pictures to text by terminal after the completion of writing, obtain original image.
It takes pictures for example, mobile phone can be used in user to text, and the photo is passed through to the client being arranged in terminal
End is uploaded to server.The server can be the background server of the client.
For another example the equipment dedicated for writing can have a writing region, can place in the region for writing
Paper, user can write on it.The equipment can also be arranged in the positions such as the top of paper or oblique upper to be taken the photograph
As head.User after finishing writing, can press the button in equipment, trigger the camera in equipment and acquire original image, and by its
Feed back to server.
The setting position of camera can be adjusted according to actual needs, which was photographed and writes area
Domain.
After original image is uploaded to server by terminal, server can receive the original image, and be based on original graph
As to including writing words evaluate.
Step 202, the original image is pre-processed to obtain processing image.
It is generally color image by the image that terminal acquires, therefore, which can be pre-processed, in order to
The more accurate font information into image, and then more accurately text can be evaluated.
Wherein it is possible to carry out binary conversion treatment to original image, color image is adjusted to black white image.It can specifically incite somebody to action
The gray value of pixel on image is set as 0 or 255, that is, whole image is showed apparent black and white effect.
Specifically, noise reduction operation can also be carried out to the image after binaryzation, to reduce content outside font to font
Evaluate bring interference.For example, there may be impurity on paper, the position of paper is caused to be arranged to black, and font covers
Cover area is also set to black, then being easy when evaluating font using impurity position as a part of font.
Step 203, the writing words for including in the processing image are split, obtain at least one cutting part
Point.
Further, a parted pattern can be trained in advance, for dividing writing words.
When practical application, some texts can be preset, and by way of manually marking, mark out these texts
Segmentation result.Mould can be divided according to the corresponding bianry image of these texts, including the bianry image of segmentation result, trained one
Type.
For example, the image of a large amount of texts soft font of corresponding standard can be acquired, which can be obtained by pretreatment
As corresponding bianry image.It can also mark out by way of manually marking and need divided part in bianry image, into
And obtain include segmentation result bianry image, that is, one group of not divided character image and these texts are divided
Character image, neural metwork training can be carried out by this group of data, and then obtain parted pattern.
Processing image can be inputted into parted pattern, parted pattern is enabled to export segmentation result.
Wherein, in the training process, interactive process can also be added, for example, can export based on current training result
The segmentation result of text, user can modify to segmentation result, and according to modified segmentation result continue to model into
Row training.
Optionally, Text segmentation rule can be preset, and based on the writing text in these regular dividing processing images
Word.
Wherein, different texts can have different segmentation rules, a database can be preset, wherein being stored with
The corresponding segmentation rule of different literals.
In one embodiment, server can identify processing image, determine the text that user writes, and obtain
The corresponding segmentation rule of the text is taken, then the text in processing image is split based on segmentation rule.
In another embodiment, server first can also send a practice text to terminal, be somebody's turn to do so that user writes
Practice text.After passing original image at the terminal, the corresponding segmentation rule of the available practice text of server, then be based on dividing
Rule is cut to be split the text in processing image.
Wherein, by being split to writing words, at least one partitioning portion can be obtained.For example, user carries out pen
Practice is drawn, a point has been write, when being split to the point, its entirety may be used as to a partitioning portion.For another example user
A complete text has been write, multiple partitioning portions may be divided at this time.
Fig. 2A is the font segmentation schematic diagram shown in an exemplary embodiment of the invention.
As shown in Figure 2 A, Chinese character " forever " can be partitioned into 10 parts.
Step 204, the corresponding feature of the partitioning portion is extracted, the evaluation of the writing words is determined according to the feature
Information, and feed back the evaluation information.
Specifically, be directed to each partitioning portion, server can extract including feature.For example, can instruct in advance
Practice a Feature Selection Model, for extracting the feature in each partitioning portion.
Further, the corresponding bianry image of writing words, the bianry image including segmentation result can be acquired in advance, led to
The feature in training pattern extraction image is crossed, and identifies and wherein distinguishes biggish feature, using these features as to be determined
Feature.For example, the text of student's writing can be acquired, as writing words, and the bianry image of these texts is obtained.Packet
The bianry image for including segmentation result for example can be the corresponding image of standard glyph.
When practical application, by comparing the difference of writing words and grapholect, it can recognize that writing words and mark
Difference characteristic between font, and these features are to need the feature of key evaluation.It can be corresponding more by a text
A writing words bianry image training pattern, so that the model determines the feature for needing to extract in each partitioning portion of the text.
Bianry image including segmentation result can be a complete bi-level fonts figure, wherein being labelled with each departmentalism
Point.
Wherein it is possible to extract high dimensional feature to the component diagram of text bianry image and mark, analyze in data variation compared with
Big feature.According to preset characteristic dimension, dimensionality reduction behaviour is carried out to high dimensional feature in the way of similar principal component analysis
Make, obtains the suitable low-dimensional feature of dimension scale, the dimension of low-dimensional feature is determined according to text specifying information.
Wherein, server can evaluate entire writing words according to the feature of extraction, to determine evaluation information.
Specifically, server can be according to each partitioning portion of characteristic evaluating of extraction, and commenting each partitioning portion
Valence result is combined, and generates evaluation information.Can also to terminal Feedback Evaluation information, enable the terminal to text, image,
The various ways such as voice show evaluation information to user.
Further, an evaluation model can be trained in advance, for the characteristic evaluating writing words according to extraction.For example,
A large amount of writing words, such as the soft pen calligraphy that acquisition student writes can be acquired, and these texts are commented by expert
Valence obtains the corresponding evaluation result of writing words.And in-service evaluation result marks writing words, obtains the data of training, makes
With these data evaluation model can be obtained with training pattern.
When practical application, it is also based on the feature that above-mentioned steps extract each writing words, and by the corresponding spy of text
Levy the training data of vector, evaluation result as training pattern.
Wherein, multiple indexs be may include in evaluation result, for example, each stroke in a text may have phase
The index answered can also be by adjusting the side of feature in order to the relationship between the variation and evaluation index of each feature of determination
Formula generates new text, then acquires the evaluation result of these texts.For example, the thinner of the horizontal adjustment of text can be obtained
New text, and by the expert opinion new text, obtain evaluation result.
Specifically, the text after feature and its corresponding evaluation result can will be adjusted in training as training data,
Training pattern is inputted, so that model can determine that each feature is corresponding and comment according to the feature difference between text, evaluation difference
Valence index.Such as horizontal width is when being n, the evaluation index of text be it is horizontal very well, when horizontal width is n+3, the evaluation of text refers to
It is designated as the width criteria horizontal too wide, then that training pattern can be horizontal based on the determination of these differences.
Method provided in this embodiment is for assessing font, and specifically for assessing the font of soft pen calligraphy, this method is by setting
The equipment for being equipped with method provided in this embodiment executes, which realizes usually in a manner of hardware and/or software.
Font evaluation method provided in this embodiment, comprising: receive the original image including writing words；To described original
Image is pre-processed to obtain processing image；To it is described processing image in include the writing words be split, obtain to
A few partitioning portion；The corresponding feature of the partitioning portion is extracted, the evaluation of the writing words is determined according to the feature
Information, and feed back the evaluation information.Method provided in this embodiment, can by include writing words original image into
Row analysis, determines the evaluation information of text, is determined by way of comparing copybook present in writing without user
Problem, method provided in this embodiment can be accurately to the problems in user feedback its writing words, and then more accurately refers to
User is led to practice calligraphy.
Fig. 3 is the flow chart of the font evaluation method shown in another exemplary embodiment of the present invention.
As shown in figure 3, font evaluation method provided in this embodiment, comprising:
Step 301, the original image including writing words is received.
Step 301 is similar with the concrete principle of step 201 and implementation, and details are not described herein again.
Step 302, binaryzation, denoising are carried out to the original image, obtains the processing image.
Wherein it is possible to carry out binary conversion treatment to original image, color image is adjusted to black white image.It can specifically incite somebody to action
The gray value of pixel on image is set as 0 or 255, that is, whole image is showed apparent black and white effect.
Specifically, noise reduction operation can also be carried out to the image after binaryzation, to reduce content outside font to font
Evaluate bring interference.For example, there may be impurity on paper, the position of paper is caused to be arranged to black, and font covers
Cover area is also set to black, then being easy when evaluating font using impurity position as a part of font.
Step 303, the writing words for including in the processing image are carried out by parted pattern trained in advance
Segmentation.
It further, can be previously according to training image, segmented image training parted pattern, for the processing image
In include the writing words be split.
When practical application, some character images, such as the character image of standard glyph can be acquired, and two-value is carried out to it
Change, the processing of noise reduction obtains training image.Wherein it is possible to mark each segmentation in training image by way of manually marking
Part obtains segmented image.
Wherein it is possible to training image, segmented image input training pattern are trained, so that the model can adjust
Internal data can handle these training images, obtain corresponding segmented image.
Specifically, since Chinese character quantity is more, and structure is complicated, if training image and its corresponding segmented image quantity
It is few, obtained parted pattern may be trained not accurate enough, therefore, in method provided in this embodiment, artificial school can also be added
Pair process, make parted pattern more accurate by manual intervention.
Further, a training image can be inputted into training pattern, it is made to export segmented image, user can be actively right
Segmented image is adjusted, and complies with the needs of calligraphy teaching.It can be continued according to modified segmented image to model
It is trained, keeps it more accurate.
When practical application, if training parted pattern uses the corresponding training image of standard glyph and segmented image,
Directly by the corresponding binary image input model of writing words, the segmentation result that may be exported is not very accurately, therefore, originally
Embodiment provide method can also include:
Practice text is determined according to the writing words, so that the parted pattern is according to the practice text to the book
Text is write to be split.
Wherein, parted pattern can determine the partitioning scheme of the text according to practice text, and based on partitioning scheme to book
Text is write to be split.Such as the text of user's input is " eight ", if the opening of the word is smaller, parted pattern may will be opened
It is split as a whole at mouthful.
Specifically, determining that its corresponding practice text, parted pattern can determine this if identify to text in advance
Practicing text is " eight ", and determines " eight " corresponding partitioning scheme, is then split aperture position respectively.
Further, before user's writing words, practice text can also be sent from server to terminal, and then specified
User needs the text write.In this case, parted pattern can directly divide writing words according to the practice text
It cuts.
When practical application, the partitioning portion includes following any information:
The combination of single stroke, some stroke, at least two strokes.
Single stroke refers to a complete stroke, such as cross, point, perpendicular etc., and some stroke refers to a complete stroke
A part, such as the perpendicular turn part hooked can be used as a partitioning portion.The combination of two strokes for example can be two points
Combination.
Wherein, when dividing text, a stroke can be used as an individual partitioning portion, a part of the stroke
It can be used as individual partitioning portion, which can also be with other stroke combinations as a partitioning portion.There may be it
A kind of middle mode, there may also be various ways.
Step 304, the feature that the partitioning portion includes is extracted by Feature Selection Model trained in advance.
Specifically, training image can be acquired in advance, which for example can be the character image of user's writing, tool
Body can be binary image.Such as it can be taken pictures, be pre-processed to these texts, be obtained after student has write text
Training image.
The corresponding component mark figure of text can also be acquired, which can be each point marked in the literature
The figure of part is cut, such as the segmentation figure of the parted pattern output completed based on training being split to text；It can also be with
It is the figure for marking the contents such as stroke, stroke combination in the literature according to demand.Component mark figure is also possible to binary image.It should
Component mark legend such as can be the image by being labeled to standard glyph.
Training image, component mark figure input training pattern can be trained into completion to be trained to the model
Model is Feature Selection Model.
Wherein, training pattern can extract training image, the high dimensional feature in component mark figure, and the feature both compared
Difference.The feature of same position in font can specifically be extracted.For example, training pattern can extract in training image a slash
Width characteristics, extracting parts mark the width characteristics of the high dimensional feature slash in figure, then the two are compared.If the two is poor
It is different larger, it may be considered that this feature is the feature for needing emphasis to assess in word evaluation.Default difference can be set, if special
Difference is greater than the default difference between sign, then it is assumed that the two differs greatly.For example, can calculate the Euclideans of two feature vectors away from
From, and according to distance and the comparison result of distance threshold, determine difference degree.
Specifically, different training images can be inputted for the same text, such as the text figure that different user is write
Picture, and then training pattern study can be made to the difference characteristic of the text and standard glyph.
Further, however, it is determined that characteristic dimension it is higher, can also be utilized similar main according to preset characteristic dimension
The mode of constituent analysis carries out dimensionality reduction operation to high dimensional feature, obtains the suitable low-dimensional feature of dimension scale, specifically can basis
Text specifying information determines the dimension of low-dimensional feature, calculation are as follows: dimension=stroke quantity × 4.
Step 305, evaluation index corresponding with feature is obtained, and determines that partitioning portion is corresponding according to characteristic evaluating index
Evaluation result.
Wherein it is possible to pass through the evaluation previously according to training image and its corresponding evaluation data evaluation of training model
Model determines the corresponding evaluation result of the partition member according to the feature.
Specifically, training image can be acquired in advance, which is, for example, that the text write to user is taken pictures
And the image obtained by pretreatment, such as the image after binaryzation.Training image can also be handled, such as based on upper
It states parted pattern, Feature Selection Model that training is completed to handle training image, to extract each cutting part subpackage in image
The feature included, then using the corresponding evaluation data of these features and image as training data, evaluation of training model.
Further, the evaluation data of the feature and training image of training image or extraction can be inputted to training mould
Type is trained.
When practical application, can also adjust it is same practice text in include feature obtain it is corresponding with the practice text
Different training images；According to the feature difference between the training image, the difference between training image evaluation data, training
The evaluation model.
Wherein it is possible to be adjusted to the writing font of same practice text, for example, on the basis of a standard glyph
It is adjusted, changes the feature in image, and then change the font in image.It for example by point " forever " and lower partial distance is one
A feature, then the adjustable distance is to obtain different fonts.The edge radian of point can also be used as a feature, can adjust
The edge radian of integral point, the different corresponding fonts of radian are also different.
Specifically, can by adjusting extraction feature to change font, these boundaries for going to equipment can also be exported
Different fonts are evaluated by expert in face, to obtain the evaluation data of the font after eigentransformation.
It further, can be using the font after adjustment feature as training image, by the corresponding evaluation data of each font
As another training data.The feature in training image, these features can be extracted by parted pattern, Feature Selection Model
It can be entered in current training pattern (the not evaluation model that training finishes), which can be according between training image
Difference between feature difference, training image evaluation data, determines the corresponding evaluation index of each feature.For example, when " forever "
Point and lower partial distance when being greater than n distance farther out, be closer when the point of " forever " and lower partial distance is less than m.
When practical application, during evaluating text, Feature Selection Model can be to evaluation model output character
In characteristic, such as can be the data of vector form.Trained evaluation model can using these characteristics as
Input, and evaluation index corresponding with these features is obtained, and the corresponding evaluation of partition member is determined according to characteristic evaluating index
As a result.For example, point is excessive with lower partial distance when input text is " forever ", then the radian of such as point is excessively gentle.
Step 306, evaluation information is determined according to evaluation result.
Wherein it is possible to determine the overall evaluation information of text according to the evaluation result of each partitioning portion.It can integrate point
The evaluation result for cutting component, obtains evaluation information, and it is poor pre- that evaluation can also be filtered out in the evaluation result of partition member
If quantity is as a result, as evaluation information.
For example, writing words are divided into 7 parts, wherein there is the evaluation information of 5 parts poor, can therefrom select to comment
3 worst parts of valence information are as evaluation information.Specifically semantics recognition can be carried out according to the text in evaluation result, with true
Determine the positive negative emotions of evaluation result.
Specifically, can be identified in original image according to evaluation information, and feed back to terminal so that terminal to
Family shows the image for having evaluation information.For example, the partitioning portion circle being evaluated can be elected, and mark corresponds to wherein
Evaluation result.
Fig. 4 is the structure chart of the font evaluating apparatus shown in an exemplary embodiment of the invention.
As shown in figure 4, font evaluating apparatus provided in this embodiment, comprising:
Receiving module 41, for receiving the original image including writing words；
Preprocessing module 42 obtains processing image for being pre-processed to the original image；
Divide module 43, for being split to the writing words for including in the processing image, obtains at least one
A partitioning portion；
Extraction module 44, the feature for including for extracting the partitioning portion；
Evaluation module 45 for determining the evaluation information of the writing words according to the feature, and feeds back the evaluation
Information.
Font evaluating apparatus provided in this embodiment, including receiving module, for receiving the original graph including writing words
Picture；Preprocessing module obtains processing image for being pre-processed to original image；Divide module, for in processing image
Including writing words be split, obtain at least one partitioning portion；Extraction module, the spy for including for extracting partitioning portion
Sign；Evaluation module, for determining the evaluation information of writing words, and Feedback Evaluation information according to feature.It is provided in this embodiment
Device, can by include writing words original image analyze, determine the evaluation information of text, without
User determines the problem of writing by way of comparing copybook, and device provided in this embodiment can be accurately to user
The problems in its writing words are fed back, and then more accurately user are instructed to practice calligraphy.
The concrete principle and implementation of font evaluating apparatus provided in this embodiment with embodiment class shown in Fig. 2
Seemingly, details are not described herein again.
Fig. 5 is the structure chart of the font evaluating apparatus shown in another exemplary embodiment of the present invention.
As shown in figure 4, on the basis of the above embodiments, device provided in this embodiment, optionally, the pretreatment mould
Block 42 is specifically used for:
Binaryzation, denoising are carried out to the original image, obtain the processing image.
Optionally, the partitioning portion includes following any information:
The combination of single stroke, some stroke, at least two strokes.
It optionally, further include parted pattern training module 46, for according to training image, segmented image training segmentation mould
Type, the parted pattern are used to be split the writing words for including in the processing image.
Optionally, the segmentation module 43 is specifically used for:
Practice text is determined according to the writing words, so that the parted pattern is according to the practice text to the book
Text is write to be split.
Described device further includes sending module 47, for receiving the original graph including writing words in the receiving module 41
Before picture, practice text is sent to terminal device, so that user writes the writing words according to the practice text；
Optionally, the segmentation module 43 is specifically used for:
The writing words are split according to the practice text by the parted pattern.
It optionally, further include extracting model training module 48, for pre- according to meeting between training image, component mark figure
If the feature training characteristics of difference extract model, the Feature Selection Model is for extracting the feature that the partitioning portion includes.
Optionally, the evaluation module 45 is specifically used for:
Evaluation index corresponding with the feature is obtained, and the partition member pair is determined according to the characteristic evaluating index
The evaluation result answered；
The evaluation information is determined according to the evaluation result.
Optionally, described device further includes evaluation model training module 49, is specifically used for according to training image and its correspondence
Evaluation data evaluation of training model, the evaluation model is used to according to the feature determine the corresponding evaluation of the partition member
As a result.
Optionally, evaluation model training module 49 is specifically used for:
It adjusts the feature for including in same practice text and obtains different training images corresponding from the practice text；
According between the training image feature difference, the training image evaluation data between difference, training described in
Evaluation model.
The concrete principle and implementation of device provided in this embodiment are similar with embodiment shown in Fig. 3, herein not
It repeats again.
Fig. 6 is the structure chart of the font valuator device shown in an exemplary embodiment of the invention.
As shown in fig. 6, font valuator device provided in this embodiment includes:
Memory 61；
Processor 62；And
Computer program；
Wherein, the computer program is stored in the memory 61, and be configured to by the processor 62 execute with
Realize any font evaluation method as described above.
The present embodiment also provides a kind of computer readable storage medium, is stored thereon with computer program,
The computer program is executed by processor to realize any font evaluation method as described above.
The present embodiment also provides a kind of computer program, including program code, when computer runs the computer program
When, said program code executes any font evaluation method as described above.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above-mentioned each method embodiment can lead to
The relevant hardware of program instruction is crossed to complete.Program above-mentioned can be stored in a computer readable storage medium.The journey
When being executed, execution includes the steps that above-mentioned each method embodiment to sequence；And storage medium above-mentioned include: ROM, RAM, magnetic disk or
The various media that can store program code such as person's CD.
Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention., rather than its limitations；To the greatest extent
Pipe present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: its according to
So be possible to modify the technical solutions described in the foregoing embodiments, or to some or all of the technical features into
Row equivalent replacement；And these are modified or replaceed, various embodiments of the present invention technology that it does not separate the essence of the corresponding technical solution
The range of scheme.
Claims (13)
1. a kind of font evaluation method characterized by comprising
Receive the original image including writing words；
The original image is pre-processed to obtain processing image；
The writing words for including in the processing image are split, at least one partitioning portion is obtained；
The feature that the partitioning portion includes is extracted, the evaluation information of the writing words is determined according to the feature, and is fed back
The evaluation information.
2. the method according to claim 1, wherein described pre-processed to the original image is handled
Image, comprising:
Binaryzation, denoising are carried out to the original image, obtain the processing image.
3. the method according to claim 1, wherein the partitioning portion includes following any information:
The combination of single stroke, some stroke, at least two strokes.
4. the method according to claim 1, wherein further include:
According to training image, segmented image training parted pattern, the parted pattern is used for including in the processing image
The writing words are split.
5. according to the method described in claim 4, it is characterized in that, described to the writing text for including in the processing image
Word is split, comprising:
Practice text is determined according to the writing words, so that the parted pattern is according to the practice text to the writing text
Word is split.
6. according to the method described in claim 4, it is characterized in that, it is described receive include writing words original image before,
Further include:
Practice text is sent to terminal device, so that user writes the writing words according to the practice text；
It is described that the writing words for including in the processing image are split, comprising:
The writing words are split according to the practice text by the parted pattern.
7. the method according to claim 1, wherein further include:
Model is extracted according to the feature training characteristics for meeting default difference between training image, component mark figure, the feature mentions
Modulus type is for extracting the feature that the partitioning portion includes.
8. the method according to claim 1, wherein described determine commenting for the writing words according to the feature
Valence information, comprising:
Evaluation index corresponding with the feature is obtained, and determines that the partition member is corresponding according to the characteristic evaluating index
Evaluation result；
The evaluation information is determined according to the evaluation result.
9. according to the method described in claim 8, it is characterized by further comprising:
According to training image and its corresponding evaluation data evaluation of training model, the evaluation model is used for true according to the feature
Determine the corresponding evaluation result of the partition member.
10. according to the method described in claim 9, it is characterized in that, described according to training image and its corresponding evaluation data
Evaluation of training model, comprising:
It adjusts the feature for including in same practice text and obtains different training images corresponding from the practice text；
According to the feature difference between the training image, the difference between training image evaluation data, the training evaluation
Model.
11. a kind of font evaluating apparatus characterized by comprising
Receiving module, for receiving the original image including writing words；
Preprocessing module obtains processing image for being pre-processed to the original image；
Divide module, for being split to the writing words for including in the processing image, obtains at least one segmentation
Part；
Extraction module, the feature for including for extracting the partitioning portion；
Evaluation module for determining the evaluation information of the writing words according to the feature, and feeds back the evaluation information.
12. a kind of font valuator device characterized by comprising
Memory；
Processor；And
Computer program；
Wherein, the computer program stores in the memory, and is configured to be executed by the processor to realize such as power
Benefit requires any method of 1-10.
13. a kind of computer readable storage medium, which is characterized in that it is stored thereon with computer program,
The computer program is executed by processor to realize the method as described in claim 1-10 is any.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782326.7A CN110503101A (en) | 2019-08-23 | 2019-08-23 | Font evaluation method, device, equipment and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910782326.7A CN110503101A (en) | 2019-08-23 | 2019-08-23 | Font evaluation method, device, equipment and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110503101A true CN110503101A (en) | 2019-11-26 |
Family
ID=68589155
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910782326.7A Pending CN110503101A (en) | 2019-08-23 | 2019-08-23 | Font evaluation method, device, equipment and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110503101A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111554149A (en) * | 2020-05-15 | 2020-08-18 | 黑龙江德亚文化传媒有限公司 | System and method for copybook scoring |
| CN112308745A (en) * | 2020-02-28 | 2021-02-02 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN112307866A (en) * | 2020-02-28 | 2021-02-02 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1571979A (en) * | 2001-10-15 | 2005-01-26 | 西尔弗布鲁克研究有限公司 | A method and apparatus for decoding handwritten characters |
| CN102682641A (en) * | 2012-05-10 | 2012-09-19 | 梁洺华 | Method and system for learning and practicing characters on electronic device |
| CN103080878A (en) * | 2010-08-24 | 2013-05-01 | 诺基亚公司 | Method and apparatus for segmenting strokes of overlapped handwriting into one or more groups |
| CN106326907A (en) * | 2015-06-23 | 2017-01-11 | 王东锐 | Handwriting automatic evaluation method and system |
| CN107622710A (en) * | 2017-08-30 | 2018-01-23 | 中南出版传媒集团股份有限公司湖南教育出版社分公司 | Digitize writing practising method and system |
| CN108734165A (en) * | 2018-05-07 | 2018-11-02 | 百度在线网络技术（北京）有限公司 | Calligraphy quality determining method, device and electronic equipment based on artificial intelligence |
| CN108764070A (en) * | 2018-05-11 | 2018-11-06 | 西北大学 | A kind of stroke dividing method and calligraphic copying guidance method based on writing video |
| CN109886180A (en) * | 2013-12-17 | 2019-06-14 | 微软技术许可有限责任公司 | For overlapping the user interface of handwritten text input |
| CN110070089A (en) * | 2019-04-24 | 2019-07-30 | 京东方科技集团股份有限公司 | Calligraphy guidance method and device, computer equipment and medium |
-
2019
- 2019-08-23 CN CN201910782326.7A patent/CN110503101A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1571979A (en) * | 2001-10-15 | 2005-01-26 | 西尔弗布鲁克研究有限公司 | A method and apparatus for decoding handwritten characters |
| CN103080878A (en) * | 2010-08-24 | 2013-05-01 | 诺基亚公司 | Method and apparatus for segmenting strokes of overlapped handwriting into one or more groups |
| CN102682641A (en) * | 2012-05-10 | 2012-09-19 | 梁洺华 | Method and system for learning and practicing characters on electronic device |
| CN109886180A (en) * | 2013-12-17 | 2019-06-14 | 微软技术许可有限责任公司 | For overlapping the user interface of handwritten text input |
| CN106326907A (en) * | 2015-06-23 | 2017-01-11 | 王东锐 | Handwriting automatic evaluation method and system |
| CN107622710A (en) * | 2017-08-30 | 2018-01-23 | 中南出版传媒集团股份有限公司湖南教育出版社分公司 | Digitize writing practising method and system |
| CN108734165A (en) * | 2018-05-07 | 2018-11-02 | 百度在线网络技术（北京）有限公司 | Calligraphy quality determining method, device and electronic equipment based on artificial intelligence |
| CN108764070A (en) * | 2018-05-11 | 2018-11-06 | 西北大学 | A kind of stroke dividing method and calligraphic copying guidance method based on writing video |
| CN110070089A (en) * | 2019-04-24 | 2019-07-30 | 京东方科技集团股份有限公司 | Calligraphy guidance method and device, computer equipment and medium |
Non-Patent Citations (3)
| Title |
|---|
| 孟祥旭: "《智能CAD与数字娱乐新进展》", 30 September 2006, 山东大学出版社 * |
| 张巍: "基于条件生成对抗网络的书法字笔画分割", 《自动化学报》 * |
| 黄襄念: "《文字识别原理与策略》", 30 November 2002, 西南交通大学出版社 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112308745A (en) * | 2020-02-28 | 2021-02-02 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN112307866A (en) * | 2020-02-28 | 2021-02-02 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating information |
| CN111554149A (en) * | 2020-05-15 | 2020-08-18 | 黑龙江德亚文化传媒有限公司 | System and method for copybook scoring |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492643B (en) | Certificate identification method and device based on OCR, computer equipment and storage medium | |
| US10699111B2 (en) | Page segmentation of vector graphics documents | |
| CN109190722B (en) | Font style migration transformation method based on Manchu character picture | |
| CN110032938B (en) | Tibetan recognition method and device and electronic equipment | |
| CN111753767A (en) | Method and device for automatically correcting operation, electronic equipment and storage medium | |
| US8494273B2 (en) | Adaptive optical character recognition on a document with distorted characters | |
| CN110210413A (en) | A kind of multidisciplinary paper content detection based on deep learning and identifying system and method | |
| JP2020527260A (en) | Text detection analysis methods, devices and devices | |
| CN111758116B (en) | Face image recognition system, recognizer generation device, recognition device, and face image recognition system | |
| WO2021042505A1 (en) | Note generation method and apparatus based on character recognition technology, and computer device | |
| CN110503101A (en) | Font evaluation method, device, equipment and computer readable storage medium | |
| CN111046760B (en) | Handwriting identification method based on domain countermeasure network | |
| CN112686219B (en) | Handwritten text recognition method and computer storage medium | |
| CN111401099A (en) | Text recognition method, device and storage medium | |
| CN112434568B (en) | Drawing identification method and device, storage medium and computing equipment | |
| CN113177435A (en) | Test paper analysis method and device, storage medium and electronic equipment | |
| CN114663904A (en) | PDF document layout detection method, device, equipment and medium | |
| CN112749696A (en) | Text detection method and device | |
| CN113901952A (en) | Print form and handwritten form separated character recognition method based on deep learning | |
| US20220319215A1 (en) | Image processing method, device, electronic apparatus, and storage medium | |
| CN113592807A (en) | Training method, image quality determination method and device, and electronic equipment | |
| CN113901933A (en) | Electronic invoice information extraction method, device and equipment based on artificial intelligence | |
| CN114330234A (en) | Layout structure analysis method and device, electronic equipment and storage medium | |
| CN109147002B (en) | Image processing method and device | |
| CN117095414A (en) | Handwriting recognition system and recognition method based on dot matrix paper pen |
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: 20191126 |