CN109919037A - A kind of text positioning method and device, text recognition method and device - Google Patents
A kind of text positioning method and device, text recognition method and device Download PDFInfo
- Publication number
- CN109919037A CN109919037A CN201910105748.0A CN201910105748A CN109919037A CN 109919037 A CN109919037 A CN 109919037A CN 201910105748 A CN201910105748 A CN 201910105748A CN 109919037 A CN109919037 A CN 109919037A
- Authority
- CN
- China
- Prior art keywords
- text
- line
- image
- identified
- text image
- 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.)
- Granted
Links
Landscapes
- Image Analysis (AREA)
- Character Input (AREA)
Abstract
This application provides a kind of text recognition methods, belong to text recognition technique field, and accuracy rate is low during solving the problems, such as prior art text identification.The described method includes: obtaining line of text image to be identified;The line of text image to be identified is input to line of text identification model trained in advance, determine that the corresponding line of text recognition result of the line of text image to be identified, the line of text recognition result are used to indicate the line of text attribute of the line of text image to be identified corresponding position;Image-region corresponding with the line of text attribute in the line of text image to be identified is determined according to line of text recognition result.Text positioning method disclosed in the present application determines the distributed areas of the text of different line of text attributes in the line of text image to be identified of complicated arrangement by the line of text identification model of training, help to be identified for different text filed use text image corresponding with the line of text attribute of this article one's respective area identification engines, to promote the accuracy of text identification.
Description
Technical field
This application involves text recognition technique field more particularly to a kind of text positioning method and device, text identification sides
Method and device.
Background technique
File and picture identification process be usually will style of writing this image or column text image be input in advance trained text
This image recognition engine is to obtain corresponding text code.Column text is by being rotated by 90 ° to obtain style of writing originally, therefore, usually row
Text and column text are referred to as style of writing originally.
Text image identification engine in the prior art is that the image of image or single-row text based on single file text carries out
Trained, therefore, for the case where the single file text and multiline text of mixed distribution, text image is known in the text image of input
Other engine is identified as single file text.
For example, most common is exactly the line of text image that texts are annotated by single-row body text and two column in ancient books document,
And existing text image identification engine can identify the line of text of two column annotation texts as uniline body text, it is clear that this
Single-row line of text and multiple row line of text be it is different, therefore, be thus easy to cause biserial annotation text line of text erroneous judgement
It is lower to the recognition accuracy of the image of the column text so as to cause text image identification engine for single-row body text.
To sum up, at least there is recognition accuracy when the text image for carrying out complicated arrangement is identified in the prior art
Low problem.
Summary of the invention
The embodiment of the present application provides a kind of text positioning method, helps to solve text recognition method in the prior art and exists
The low problem of accuracy rate.
In a first aspect, the embodiment of the present application provides a kind of text positioning method, comprising:
Obtain line of text image to be identified;
The line of text image to be identified is input to line of text identification model trained in advance, determines the text to be identified
The corresponding line of text recognition result of current row image, wherein the line of text recognition result is used to indicate the line of text to be identified
The line of text attribute of image corresponding position;
Figure corresponding with the line of text attribute in the line of text image to be identified is determined according to line of text recognition result
As region.
Optionally, described that the line of text image to be identified is input to line of text identification model trained in advance, it determines
Before the step of line of text image to be identified corresponding line of text recognition result, comprising:
Obtain line of text identification model training sample, wherein the sample data of the training sample be preset height and
The line of text image of predetermined width, the sample label of the training sample are used to indicate corresponding position in the line of text image
Line of text attribute;
Using the sample data of the training sample as the input of the line of text identification model, with line of text identification
The output and the minimum target of error of the sample label of the training sample of model, the training line of text identification model,
In, the line of text identification model is constructed based on convolutional neural networks.
Optionally, the step of obtaining the training sample of line of text identification model, comprising:
The line of text image of several preset heights and predetermined width is obtained as sample data, constructs sample data sets;
For the corresponding line of text image of each sample data in the sample data sets, by according to default step
Long mobile specified sliding window is scanned the line of text image along picture traverse direction, to mark the finger according to scanning result
Determine the line of text attribute of line of text image described in the position sequentially passed through in sliding window moving process;
According to the text of line of text image described in the position sequentially passed through in the specified sliding window moving process of label
Current row attribute determines the sample label of respective sample data;Wherein, the height of the specified sliding window is the line of text image
First preset ratio of the preset height, the width of the specified sliding window are the predetermined width of the line of text image
Second preset ratio.
Optionally, the step of acquisition line of text image to be identified, comprising:
By stretch along width and/or short transverse to line of text image to be identified or compression processing, will it is described to
Identify that line of text Image Adjusting is the line of text image to be identified of the preset height and the predetermined width.
Optionally, the line of text recognition result includes line of text image to be identified corresponding position described in sequence identification
The classification results of line of text attribute, it is described according to line of text recognition result determine in the line of text image to be identified with the text
The step of current row attribute corresponding image-region, comprising:
The text diagram image position being sequentially distributed in the line of text image to be identified is determined according to the width of the specified sliding window
It sets, the text image position of the sequence distribution is sequentially corresponding with the classification results of the line of text attribute;
According to the classification results of the line of text attribute, the identical text image position of adjacent and classification results is gathered
It closes, determines the image-region in the line of text image to be identified corresponding from different line of text attributes.
Second aspect, the embodiment of the present application also provides a kind of text recognition methods, comprising: passes through aforementioned first aspect public affairs
The text positioning method opened determines image-region corresponding from different line of text attributes in line of text image to be identified;
By the text image identification model with each line of text attributes match, respectively to corresponding line of text attribute pair
Line of text image to be identified in the image-region answered is identified, determines the line of text image to be identified in respective image region
Recognition result;
According to the position in described image region, the recognition result of the line of text image to be identified in each image-region is carried out
Fusion, determines the corresponding text of the line of text image to be identified.
The third aspect, the embodiment of the present application also provides a kind of String localization devices, comprising:
Line of text image collection module to be identified, for obtaining line of text image to be identified;
Line of text recognition result determining module, for the line of text image to be identified to be input to text trained in advance
Row identification model determines the corresponding line of text recognition result of the line of text image to be identified, wherein the line of text identification knot
Fruit is used to indicate the line of text attribute of the line of text image to be identified corresponding position;
Image-region determining module, for according to line of text recognition result determine in the line of text image to be identified with institute
State the corresponding image-region of line of text attribute.
Optionally, the line of text image to be identified is being input to line of text identification model trained in advance, is determining institute
Before the step of stating line of text image to be identified corresponding line of text recognition result, further includes:
Sample collection module, for obtaining the training sample of line of text identification model, wherein the sample of the training sample
Data are the line of text image of preset height and predetermined width, and the sample label of the training sample is used to indicate the line of text
The line of text attribute of corresponding position in image;
Line of text identification model training module, for being identified using the sample data of the training sample as the line of text
The input of model, with the output of the line of text identification model and the minimum mesh of error of the sample label of the training sample
Mark, the training line of text identification model, wherein the line of text identification model is constructed based on convolutional neural networks.
Optionally, the sample collection module is further used for:
The line of text image of several preset heights and predetermined width is obtained as sample data, constructs sample data sets;
For the corresponding line of text image of each sample data in the sample data sets, by according to default step
Long mobile specified sliding window is scanned the line of text image along picture traverse direction, to mark the finger according to scanning result
Determine the line of text attribute of line of text image described in the position sequentially passed through in sliding window moving process;
According to the text of line of text image described in the position sequentially passed through in the specified sliding window moving process of label
Current row attribute determines the sample label of respective sample data;Wherein, the height of the specified sliding window is the line of text image
First preset ratio of the preset height, the width of the specified sliding window are the predetermined width of the line of text image
Second preset ratio.
Optionally, when obtaining line of text image to be identified, the line of text image collection module to be identified is further used
In:
By stretch along width and/or short transverse to line of text image to be identified or compression processing, will it is described to
Identify that line of text Image Adjusting is the line of text image to be identified of the preset height and predetermined width.
Optionally, the line of text recognition result includes line of text image to be identified corresponding position described in sequence identification
The classification results of line of text attribute, described image area determination module, are further used for:
The text diagram image position being sequentially distributed in the line of text image to be identified is determined according to the width of the specified sliding window
It sets, the text image position of the sequence distribution is sequentially corresponding with the classification results of the line of text attribute;
According to the classification results of the line of text attribute, the identical text image position of adjacent and classification results is gathered
It closes, determines the image-region in the line of text image to be identified corresponding from different line of text attributes.
Fourth aspect, the embodiment of the present application also provides a kind of text identification devices, comprising:
Line of text image-region determining module to be identified, for true by text positioning method described in aforementioned first aspect
Image-region corresponding from different line of text attributes in fixed line of text image to be identified;
Subregion identification module, for passing through the text image identification model with each line of text attributes match, difference
Line of text image to be identified in image-region corresponding with corresponding line of text attribute is identified, determines respective image region
The recognition result of interior line of text image to be identified;
Recognition result Fusion Module, for the position according to described image region, to the text to be identified in each image-region
The recognition result of current row image is merged, and determines the corresponding text of the line of text image to be identified.
5th aspect the embodiment of the present application also provides a kind of electronic equipment, including memory, processor and is stored in institute
The computer program that can be run on memory and on a processor is stated, the processor realizes this when executing the computer program
Apply for text positioning method and/or text recognition method as described in the examples.
6th aspect, the embodiment of the present application also provides a kind of computer readable storage mediums, are stored thereon with computer
Program, the program realize text positioning method described in the embodiment of the present application and/or text recognition method when being executed by processor
The step of.
In this way, text positioning method disclosed in the embodiment of the present application then, will by obtaining line of text image to be identified
The line of text image to be identified is input to line of text identification model trained in advance, determines the line of text image pair to be identified
The line of text recognition result answered, the line of text recognition result are used to indicate the line of text image to be identified corresponding position
Line of text attribute;Finally, according to line of text recognition result determine in the line of text image to be identified with the line of text attribute
Corresponding image-region helps to solve the problems, such as that text identification accuracy rate is low in the prior art.The embodiment of the present application discloses
Text positioning method not identical text is determined in the line of text image to be identified of complicated arrangement by the line of text identification model of training
The distributed areas of the text (such as single file text or multiline text) of current row attribute, facilitate for it is different it is text filed using with
The corresponding text image identification engine of the line of text attribute of this article one's respective area identifies corresponding text filed image, to mention
Rise the accuracy of text identification.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to required in the embodiment of the present application description
Attached drawing to be used is briefly described, it should be apparent that, the drawings in the following description are only some examples of the present application,
For those of ordinary skill in the art, without any creative labor, it can also obtain according to these attached drawings
Obtain other attached drawings.
Fig. 1 is the text positioning method flow chart of the embodiment of the present application one;
Fig. 2 is the text positioning method flow chart of the embodiment of the present application two;
Fig. 3 is the original image that training sample is acquired in the embodiment of the present application two;
Fig. 4 is the line of text image schematic diagram based on Image Acquisition in Fig. 3 in the embodiment of the present application two;
Fig. 5 is the sample data schematic diagram for obtain after stretch processing to the line of text image of acquisition;
Fig. 6 is the schematic diagram that the sample data based on Fig. 5 carries out sample mark;
Fig. 7 is the schematic diagram for carrying out sliding window scanning in the embodiment of the present application two to the sample data of Fig. 6;
Fig. 8 is the line of text identification model schematic network structure used in the embodiment of the present application two;
Fig. 9 is that the recognition result of the text image to be identified in the embodiment of the present application two and input picture corresponding relationship are illustrated
Figure;
Figure 10 is the final positioning result schematic diagram of text image to be identified in the embodiment of the present application two;
Figure 11 is the text recognition method flow chart of the embodiment of the present application three;
Figure 12 is one of String localization apparatus structure schematic diagram of the embodiment of the present application four;
Figure 13 is the two of the String localization apparatus structure schematic diagram of the embodiment of the present application four;
Figure 14 is the text identification apparatus structure schematic diagram of the embodiment of the present application five.
Specific embodiment
Below in conjunction with the attached drawing in the embodiment of the present application, technical solutions in the embodiments of the present application carries out clear, complete
Site preparation description, it is clear that described embodiment is some embodiments of the present application, instead of all the embodiments.Based on this Shen
Please in embodiment, every other implementation obtained by those of ordinary skill in the art without making creative efforts
Example, shall fall in the protection scope of this application.
Different line of text attributes in the embodiment of the present application can for single file text or with duplicate rows text, or
Different literals font or different literals type etc..Understand this programme for the ease of reader, in the embodiment of the present application with difference
Line of text attribute is single file text or illustrates the specific embodiment of text positioning method with duplicate rows text.
Embodiment one:
A kind of text positioning method is present embodiments provided, as shown in Figure 1, which comprises step 10 to step 12.
Step 10, line of text image to be identified is obtained.
Line of text image to be identified described in the embodiment of the present application is the image of pre-set dimension.The line of text to be identified
It can only include the line of text image of single file text in image, or only include the line of text image of multiline text, may be used also
Think the line of text image for not only including single file text but also including the mixing arrangement of multiline text.
When it is implemented, promoting the efficiency of String localization to reduce operand, it is preferable that the text to be identified of acquisition
Row image is gray level image.
Step 11, the line of text image to be identified is input to in advance trained line of text identification model, determine described in
The corresponding line of text recognition result of line of text image to be identified.
Wherein, the line of text recognition result is used to indicate the line of text of the line of text image to be identified corresponding position
Attribute.
The application is when it is implemented, before identifying the line of text image to be identified, it is necessary first to training text
Current row identification model.The line of text identification model is based on convolutional neural networks training, by carrying out to input line of text image
Multiple convolution operation simultaneously carries out feature extraction and mapping, and line of text image to be identified described in final output is along picture traverse direction
The line of text attributive classification result for each image-region that (as from left to right or from right to left) successively divides.Wherein, Mei Getu
As region size and the training line of text identification model when, adopted when to the line of text image pattern of acquisition, sample label is set
The size of sliding window is identical;It include: single file text and multiline text citing, the line of text identification with the line of text attribute
As a result the line of text image to be identified for being input to the line of text identification model is used to indicate along picture traverse direction successively with institute
Sliding window size is stated to be divided to obtain each image-region to be single file text or multiline text.
Step 12, according to line of text recognition result determine in the line of text image to be identified with the line of text attribute pair
The image-region answered.
After the recognition result that line of text image to be identified has been determined, next further recognition result is gathered
It closes.Since the line of text for including in line of text image to be identified may arrange for single file text row or multiline text mixing, and
And the position and length of single file text or multiline text are not fixed, therefore, it is necessary to the line of text obtained to abovementioned steps identifications to tie
Adjacent single file text region and adjacent multiline text region are polymerize in fruit, to determine in the line of text image to be identified
The image-region of single file text and more line of text images.
When it is implemented, above-mentioned line of text recognition result can be expressed as character array, each element of array is respectively referred to
Show the line of text attribute for the image-region being arranged successively in the line of text image to be identified.Wherein, the line of text figure to be identified
The image-region being arranged successively as in is by waiting knowing to this with the width equal with the width for the sliding window mentioned in abovementioned steps
Other line of text image sequentially divides obtained multiple images region along the progress of picture traverse direction.
Text positioning method disclosed in the embodiment of the present application, by obtaining line of text image to be identified, then, will it is described to
Identification line of text image is input to line of text identification model trained in advance, determines the corresponding text of the line of text image to be identified
Current row recognition result, the line of text recognition result are used to indicate the line of text of the line of text image to be identified corresponding position
Attribute;Finally, being determined according to line of text recognition result corresponding with the line of text attribute in the line of text image to be identified
Image-region, it is accurate because multiline text is regarded text identification caused by single file text identifies in the prior art to help to solve
The low problem of rate.Text positioning method disclosed in the embodiment of the present application determines complicated row by the line of text identification model of training
The distributed areas of the text (such as single file text or multiline text) of different line of text attributes, have in the line of text image to be identified of cloth
Help for different text filed use text image corresponding with the line of text attribute of this article one's respective area identification engines to phase
Text filed image is answered to be identified, to promote the accuracy of text identification.
Embodiment two:
A kind of text positioning method is present embodiments provided, as shown in Figure 2, which comprises step 20 to step 23.
Step 20, training text row identification model.
In some embodiments of the present application, line of text image to be identified is input to line of text trained in advance and identifies mould
Type, before the step of determining the line of text image to be identified corresponding line of text recognition result, further includes: the identification of training text row
Model.When it is implemented, training text row identification model further comprises: the training sample of line of text identification model is obtained,
In, the sample data of training sample is the line of text image of preset height and predetermined width, and the sample label of the training sample is used
In the line of text attribute for indicating corresponding position in above-mentioned line of text image;Using the sample data of the training sample as line of text
The input of identification model, with the output and the minimum target of error of the sample label of the training sample of line of text identification model,
Training this article current row identification model, wherein this article current row identification model is constructed based on convolutional neural networks.
When it is implemented, firstly the need of training sample is collected.
In some embodiments of the present application, obtain line of text identification model training sample the step of, comprising: if obtain
The line of text image of dry preset height and predetermined width constructs sample data sets as sample data;For sample data set
The corresponding line of text image of each sample data in conjunction, by specifying sliding window to this article current row figure according to preset step-length is mobile
As being scanned along picture traverse direction, to mark everybody sequentially passed through in the specified sliding window moving process according to scanning result
Set the line of text attribute of place's this article current row image;According to the position sequentially passed through in the specified sliding window moving process of label
The line of text attribute of this article current row image, determines the sample label of respective sample data;Wherein, the height of the specified sliding window is to be somebody's turn to do
First preset ratio of the aforementioned preset height of line of text image, the width of the specified sliding window are the aforementioned pre- of this article current row image
If the second preset ratio of width.Further, the first preset ratio and the second preset ratio are according to line of text figure to be identified
The length and text distribution height of picture determine;Preset step-length is generally equal to the width of sliding window.In the present embodiment, for the ease of retouching
State, by from left to right line of text image is scanned along picture traverse direction or convolution algorithm for, illustrate String localization side
The specific technical solution of method.
In some embodiments of the present application, can choose the image of ancient books, document as original image, then to it is original into
The processing of row gray processing, and every a line or the corresponding image of each column content are partitioned into as line of text image.When with as shown in Figure 3
Local chronicle image as original image, it is available every by being pre-processed to original image when acquiring training sample
The image of one column text, such as the image of the column text in rectangular area 310.Then, by 90 degree of the image rotation of each column text,
A width line of text image is obtained, as shown in Figure 4.
When it is implemented, the text image for training text row identification model can have unified size, it is therefore, right
The line of text image acquired in through different original images carries out stretching or compression processing, to normalize to unified ruler
It is very little.When such as carrying out stretch processing to line of text image, the height of the line of text image stretched will be needed to be adjusted to preset height (such as
64), width adjustment is to predetermined width (such as 1280).Stretched processing will obtain line of text image as shown in Figure 5.
Pass through the available several line of text images of the above method, sample of every width line of text image as a training sample
Notebook data constructs sample data sets.The application when it is implemented, aforementioned preset height according to the row of text image to be identified
High, ancient books file col width, the high data of row determine;Aforementioned predetermined width is true according to data such as long, the row length of column of ancient books file
It is fixed.
Further, sample label is arranged to each sample data in sample data sets.
Firstly, to each sample data (i.e. each width line of text image) in sample data sets, manually mark is different
The coordinate of the image-region of line of text attribute and corresponding line of text attribute.For example, marking each duplicate rows in every width line of text image
Text filed top left co-ordinate and bottom right angular coordinate and the text filed quantity of duplicate rows.With line of text image shown in fig. 5
Citing, annotation results schematic diagram are as shown in Figure 6.Wherein, bound [] indicates the text filed top left co-ordinate and bottom right of duplicate rows
Angular coordinate;DNum indicates the quantity of duplicate rows text in this article current row image.
When it is implemented, passing through cunning to each sample data (i.e. each width line of text image) in sample data sets
The text pixel of window scan text row image corresponding position is distributed the sample label for determining the sample data.For example, with sample
Height of the height of line of text image as specified sliding window in data acquisition system, alternatively, with line of text image in sample data sets
Height height of the first preset ratio (such as 4/5 to 6/5) as specified sliding window;With line of text image in sample data sets
Width width of the second preset ratio (such as 1/40) as specified sliding window, determine specified sliding window.
Then, to each sample data in sample data sets, i.e., each width line of text image, from this article current row figure
The left-most position of picture rises, and with the mobile specified sliding window of preset step-length, each position of the specified sliding window corresponds to this article current row
One position of image, the corresponding image-region in the position of line of text image, the width of each image-region are specified with this
The width of sliding window is equal.It is as shown in Figure 7 by several image-regions for specifying sliding window to be scanned determining line of text image.This
In embodiment, aforementioned preset step-length is the width of 1 specified sliding window, when it is implemented, can be according to the accurate of line of text positioning
Property and the comprehensive step-length for determining that specified sliding window is mobile of operation efficiency.
Further, for determining each image-region, further according to the abscissa of the image-region and to this article
The text filed upper left corner abscissa of the duplicate rows of current row image labeling and lower right corner abscissa determine whether the image-region is double
Style of writing one's respective area.It, can will be in the corresponding sample label of the image-region if the image-region is that duplicate rows is text filed
Data are to be set to indicate that the line of text attribute tags of duplicate rows text, such as 0, i.e. the scanning result of the image-region is 0;If should
The non-duplicate rows of image-region is text filed, then can be to be set to indicate that list by the data in the corresponding sample label of the image-region
The line of text attribute tags of style of writing originally, such as 1, the i.e. scanning result of the image-region are 1.
It, can be with when judging a certain image-region is text filed duplicate rows or single file text region in specific implementation process
According to the text filed area accounting judgement of duplicate rows in the image-region.For example, it is text filed to work as duplicate rows in a certain image-region
Area be greater than the image-region area 1/2 when, it is determined that the image-region be duplicate rows it is text filed, otherwise, it determines the figure
As region is single file text region.
Later, each of this article current row image successively determining in above-mentioned specified sliding window from left to right moving process will be passed through
The array of 40 dimensions of the line of text attribute tags composition of image-region, as the sample label of this article current row image, i.e., corresponding sample
The sample label of notebook data.The sample label of line of text image shown in Fig. 7 can identify are as follows:
Label=[0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,1,1,1,1,0,
0,0,0,0,0,0,0,0,0,0]。
When it is implemented, accuracy and the comprehensive determination of operation efficiency that aforementioned second preset ratio is positioned according to line of text.
The application is when it is implemented, also need to construct line of text identification model.
In the embodiment of the present application, line of text identification model is constructed based on convolutional neural networks.When it is implemented, can construct
The line of text identification model of network structure as shown in Figure 8.This article current row identification model from input side to outlet side successively are as follows: by
Tie up convolutional layers, the maximum pond layer that convolution kernel size is 2 × 2, by rolling up in the 2 of 16 convolution kernels composition that convolution kernel size is 7 × 7
Tie up convolutional layers, the maximum pond layer that convolution kernel size is 2 × 2, by convolution in the 2 of 32 convolution kernels composition that product core size is 5 × 5
Tie up convolutional layers, the maximum pond layer that convolution kernel size is 2 × 2, by convolution kernel in the 2 of 64 convolution kernels composition that core size is 3 × 3
It is maximum pond layer that 2 dimension convolutional layers that 64 convolution kernels that size is 3 × 3 are constituted, convolution kernel size are 2 × 2, big by convolution kernel
It is small be 2 × 2 128 convolution kernels constitute 2 dimension convolutional layers, convolution kernel size be 2 × 2 maximum pond layer, batch standardize layer,
2 dimension convolutional layers, a vector transposition process layer, the vector being made of 1 convolution kernel that convolution kernel size is 1 × 1 flatten layer.Its
In, the output of the last one convolutional layer indicates the classification results of each image-region in the line of text image of input.
When it is implemented, opencv (cross-platform computer vision library based on BSD license (open source) distribution) can be passed through
Or other existing kits read each sample data (i.e. line of text image), obtain 64 × 1280 × 1 gray level image, then
The gray level image of reading is input to aforementioned texts row identification model.Size of the gray level image after first time convolution is 64
× 1280 × 16, it is then passed through primary maximum pondization operation, the size by maximum Chi Huahou for the first time is 32 × 640 × 16.
Successively convolution sum pondization processing by above layers will obtain size as the vector of 2 × M × 1, such as [[0.2,0.6,
0.3,…,0.9],[0.8,,0.4,0.7,…,0.1]].Later, the first peacekeeping second dimension of the vector of 2 × M × 1 is turned
Processing is set, keeps the vector corresponding in width direction with the line of text image of input, the vector that size is M × 2 × 1 will be obtained.Most
Afterwards, pass through, TimeDistrition (Flatten) indicates that the data of the last bidimensional to transposition treated vector flatten
Processing, obtains the output of model, such as [[0.2,0.8], [0.6,0.4], [0.3,0.7], [0.9,0.1] ...], wherein M is equal to
40, for the quantity for the image-region that each sample data is designated to determine when sliding window scanning.
Later, based on the training sample training text row identification model in above-mentioned sample set.The training text that training obtains
Current row identification model can the line of text image to a certain length handle, and export and each image-region of input picture
Corresponding line of text attributive character.
The process of the deconvolution parameter of each layer in the really continuous solution of the training process of model, Optimized model, by reversed
Transmission method, with the output of line of text identification model and the minimum mesh of error of the sample label of the line of text image accordingly inputted
Mark solves optimized parameter, is finally completed the training of this article current row identification model.The specific training process of model is referring to existing skill
Art repeats no more in the present embodiment.
The application prevents mould when it is implemented, equilibrium treatment can be done to the sample data in sample data sets first
Type training is inclined.Meanwhile the sample in sample data sets is carried out to upset the extensive effect handled to be got well at random, with ratio
0.8 for total sample is used as training set, is left to be used as test set, to verify the extensive energy for the line of text identification model that training obtains
Power.
Step 21, line of text image to be identified is obtained.
Line of text image to be identified described in the embodiment of the present application is the image of pre-set dimension.The line of text to be identified
It can only include the line of text image of single file text in image, or only include the line of text image of multiline text, may be used also
Think and as shown in Figure 3 not only included single file text and included the line of text image arranged of mixing of multiline text.
When it is implemented, promoting the efficiency of String localization to reduce operand, it is preferable that the text to be identified of acquisition
Row image is gray level image.
In some embodiments of the present application, the step of acquisition line of text image to be identified, comprising: known by treating
Other line of text image carries out stretching or compression processing along width and/or short transverse, is by the line of text Image Adjusting to be identified
The line of text image to be identified of the preset height and predetermined width.Because training sample is pre- during model training
If the line of text image of height and predetermined width, therefore, during being identified, if the width of line of text image to be identified
Degree is not equal to aforementioned predetermined width, then needs in the width direction to carry out the line of text image to be identified at stretching or compression
Reason, by the width adjustment of the line of text image to be identified to aforementioned predetermined width.Further, if line of text image to be identified
Height be not equal to aforementioned preset height, it is also necessary to the line of text image to be identified is stretched or is compressed along short transverse
Processing, is adjusted to aforementioned preset height for the height of the line of text image to be identified, such as by the size of line of text image to be identified
It is adjusted to 64 × 1280.
Step 22, the line of text image to be identified is input to in advance trained line of text identification model, determine described in
The corresponding line of text recognition result of line of text image to be identified.
Wherein, the line of text recognition result is used to indicate the line of text of the line of text image to be identified corresponding position
Attribute.
The application is when it is implemented, before identifying the line of text image to be identified, it is necessary first to training text
Current row identification model.The line of text identification model is based on convolutional neural networks training, by carrying out to input line of text image
Multiple convolution operation simultaneously carries out feature extraction and mapping, and line of text image to be identified described in final output successively divides from left to right
Each image-region line of text attributive classification result.Wherein, the size of each image-region and the training line of text are known
When other model, the size of the sliding window used when sample label is arranged to the line of text image pattern of acquisition is identical;With the text
Row attribute includes: single file text and multiline text citing, and the line of text recognition result, which is used to indicate, is input to the line of text
The line of text image to be identified of identification model successively divide with the sliding window size (as from left to right) in the width direction
It is single file text or multiline text to each image-region.
Step 23, according to line of text recognition result determine in the line of text image to be identified with the line of text attribute pair
The image-region answered.
After the recognition result that line of text image to be identified has been determined, next further recognition result is gathered
It closes.In some embodiments of the present application, line of text recognition result includes that line of text image to be identified described in sequence identification is corresponding
The classification results of line of text attribute at position, it is described to be determined in the line of text image to be identified according to line of text recognition result
The step of image-region corresponding with the line of text attribute, comprising: determined according to the width of the specified sliding window described wait know
The text image position being sequentially distributed in other line of text image, the text image position of sequence distribution sequentially with the text
The classification results of row attribute are corresponding;According to the classification results of the line of text attribute, text identical to adjacent and classification results
Picture position is polymerize, and determines the image-region in the line of text image to be identified corresponding from different line of text attributes.
It take sample label when training text row identification model as the array of 40 elements, i.e., each sample data is designated to slide
The image-region determined when window scans is 40 citings, and the line of text recognition result of this article current row identification model will be 40 × 2
Two-dimensional array, such as it is expressed as A [40,2], wherein the value of each elements A in array is used to indicate line of text image to be identified
The image-region of middle corresponding position is identified as the probability of single file text or duplicate rows text.For example, array element A [0,0] can be used
The image-region of first specified sliding window position is identified as the probability of single file text in expression line of text image to be identified;Array
It is double that elements A [0,1] can be used to indicate that the image-region of first specified sliding window position in line of text image to be identified is identified as
The probability of style of writing originally.When the probability that the image-region of a certain position in line of text image to be identified is identified as single file text is greater than in advance
If when threshold value, it is determined that the line of text attribute of the image-region is single file text.For example, work as A [0,0] > A [0,1] when, indicate to
The image-region of first specified sliding window position is identified as single file text in identification line of text image, such as by its line of text attribute
Labeled as 0;Conversely, determining that the line of text attribute of the image-region is duplicate rows text, such as its line of text attribute is labeled as 1.
Wherein, the width of each image-region corresponds to the width of aforementioned specified sliding window.
The each image-region being sequentially distributed from left to right in line of text image to be identified can be determined according to preceding method
Line of text attribute label, such as:
0,0,0,0,0,0,0,0,0,0,0,0,0,1,1,1,1,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
0,0,0,0,0,0.Wherein, a figure in the corresponding line of text image to be identified of each line of text attribute label of sequential
As the classification results in region.The corresponding relationship of line of text attribute label and an image-region in line of text image to be identified is such as
Shown in Fig. 9.
Further, identical image-region is marked to polymerize adjacent and line of text attribute, to adjacent and classification
As a result identical text image position is polymerize, and is determined in line of text image to be identified corresponding from different line of text attributes
Image-region.It is if 13, the left side line of text attribute label in Fig. 9 is 0, then this 13 line of text attribute labels are corresponding
Image district in line of text image to be identified is polymerize, by the line of text attribute of the image-region obtained after polymerization be determined as with
The corresponding line of text attribute of line of text attribute label 0, i.e. single file text.For another example the left side in Fig. 9 plays the 14th to the 18th
Line of text attribute label be 1, then by this 5 line of text attributes mark the image district in corresponding line of text image to be identified into
Row polymerization, is determined as line of text corresponding with line of text attribute label 1 for the line of text attribute of the image-region obtained after polymerization
Attribute, i.e. multiline text.In this way, the image district of the different line of text attributes in line of text image to be identified can be determined
Domain.As shown in Figure 10, it identifies and polymerize by aforementioned texts row, finally determine 3 figures for including in line of text image to be identified
As region 1010,1011 and 1012, wherein image-region 1010 and 1012 is the image-region of corresponding single file text, image district
Domain 1011 is the image-region of corresponding single file text.
Text positioning method disclosed in the embodiment of the present application, by being identified based on the preparatory training text row of convolutional neural networks
The line of text image to be identified is input to line of text trained in advance after getting line of text image to be identified by model
Identification model determines that the corresponding line of text recognition result of the line of text image to be identified, this article current row recognition result are used to indicate
The line of text attribute of the line of text image to be identified corresponding position;Finally, determining that this is to be identified according to line of text recognition result
Image-region corresponding with above-mentioned line of text attribute in line of text image helps to solve text identification accuracy rate in the prior art
Low problem.Text positioning method disclosed in the embodiment of the present application determines complicated arrangement by the line of text identification model of training
Line of text image to be identified in different line of text attributes text (such as single file text or multiline text) distributed areas, help
Identify engine to corresponding in the text filed use text image corresponding to the line of text attribute of this article one's respective area for being directed to different
Text filed image is identified, to promote the accuracy of text identification.
Embodiment three:
Correspondingly, as shown in figure 11, the embodiment of the present application also discloses a kind of text recognition method, including step 1101 to
Step 1103.
Step 1101, image-region corresponding from different line of text attributes in line of text image to be identified is determined.
When it is implemented, passing through String localization described in embodiment one or embodiment two for line of text image to be identified
Method determines image-region corresponding from different line of text attributes in the line of text image to be identified, such as corresponding with single file text
Image-region, image-region corresponding with multiline text.
Step 1102, by the text image identification model with each line of text attributes match, respectively to corresponding line of text
Line of text image to be identified in the corresponding image-region of attribute is identified, determines the text to be identified in respective image region
The recognition result of row image.
Next, being carried out respectively by single file text image recognition model pair each image-region corresponding with single file text
Identification, obtains corresponding uniline recognition result;Pass through multiline text image recognition model pair each image corresponding with multiline text
Region is identified respectively, obtains corresponding multirow recognition result.
Step 1103, according to the position of above-mentioned each image-region, to the line of text image to be identified in each image-region
Recognition result is merged, and determines the corresponding text of the line of text image to be identified.
Finally, to obtained uniline recognition result and multirow recognition result, according to corresponding image-region in text to be identified
Position in current row image is spliced, and the recognition result in the line of text image to be identified is obtained.
When it is implemented, the difference line of text attribute can for single file text or with duplicate rows text, or it is different
Character script or different literals type.
Text recognition method disclosed in the embodiment of the present application, by determination line of text image to be identified from different line of text
The corresponding image-region of attribute, then, by the text image identification model with each line of text attributes match, respectively to
Accordingly the line of text image to be identified in the corresponding image-region of line of text attribute is identified, is determined in respective image region
The recognition result of line of text image to be identified, according to the position in described image region, to the text to be identified in each image-region
The recognition result of row image is merged, and is determined the corresponding text of the line of text image to be identified, is helped to promote complicated row
The identification accuracy of the text image of cloth.
Example IV:
Correspondingly, the embodiment of the present application also discloses a kind of String localization device, as shown in figure 12, described device includes:
Line of text image collection module 121 to be identified, for obtaining line of text image to be identified;
Line of text recognition result determining module 122, for above-mentioned line of text image to be identified to be input to training in advance
Line of text identification model determines the corresponding line of text recognition result of above-mentioned line of text image to be identified, wherein above-mentioned line of text is known
Other result is used to indicate the line of text attribute of above-mentioned line of text image to be identified corresponding position;
Image-region determining module 123, for being determined in above-mentioned line of text image to be identified according to line of text recognition result
Image-region corresponding with each line of text attribute.
Optionally, above-mentioned line of text image to be identified is being input to in advance trained line of text identification model, in determination
Before stating the corresponding line of text recognition result of line of text image to be identified, as shown in figure 13, above-mentioned String localization device further include:
Sample collection module 124, for obtaining the training sample of line of text identification model, wherein above-mentioned training sample
Sample data is the line of text image of preset height and predetermined width, and the sample label of training sample is used to indicate line of text image
The line of text attribute of middle corresponding position;
Line of text identification model training module 125, for using the sample data of above-mentioned training sample as above-mentioned line of text
The input of identification model, it is minimum with the error of the sample label of above-mentioned training sample with the output of above-mentioned line of text identification model
Target, the training line of text identification model, wherein above-mentioned line of text identification model is constructed based on convolutional neural networks.
Optionally, the sample collection module 124 is further used for:
The line of text image of several preset heights and predetermined width is obtained as sample data, constructs sample data sets;
For the corresponding line of text image of each sample data in the sample data sets, by according to default step
Long mobile specified sliding window is scanned above-mentioned line of text image along picture traverse direction, to mark above-mentioned finger according to scanning result
Determine the line of text attribute of the above-mentioned line of text image in the position sequentially passed through in sliding window moving process;
According to the text of the above-mentioned line of text image in position sequentially passed through in the above-mentioned specified sliding window moving process of label
Current row attribute determines the sample label of respective sample data;Wherein, the height of above-mentioned specified sliding window is above-mentioned line of text image
First preset ratio of the preset height, the width of above-mentioned specified sliding window are the above-mentioned predetermined width of above-mentioned line of text image
Second preset ratio.
Optionally, when obtaining line of text image to be identified, above-mentioned line of text image collection module 121 to be identified is further
For:
By carrying out stretching or compression processing along width and/or short transverse to line of text image to be identified, this is waited knowing
Other line of text Image Adjusting is the line of text image to be identified of above-mentioned preset height and predetermined width.
Optionally, the line of text recognition result includes the text of the sequence identification line of text image to be identified corresponding position
The classification results of current row attribute, above-mentioned image-region determining module 123, are further used for:
The text diagram image position being sequentially distributed in above-mentioned line of text image to be identified is determined according to the width of the specified sliding window
It sets, the text image position of said sequence distribution is sequentially corresponding with the classification results of above-mentioned line of text attribute;
According to the classification results of the line of text attribute, the identical text image position of adjacent and classification results is gathered
It closes, determines the image-region in the line of text image to be identified corresponding from different line of text attributes.The embodiment of the present application is public
Then the line of text image to be identified is input to pre- by the String localization device opened by obtaining line of text image to be identified
First trained line of text identification model, determines the corresponding line of text recognition result of the line of text image to be identified, the text
Row recognition result is used to indicate the line of text attribute of the line of text image to be identified corresponding position;Finally, according to line of text
Recognition result determines image-region corresponding with the line of text attribute in the line of text image to be identified, helps to solve existing
There is the problem that text recognition accuracy is low in technology.The text that String localization device disclosed in the embodiment of the present application passes through training
Row identification model determines text (such as single file text or more of different line of text attributes in the line of text image to be identified of complicated arrangement
Style of writing this) distributed areas, facilitate for different text filed using corresponding with the line of text attribute of this article one's respective area
Text image identifies that engine identifies corresponding text filed image, to promote the accuracy of text identification.
Embodiment five:
Correspondingly, the embodiment of the present application also discloses a kind of text identification device, as shown in figure 14, described device includes:
Line of text image-region determining module 141 to be identified, for passing through text described in embodiment one and embodiment two
Localization method determines image-region corresponding from different line of text attributes in line of text image to be identified;
Subregion identification module 142, for dividing by aforementioned and each line of text attributes match text image identification model
The other line of text image to be identified in image-region corresponding with corresponding line of text attribute identifies, determines respective image area
The recognition result of line of text image to be identified in domain;
Recognition result Fusion Module 143, for the position according to described image region, to be identified in each image-region
The recognition result of line of text image is merged, and determines the corresponding text of line of text image to be identified.
Text identification device disclosed in the embodiment of the present application, by determination line of text image to be identified from different line of text
The corresponding image-region of attribute, then, by the text image identification model with each line of text attributes match, respectively to
Accordingly the line of text image to be identified in the corresponding image-region of line of text attribute is identified, is determined in respective image region
The recognition result of line of text image to be identified, according to the position in described image region, to the text to be identified in each image-region
The recognition result of row image is merged, and is determined the corresponding text of the line of text image to be identified, is helped to promote complicated row
The identification accuracy of the text image of cloth.
Correspondingly, the embodiment of the present application also discloses a kind of electronic equipment, the electronic equipment, including memory, processing
Device and it is stored in the computer program that can be run on the memory and on a processor, the processor executes the computer
Text positioning method described in the embodiment of the present application one and embodiment two is realized when program, and/or, realize the embodiment of the present application three
The text recognition method.The electronic equipment can be mobile phone, PAD, tablet computer, human face recognition machine etc..
Correspondingly, being stored thereon with computer journey the embodiment of the present application also provides a kind of computer readable storage medium
The step of sequence, which realizes text positioning method described in the embodiment of the present application one and embodiment two when being executed by processor,
And/or the step of realizing text recognition method described in the embodiment of the present application three.
The Installation practice of the application is corresponding with method, the specific implementation side of each module and each unit in Installation practice
Formula is embodiment referring to method, and details are not described herein again.
Those of ordinary skill in the art may be aware that list described in conjunction with the examples disclosed in the embodiments of the present disclosure
Member and algorithm steps can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions are actually
It is implemented in hardware or software, the specific application and design constraint depending on technical solution.Professional technician
Each specific application can be used different methods to achieve the described function, but this realization is it is not considered that exceed
Scope of the present application.
One with ordinary skill in the art would appreciate that in embodiment provided herein, it is described to be used as separation unit
The unit of explanation may or may not be physically separated, it can and it is in one place, or can also be distributed
Onto multiple network units.In addition, each functional unit in each embodiment of the application can integrate in a processing unit
In, it is also possible to each unit and physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer readable storage medium.Based on this understanding, the technical solution of the application can be produced with software
The form of product embodies, which is stored in a storage medium, including some instructions are used so that one
Platform computer equipment (can be personal computer, server or the network equipment etc.) executes described in each embodiment of the application
The all or part of the steps of method.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, ROM, RAM, magnetic or disk etc.
The various media that can store program code.
The above, the only specific embodiment of the application, but the protection scope of the application is not limited thereto, ability
Domain those of ordinary skill is it is to be appreciated that unit described in conjunction with the examples disclosed in the embodiments of the present disclosure and algorithm steps
Suddenly, it can be realized with the combination of electronic hardware or computer software and electronic hardware.These functions actually with hardware still
Software mode executes, specific application and design constraint depending on technical solution.Professional technician can be to each
Specific application is to use different methods to achieve the described function, but this realization is it is not considered that exceed the model of the application
It encloses.
Claims (14)
1. a kind of text positioning method characterized by comprising
Obtain line of text image to be identified;
The line of text image to be identified is input to line of text identification model trained in advance, determines the line of text to be identified
The corresponding line of text recognition result of image, wherein the line of text recognition result is used to indicate the line of text image to be identified
The line of text attribute of corresponding position;
Image district corresponding with the line of text attribute in the line of text image to be identified is determined according to line of text recognition result
Domain.
2. the method according to claim 1, wherein described be input to the line of text image to be identified in advance
Trained line of text identification model, before the step of determining the line of text image to be identified corresponding line of text recognition result,
Include:
Obtain the training sample of line of text identification model, wherein the sample data of the training sample is preset height and presets
The line of text image of width, the sample label of the training sample are used to indicate the text of corresponding position in the line of text image
Current row attribute;
Using the sample data of the training sample as the input of the line of text identification model, with the line of text identification model
Output and the minimum target of error of the sample label of the training sample, the training line of text identification model, wherein institute
Stating line of text identification model is constructed based on convolutional neural networks.
3. according to the method described in claim 2, it is characterized in that, obtain line of text identification model training sample the step of,
Include:
The line of text image of several preset heights and predetermined width is obtained as sample data, constructs sample data sets;
For the corresponding line of text image of each sample data in the sample data sets, by being moved according to preset step-length
Dynamic specified sliding window is scanned the line of text image along picture traverse direction, to mark the specified cunning according to scanning result
The line of text attribute of line of text image described in the position sequentially passed through in window moving process;
According to the line of text of line of text image described in the position sequentially passed through in the specified sliding window moving process of label
Attribute determines the sample label of respective sample data;Wherein, the height of the specified sliding window is the described of the line of text image
First preset ratio of preset height, the width of the specified sliding window are the second of the predetermined width of the line of text image
Preset ratio.
4. according to the method described in claim 2, it is characterized in that, the step of acquisition line of text image to be identified, comprising:
It, will be described to be identified by carrying out stretching or compression processing along width and/or short transverse to line of text image to be identified
Line of text Image Adjusting is the line of text image to be identified of the preset height and the predetermined width.
5. according to the method described in claim 3, it is characterized in that, the line of text recognition result include described in sequence identification to
Identify the classification results of the line of text attribute of line of text image corresponding position, it is described according to the determination of line of text recognition result
In line of text image to be identified the step of image-region corresponding with the line of text attribute, comprising:
The text image position being sequentially distributed in the line of text image to be identified, institute are determined according to the width of the specified sliding window
The text image position for stating sequence distribution is sequentially corresponding with the classification results of the line of text attribute;
According to the classification results of the line of text attribute, the identical text image position of adjacent and classification results is polymerize,
Determine the image-region in the line of text image to be identified corresponding from different line of text attributes.
6. a kind of text recognition method characterized by comprising
By text positioning method described in any one of claim 1 to 5 determine in line of text image to be identified from different texts
The corresponding image-region of row attribute;
By the text image identification model with each line of text attributes match, respectively to corresponding with corresponding line of text attribute
Line of text image to be identified in image-region is identified, determines the knowledge of the line of text image to be identified in respective image region
Other result;
According to the position in described image region, the recognition result of the line of text image to be identified in each image-region is melted
It closes, determines the corresponding text of the line of text image to be identified.
7. a kind of String localization device characterized by comprising
Line of text image collection module to be identified, for obtaining line of text image to be identified;
Line of text recognition result determining module is known for the line of text image to be identified to be input to line of text trained in advance
Other model determines the corresponding line of text recognition result of the line of text image to be identified, wherein the line of text recognition result is used
In the line of text attribute for indicating the line of text image to be identified corresponding position;
Image-region determining module, for according to line of text recognition result determine in the line of text image to be identified with the text
The corresponding image-region of current row attribute.
8. device according to claim 7, which is characterized in that the line of text image to be identified is being input to preparatory instruction
Experienced line of text identification model, before the step of determining the line of text image to be identified corresponding line of text recognition result, also
Include:
Sample collection module, for obtaining the training sample of line of text identification model, wherein the sample data of the training sample
For the line of text image of preset height and predetermined width, the sample label of the training sample is used to indicate the line of text image
The line of text attribute of middle corresponding position;
Line of text identification model training module, for using the sample data of the training sample as the line of text identification model
Input, with the output and the minimum target of error of the sample label of the training sample of the line of text identification model, instruction
Practice the line of text identification model, wherein the line of text identification model is constructed based on convolutional neural networks.
9. device according to claim 8, which is characterized in that the sample collection module is further used for:
The line of text image of several preset heights and predetermined width is obtained as sample data, constructs sample data sets;
For the corresponding line of text image of each sample data in the sample data sets, by being moved according to preset step-length
Dynamic specified sliding window is scanned the line of text image along picture traverse direction, to mark the specified cunning according to scanning result
The line of text attribute of line of text image described in the position sequentially passed through in window moving process;
According to the line of text of line of text image described in the position sequentially passed through in the specified sliding window moving process of label
Attribute determines the sample label of respective sample data;Wherein, the height of the specified sliding window is the described of the line of text image
First preset ratio of preset height, the width of the specified sliding window are the second of the predetermined width of the line of text image
Preset ratio.
10. device according to claim 8, which is characterized in that described to be identified when obtaining line of text image to be identified
Line of text image collection module is further used for:
It, will be described to be identified by carrying out stretching or compression processing along width and/or short transverse to line of text image to be identified
Line of text Image Adjusting is the line of text image to be identified of the preset height and the predetermined width.
11. device according to claim 9, which is characterized in that the line of text recognition result includes described in sequence identification
The classification results of the line of text attribute of line of text image to be identified corresponding position, described image area determination module, further
For:
The text image position being sequentially distributed in the line of text image to be identified, institute are determined according to the width of the specified sliding window
The text image position for stating sequence distribution is sequentially corresponding with the classification results of the line of text attribute;
According to the classification results of the line of text attribute, the identical text image position of adjacent and classification results is polymerize,
Determine the image-region in the line of text image to be identified corresponding from different line of text attributes.
12. a kind of text identification device characterized by comprising
Line of text image-region determining module to be identified, for passing through text positioning method described in any one of claim 1 to 5
Determine image-region corresponding from different line of text attributes in line of text image to be identified;
Subregion identification module, for by the text image identification model with each line of text attributes match, respectively to
Accordingly the line of text image to be identified in the corresponding image-region of line of text attribute is identified, is determined in respective image region
The recognition result of line of text image to be identified;
Recognition result Fusion Module, for the position according to described image region, to the line of text to be identified in each image-region
The recognition result of image is merged, and determines the corresponding text of the line of text image to be identified.
13. a kind of electronic equipment, including memory, processor and it is stored on the memory and can runs on a processor
Computer program, which is characterized in that the processor realizes claim 1 to 5 any one when executing the computer program
The text positioning method, and/or, realize text recognition method as claimed in claim 6.
14. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The step of text positioning method described in claim 1 to 5 any one is realized when execution, and/or, realize claim 6 institute
The step of text recognition method stated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910105748.0A CN109919037B (en) | 2019-02-01 | 2019-02-01 | Text positioning method and device and text recognition method and device |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910105748.0A CN109919037B (en) | 2019-02-01 | 2019-02-01 | Text positioning method and device and text recognition method and device |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109919037A true CN109919037A (en) | 2019-06-21 |
CN109919037B CN109919037B (en) | 2021-09-07 |
Family
ID=66961430
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910105748.0A Active CN109919037B (en) | 2019-02-01 | 2019-02-01 | Text positioning method and device and text recognition method and device |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109919037B (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110309796A (en) * | 2019-07-05 | 2019-10-08 | 上海中通吉网络技术有限公司 | Face list recognition methods, device and equipment |
CN110503189A (en) * | 2019-08-02 | 2019-11-26 | 腾讯科技(深圳)有限公司 | A kind of data processing method and device |
CN111798542A (en) * | 2020-09-10 | 2020-10-20 | 北京易真学思教育科技有限公司 | Model training method, data processing device, model training apparatus, and storage medium |
CN111950397A (en) * | 2020-07-27 | 2020-11-17 | 腾讯科技(深圳)有限公司 | Text labeling method, device and equipment for image and storage medium |
CN112541438A (en) * | 2020-12-15 | 2021-03-23 | 深兰人工智能(深圳)有限公司 | Text recognition method and device |
CN113780131A (en) * | 2021-08-31 | 2021-12-10 | 众安在线财产保险股份有限公司 | Text image orientation recognition method and text content recognition method, device and equipment |
CN114677691A (en) * | 2022-04-06 | 2022-06-28 | 北京百度网讯科技有限公司 | Text recognition method and device, electronic equipment and storage medium |
Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004094292A (en) * | 2002-08-29 | 2004-03-25 | Ricoh Co Ltd | Character recognizing device, character recognizing method, and program used for executing the method |
CN1741035A (en) * | 2005-09-23 | 2006-03-01 | 清华大学 | Blocks letter Arabic character set text dividing method |
CN105574523A (en) * | 2014-10-31 | 2016-05-11 | 株式会社东芝 | Character recognition device and method, image display device, and image retrieval device |
CN105608454A (en) * | 2015-12-21 | 2016-05-25 | 上海交通大学 | Text structure part detection neural network based text detection method and system |
CN105912225A (en) * | 2016-03-31 | 2016-08-31 | 北京奇虎科技有限公司 | Barrage display method and device based on encyclopedic items |
CN105989341A (en) * | 2015-02-17 | 2016-10-05 | 富士通株式会社 | Character recognition method and device |
US20170147536A1 (en) * | 2011-05-23 | 2017-05-25 | Nuance Communications, Inc. | Text browsing, editing and correction methods for automotive applications |
CN107016387A (en) * | 2016-01-28 | 2017-08-04 | 苏宁云商集团股份有限公司 | A kind of method and device for recognizing label |
CN107180239A (en) * | 2017-06-09 | 2017-09-19 | 科大讯飞股份有限公司 | Line of text recognition methods and system |
CN107220641A (en) * | 2016-03-22 | 2017-09-29 | 华南理工大学 | A kind of multi-language text sorting technique based on deep learning |
CN108304814A (en) * | 2018-02-08 | 2018-07-20 | 海南云江科技有限公司 | A kind of construction method and computing device of literal type detection model |
CN108376244A (en) * | 2018-02-02 | 2018-08-07 | 北京大学 | A kind of recognition methods of text font in natural scene picture |
CN108664996A (en) * | 2018-04-19 | 2018-10-16 | 厦门大学 | A kind of ancient writing recognition methods and system based on deep learning |
-
2019
- 2019-02-01 CN CN201910105748.0A patent/CN109919037B/en active Active
Patent Citations (13)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2004094292A (en) * | 2002-08-29 | 2004-03-25 | Ricoh Co Ltd | Character recognizing device, character recognizing method, and program used for executing the method |
CN1741035A (en) * | 2005-09-23 | 2006-03-01 | 清华大学 | Blocks letter Arabic character set text dividing method |
US20170147536A1 (en) * | 2011-05-23 | 2017-05-25 | Nuance Communications, Inc. | Text browsing, editing and correction methods for automotive applications |
CN105574523A (en) * | 2014-10-31 | 2016-05-11 | 株式会社东芝 | Character recognition device and method, image display device, and image retrieval device |
CN105989341A (en) * | 2015-02-17 | 2016-10-05 | 富士通株式会社 | Character recognition method and device |
CN105608454A (en) * | 2015-12-21 | 2016-05-25 | 上海交通大学 | Text structure part detection neural network based text detection method and system |
CN107016387A (en) * | 2016-01-28 | 2017-08-04 | 苏宁云商集团股份有限公司 | A kind of method and device for recognizing label |
CN107220641A (en) * | 2016-03-22 | 2017-09-29 | 华南理工大学 | A kind of multi-language text sorting technique based on deep learning |
CN105912225A (en) * | 2016-03-31 | 2016-08-31 | 北京奇虎科技有限公司 | Barrage display method and device based on encyclopedic items |
CN107180239A (en) * | 2017-06-09 | 2017-09-19 | 科大讯飞股份有限公司 | Line of text recognition methods and system |
CN108376244A (en) * | 2018-02-02 | 2018-08-07 | 北京大学 | A kind of recognition methods of text font in natural scene picture |
CN108304814A (en) * | 2018-02-08 | 2018-07-20 | 海南云江科技有限公司 | A kind of construction method and computing device of literal type detection model |
CN108664996A (en) * | 2018-04-19 | 2018-10-16 | 厦门大学 | A kind of ancient writing recognition methods and system based on deep learning |
Non-Patent Citations (4)
Title |
---|
EHSAN KIUMARSI 等: "A Hybrid Method for Text Line Extraction in Handwritten Document Images", 《2018 16TH INTERNATIONAL CONFERENCE ON FRONTIER IN HANDWRITING RECOGNITION》 * |
HAILIN YANG 等: "Recognition of Chinese Text in Historical Documents with Page-Level Annotations", 《2018 16TH INTERNATIONAL CONFERENCE ON FRONTIERS IN HANDWRITTEN RECOGNITION》 * |
张日培 等: "基于特征的藏文音节识别算法", 《电子设计工程》 * |
施银宝: "面向新闻字幕识别的视频文本定位与切分方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110309796A (en) * | 2019-07-05 | 2019-10-08 | 上海中通吉网络技术有限公司 | Face list recognition methods, device and equipment |
CN110503189A (en) * | 2019-08-02 | 2019-11-26 | 腾讯科技(深圳)有限公司 | A kind of data processing method and device |
CN110503189B (en) * | 2019-08-02 | 2021-10-08 | 腾讯科技(深圳)有限公司 | Data processing method and device |
CN111950397A (en) * | 2020-07-27 | 2020-11-17 | 腾讯科技(深圳)有限公司 | Text labeling method, device and equipment for image and storage medium |
CN111950397B (en) * | 2020-07-27 | 2021-10-22 | 腾讯科技(深圳)有限公司 | Text labeling method, device and equipment for image and storage medium |
CN111798542A (en) * | 2020-09-10 | 2020-10-20 | 北京易真学思教育科技有限公司 | Model training method, data processing device, model training apparatus, and storage medium |
CN111798542B (en) * | 2020-09-10 | 2020-12-22 | 北京易真学思教育科技有限公司 | Model training method, data processing device, model training apparatus, and storage medium |
CN112541438A (en) * | 2020-12-15 | 2021-03-23 | 深兰人工智能(深圳)有限公司 | Text recognition method and device |
CN113780131A (en) * | 2021-08-31 | 2021-12-10 | 众安在线财产保险股份有限公司 | Text image orientation recognition method and text content recognition method, device and equipment |
CN113780131B (en) * | 2021-08-31 | 2024-04-12 | 众安在线财产保险股份有限公司 | Text image orientation recognition method, text content recognition method, device and equipment |
CN114677691A (en) * | 2022-04-06 | 2022-06-28 | 北京百度网讯科技有限公司 | Text recognition method and device, electronic equipment and storage medium |
CN114677691B (en) * | 2022-04-06 | 2023-10-03 | 北京百度网讯科技有限公司 | Text recognition method, device, electronic equipment and storage medium |
Also Published As
Publication number | Publication date |
---|---|
CN109919037B (en) | 2021-09-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109919037A (en) | A kind of text positioning method and device, text recognition method and device | |
US11823046B2 (en) | Identifying subject matter of a digital image | |
Nishimoto et al. | Personal computer-based cephalometric landmark detection with deep learning, using cephalograms on the internet | |
US20210365717A1 (en) | Method and apparatus for segmenting a medical image, and storage medium | |
US10410353B2 (en) | Multi-label semantic boundary detection system | |
CN110334585A (en) | Table recognition method, apparatus, computer equipment and storage medium | |
CN108446621A (en) | Bank slip recognition method, server and computer readable storage medium | |
WO2021217857A1 (en) | Slice defect detection method and apparatus, and electronic device and readable storage medium | |
CN109754361A (en) | The anisotropic hybrid network of 3D: the convolution feature from 2D image is transmitted to 3D anisotropy volume | |
CN111801680B (en) | Systems and methods for visual search and augmented reality | |
CN112801146B (en) | Target detection method and system | |
US20160364633A1 (en) | Font recognition and font similarity learning using a deep neural network | |
CN114972548A (en) | Image processing system and method for displaying multiple images of a biological specimen | |
CN110427970A (en) | Image classification method, device, computer equipment and storage medium | |
CN110210286A (en) | Abnormality recognition method, device, equipment and storage medium based on eye fundus image | |
WO2021238548A1 (en) | Region recognition method, apparatus and device, and readable storage medium | |
CN107609575A (en) | Calligraphy evaluation method, calligraphy evaluating apparatus and electronic equipment | |
CN110223300A (en) | CT image abdominal multivisceral organ dividing method and device | |
CN109977762A (en) | A kind of text positioning method and device, text recognition method and device | |
CN112686850A (en) | Method and system for few-sample segmentation of CT image based on spatial position and prototype network | |
CN108875734A (en) | Liver canceration localization method, device and storage medium | |
CN110807139A (en) | Picture identification method and device, computer readable storage medium and computer equipment | |
CN109377441A (en) | Tongue with privacy protection function is as acquisition method and system | |
CN112464802A (en) | Automatic identification method and device for slide sample information and computer equipment | |
CN112749576A (en) | Image recognition method and device, computing equipment and computer storage medium |
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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |