CN112446369A - Bill processing method, system and storage medium - Google Patents
Bill processing method, system and storage medium Download PDFInfo
- Publication number
- CN112446369A CN112446369A CN202011220484.2A CN202011220484A CN112446369A CN 112446369 A CN112446369 A CN 112446369A CN 202011220484 A CN202011220484 A CN 202011220484A CN 112446369 A CN112446369 A CN 112446369A
- Authority
- CN
- China
- Prior art keywords
- image
- noise
- bill
- data set
- background
- 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
- 238000003672 processing method Methods 0.000 title claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000000034 method Methods 0.000 claims abstract description 27
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000004590 computer program Methods 0.000 claims description 10
- 230000003042 antagnostic effect Effects 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 5
- 238000013256 Gubra-Amylin NASH model Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 14
- 230000000694 effects Effects 0.000 description 12
- 230000006870 function Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000013075 data extraction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
Images
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
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
The invention provides a bill processing method, a bill processing system and a storage medium, wherein the bill processing method comprises the following steps: collecting a bill image of a bill; defining image noise of the bill image; generating a data set according to the bill image and the image noise; establishing and training a confrontation network model according to the data set; acquiring a to-be-processed bill image of a to-be-processed bill; and inputting the bill image to be processed into the confrontation network model to obtain the denoised target bill image. The bill processing method provided by the invention collects the bill image of the bill, defines the image noise of the bill image, generates a data set according to the bill image and the image noise, builds the countermeasure network model according to the data set, trains the countermeasure network model, processes the bill image to be processed by using the final countermeasure network model to obtain the denoised target bill image, and can process the noise problem of the overlapping or seal parts on the bill while processing the common noise.
Description
Technical Field
The invention relates to the field of image processing, in particular to a bill processing method, a bill processing system and a computer readable storage medium.
Background
In the related art, the noise is processed by using an image binarization method, and only simple noise can be processed; the deep learning detection method is used without specially processing noise, but has poor effects on problems of character overlapping, seals and the like.
Disclosure of Invention
The present invention is directed to solving at least one of the problems of the prior art or the related art.
To this end, a first aspect of the invention proposes a method for processing a document.
A second aspect of the invention provides a system for processing documents.
A third aspect of the invention provides a computer-readable storage medium.
In view of the above, an object of the present invention is to provide a method for processing a bill, including: collecting a bill image of a bill; defining image noise of the bill image; generating a data set according to the bill image and the image noise; establishing and training a confrontation network model according to the data set; acquiring a to-be-processed bill image of a to-be-processed bill; and inputting the bill image to be processed into the confrontation network model to obtain the denoised target bill image.
According to the bill processing method, the bill image of the bill is collected, the image noise of the bill image is defined, the data set is generated according to the bill image and the image noise, the countermeasure network model is built according to the data set and trained, the final countermeasure network model is used for processing the bill image to be processed, the denoised target bill image is obtained, common noise is processed, meanwhile, the noise problem of parts such as overlapping or seals on the bill can be processed, and compared with the prior art, the bill processing method has a better denoising effect.
In addition, the processing method of the bill in the above technical solution provided by the present invention may further have the following technical features:
in the above technical solution, the image noise includes: the first noise comprises any one of stamp noise, ruled line noise and original printed text noise or a combination of the stamp noise, the ruled line noise and the original printed text noise; and the second noise comprises any one or the combination of shadow noise, background shading noise, light reflection noise and miscellaneous line noise.
In the technical scheme, the first noise comprises seal noise, ruled line noise, original printed text noise and the like generated when the bill is generated, the second noise comprises image noise generated when a bill image to be processed is generated, such as shadow noise, background shading noise, reflection noise, miscellaneous line noise and the like, and the noise is simultaneously set as the image noise, so that when the bill image is processed for denoising, the original image can be kept and a better denoising effect can be realized.
In any of the above technical solutions, the step of generating a data set according to the document image and the image noise specifically includes:
removing the machine typewriter in the bill image to generate a background picture, wherein the background picture comprises image noise;
selecting a random position of a background picture to generate a first background picture, wherein the size of the first background picture is dxd, and d is a preset pixel value;
generating a second background image according to the first background image;
selecting a random text in a text character set;
drawing a random text on the first background image to obtain a third background image;
drawing a random text on the second background image to obtain a fourth background image;
combining the third background image and the fourth background image to obtain a noise image, wherein the size of the noise image is 2d multiplied by d;
repeating the steps for N times, and taking N noise pictures as a data set.
In the technical scheme, machine typewritten characters in the collected bill images are removed, image noise is kept at the same time, background pictures are generated, pictures with the size of dxd are randomly selected from the background pictures to serve as a first background picture, the first background picture is processed to generate a second background picture, and the size of the second background picture is also dxd.
Specifically, GB2312 is used as a text character set, random texts are randomly selected from the text character set, the length of each text is N, the random texts are respectively drawn in a first background image and a second background image to obtain a third background image and a fourth background image, the third background image and the fourth background image are spliced to generate a noise image with the image size of 2 dxd, the steps are repeated for N times to obtain a data set with the image size of N, the noise image obtained by the processing method is used as the data set, the third background image and the fourth background image have a comparative effect, and the image to be processed is better denoised by an confrontation network model trained and built by using the data set.
In any of the above technical solutions, the step of generating the second background map according to the first background map specifically includes:
acquiring three primary color parameter values of the first background image, and calculating the mean value of the three primary color parameter values of the first background image;
generating a blank map, wherein the size of the blank map is dxd, and d is a preset pixel value;
and setting the three primary color parameter values of the blank map as the average value of the three primary color parameter values of the first background map to obtain a second background map.
In the technical scheme, a blank picture is generated, the size of the blank picture is equal to that of a first background picture, R, G, B three primary color parameter values of the first background picture are extracted, R, G, B three primary color parameter values of the blank picture are set to be consistent with R, G, B three primary color parameter values of the first background picture, an obtained sub-picture is a second background picture, the R, G, B average value of the first background picture is used as R, G, B values of the second background picture, and noise of the picture is reduced in a mean filtering mode.
In any of the above technical solutions, the method further includes: the data set is divided into a first data set, a second data set and a third data set according to a preset proportion.
In the technical scheme, the data set is divided into three parts according to the proportion and used for building and training the confrontation network model.
Specifically, the preset ratio is 8:1: 1.
In the technical scheme, the data sets are grouped according to a preset ratio of 8:1:1 and are respectively used as different sub data sets.
In any of the above technical solutions, the step of building and training the confrontation network model according to the data set specifically includes: building a confrontation network model; taking the first data set as a training set for training the antagonistic network model; taking the second data set as a verification set for adjusting parameters of the countermeasure network model; and taking the third data set as a test set for evaluating the antagonistic network model.
In the technical scheme, a confrontation network model is built, and a first data set, a second data set and a third data set grouped according to a preset ratio of 8:1:1 are used respectively. The method comprises the steps of acquiring a first data set, acquiring a second data set, acquiring a third data set, acquiring a training set, acquiring a third data set, and acquiring a third data set, wherein the first data set is used as the training set, training an antagonistic network model, adjusting parameters and primarily evaluating the model by using the second data set with a relatively small amount as a verification set, and finally evaluating the generalization capability of the final model until a preset effect is achieved.
In any of the above technical solutions, the countermeasure network model includes: a CGAN model, a GAN model, a DCGAN model, or a WGAN model.
In the technical scheme, the countermeasure network model can be a CGAN model, a GAN model, a DCGAN model or a WGAN model, wherein the most preferred model is the CGAN model.
In view of the above, a second aspect of the present invention provides a ticket processing system, including a memory, in which a computer program executable on a processor is stored; the processor realizes the steps of the processing method of the bill in any technical scheme when executing the computer program, so that the processing system of the bill comprises all the beneficial effects of the processing method of the bill in any technical scheme.
In view of this, the third aspect of the present invention provides a computer readable storage medium, on which a computer program is stored, the computer program, when being executed by a processor, implements the processing method of the bill according to any one of the above technical solutions, and therefore, the computer readable storage medium includes all the advantages of the processing method of the bill according to any one of the above technical solutions.
Drawings
The above and/or additional aspects and advantages of the present invention will become apparent and readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings of which:
FIG. 1 is a flow diagram illustrating a method of processing documents according to an embodiment of the present invention;
FIG. 2 is a flow chart showing a specific example of the step S106 according to an embodiment of the present invention;
FIG. 3 is a schematic diagram illustrating a specific flowchart of step S106 according to an embodiment of the present invention;
FIG. 4 is a detailed flowchart of step S108 in one embodiment of the present invention;
FIG. 5 shows a schematic diagram of a third background diagram in an embodiment of the invention;
FIG. 6 shows a schematic diagram of a fourth background diagram in an embodiment of the invention;
FIG. 7a shows a partial schematic view of a document image to be processed according to the present invention;
FIG. 7b is a partial schematic view of the note image to be processed in FIG. 7a after denoising;
FIG. 8a shows a partial schematic view of yet another ticket image to be processed of the present invention;
FIG. 8b is a partial schematic view of the note image to be processed in FIG. 8a after denoising;
FIG. 9a shows a partial schematic view of yet another ticket image to be processed of the present invention;
FIG. 9b is a partial schematic view of the note image to be processed in FIG. 9a after denoising;
FIG. 10a shows a partial schematic view of yet another ticket image to be processed of the present invention;
FIG. 10b is a partial schematic view of the note image to be processed in FIG. 10a after denoising;
FIG. 11a shows a partial schematic view of yet another ticket image to be processed of the present invention;
FIG. 11b is a partial schematic view of the note image to be processed in FIG. 11a after denoising.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a more particular description of the invention will be rendered by reference to the appended drawings. It should be noted that the embodiments and features of the embodiments of the present application may be combined with each other without conflict.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present invention, however, the present invention may be practiced in other ways than those specifically described herein, and therefore the scope of the present invention is not limited by the specific embodiments disclosed below.
A method, an apparatus, and a storage medium for processing a ticket according to some embodiments of the present invention are described below with reference to fig. 1 to 11.
Example one
As shown in fig. 1, according to an embodiment of the present invention, a method for processing a bill is provided, which includes the following specific steps:
step S102, collecting bill images of bills;
step S104, defining image noise of the bill image;
step S106, generating a data set according to the bill image and the image noise;
step S108, establishing and training a confrontation network model according to the data set;
step S110, acquiring a to-be-processed bill image of a to-be-processed bill;
and step S112, inputting the bill image to be processed into the countermeasure network model to obtain the denoised target bill image.
In the embodiment, the bill image of the bill is collected, the image noise of the bill image is defined, the data set is generated according to the bill image and the image noise, the countermeasure network model is built according to the data set and is trained, the final countermeasure network model is used for processing the bill image to be processed, the denoised target bill image is obtained, meanwhile, the noise generated when the image is formed in the image to be processed and the error noise generated when the bill image is printed are overcome, and compared with the prior art, the bill image processing method has a better denoising effect.
Further, in step S102, the bill image of the real bill can be acquired by taking a picture of the scanned bill.
Further, in step S104, the image noise includes a first noise and a second noise, where the first noise includes any one or a combination of stamp noise, ruled line noise, and original printed text noise, and the second noise includes any one or a combination of shadow noise, background shading noise, reflection noise, and miscellaneous line noise.
Specifically, the first noise refers to a stamp on the bill, and the stamp has a large shape and generally overlaps with machine-typing of the bill or a form of the bill itself, or overlaps with the machine-typing of the bill due to a printing error, or overlaps with a printed text of the bill itself. The second noise refers to that the image of the bill generates common noise such as reflection, shadow and the like in the process of photographing or scanning, or background shading noise of the bill itself, or miscellaneous line noise caused by printing errors or other conditions.
In the embodiment, the noise is set as the image noise at the same time, so that when the note image is subjected to denoising processing, the original image can be kept and a better denoising effect can be realized.
Further, as shown in fig. 2, the step S106 of generating a data set according to the document image and the image noise specifically includes:
step S202, removing the typewriter symbols in the bill image to generate a background picture, wherein the background picture comprises image noise;
step S204, selecting a random position of a background picture to generate a first background picture, wherein the size of the first background picture is dxd, and d is a preset pixel value;
step S206, generating a second background image according to the first background image;
step S208, selecting a random text in the text character set;
step S210, drawing a random text on the first background image to obtain a third background image;
step S212, drawing a random text on the second background image to obtain a fourth background image;
step S214, combining the third background image and the fourth background image to obtain a noise image, wherein the size of the noise image is 2d multiplied by d;
and step S216, repeating the steps N times, and taking N noise pictures as a data set.
Specifically, in this embodiment, step S202 may specifically be to use a picture editing tool to perform a process of removing machine-printed characters on the ticket image, and the processed background picture still includes image noise, that is, the processed background picture still retains noise of overlapping or reflective portions on the ticket image.
In step S204, any position in the background picture is randomly selected, and the background picture is cut into a d × d sub-picture, where d is a preset pixel size, and if d can be 256 or 1024, the cut sub-picture is used as the first background picture.
In step S206, the first background map is processed to generate a second background map, and the specific processing method includes:
acquiring three primary color parameter values of the first background image, and calculating the mean value of the three primary color parameter values of the first background image;
generating a blank map, wherein the size of the blank map is dxd, and d is a preset pixel value;
and setting the three primary color parameter values of the blank map as the average value of the three primary color parameter values of the first background map to obtain a second background map.
Specifically, a blank image with the size of d × d is generated by using a picture editing tool, the size of the blank image is consistent with that of the first background image, R, G, B values of the first background image are extracted, the average value of the R, G, B values is set as the R, G, B value of the blank image, and the modified blank image is the second background image.
In step S208, GB2312 is selected as the text character set, and a text of length n is randomly fetched in the text character set as a random text.
As shown in fig. 5 and 6, in steps S210 and S212, the random text may be drawn at a random position of the first background map by using a function of an Image module in a PIL (Python Image Library) to obtain a third background map, and the random text may be drawn at the same position on the second background map by using the same method to generate a fourth background map.
Specifically, the random texts may be arranged at regular intervals in a row or column manner, and the coverage of the texts includes the whole area of the first background image or the second background image.
In step S214, the third background image and the fourth background image are stitched into a stitched image (noise image) with a size of 2d × d, which can be used as data in a data set for lapping and training the confrontation network model.
In step S216, the above steps S202 to S214 are repeated, and each time one data is obtained, the size of the data set after repeating N times is N.
In addition, the process of generating the data set in step S106 is achieved by obtaining the data set in large quantities through a software system, so that a process of obtaining a large number of data sets in a short time is achieved, and the data set can be labeled manually, so that the effect of manually labeling the data set is more accurate, and technicians can select the data set according to circumstances without limitation.
As shown in fig. 3, the step S106 further includes:
s218 divides the data set into a first data set, a second data set and a third data set according to a preset ratio.
In the technical scheme, a data set is divided into three parts in proportion and used for building and training a confrontation network model, wherein the proportion of a first data set is large.
Specifically, the preset ratio is 8:1:1, and when the data sets are grouped, the ratio of the first data set to the second data set or the third data set is large, such as 8:1:1 and 6:2:2, which are feasible.
Further, as shown in fig. 4, step S108 builds and trains a confrontation network model according to the data set, which specifically includes:
s402, building a confrontation network model;
s404, taking the first data set as a training set for training the antagonistic network model; taking the second data set as a verification set for adjusting parameters of the countermeasure network model; and taking the third data set as a test set for evaluating the antagonistic network model.
Specifically, in this embodiment, S402 builds a countermeasure network model, where the countermeasure network model may be a CGAN model, a GAN model, a DCGAN model, or a WGAN model.
Taking a CGAN (Conditional generated adaptive Nets) model as an example, when the CGAN model is selected, the target function of the CGAN model is:
wherein G is a generator of the CGAN model, D is a discriminator of the CGAN model, x is a generated graph and a real graph, y is a label (i.e., extra information), and z is noise.
In step S404, the first data set, the second data set, and the third data set grouped by the preset ratio in the data set are respectively used as a training set, a verification set, and a test set.
Specifically, a training set is used as input of a constructed CGAN model, the CGAN model is trained to obtain a trained model, a verification set is used as a model evaluation data set, parameters of the trained model are adjusted according to model effects to obtain an optimized model, finally, the processing effect of the optimized model is evaluated through a test set, namely unprocessed and processed data images are compared, and the model with the best effect is selected as a final confrontation network model.
Further, in steps S110 and S112, a to-be-processed bill image of the to-be-processed bill is acquired, and the to-be-processed bill image is input into the countermeasure network model, so as to obtain a denoised target bill image.
Specifically, a bill image to be processed is collected, and the bill image to be processed is input into the generated countermeasure network model, so as to obtain a denoised target bill image, as shown in fig. 7 to 11, where real _ a is a local schematic diagram of the bill image to be processed, fake _ B is a local schematic diagram of the processed target bill image, fig. 7a is a local schematic diagram of the bill image to be processed before denoising, fig. 7B is a corresponding local schematic diagram of the denoised target bill, where shading noise is processed, fig. 8a is a local schematic diagram of the bill image to be processed before denoising, fig. 8B is a corresponding local schematic diagram of the denoised target bill, where seal noise is processed, fig. 9a is a local schematic diagram of the bill image to be processed before denoising, and fig. 9B is a corresponding local schematic diagram of the denoised target bill, where ruled line noise is processed, fig. 10a is a partial schematic view of a note image to be processed before denoising, fig. 10b is a corresponding partial schematic view of a denoised target note, wherein an overlapped part of an original printed text is processed, fig. 11a is a partial schematic view of a note image to be processed before denoising, and fig. 11b is a corresponding partial schematic view of a denoised target note, wherein the original printed text is processed, contrast is visible, the processed note image has a significant noise processing effect on a ground tint, a seal or a transverse line on the note image, and characters on the note are still clear and distinguishable while noise is removed, so that data extraction during note processing in a later period is facilitated.
The bill processing method provided by the embodiment can be used for denoising bills of various types, removing noise in the bills and simultaneously keeping due characters, can be applied to various automatic bill identification systems, improves the identification rate and the anti-interference performance of the bills, and ensures the reliability of bill result identification.
Example two
According to yet another embodiment of the present invention, there is provided a processing system for a bill, including: a memory and a processor. The storage stores a computer program capable of running on the processor, and the processor executes the computer program to implement the steps of the method for processing the ticket according to any of the above embodiments.
EXAMPLE III
According to an embodiment of the invention, a computer-readable storage medium is proposed, the computer program realizing the steps of the method of processing a ticket according to any of the above-described embodiments when being executed by a processor, the computer-readable storage medium thus comprising all the advantageous effects of the method of processing a ticket according to any of the above-described embodiments.
The computer-readable storage medium includes a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk, and the like.
In the description of the present invention, the terms "plurality" or "a plurality" refer to two or more, and unless otherwise specifically defined, the terms "upper", "lower", and the like indicate orientations or positional relationships based on the orientations or positional relationships illustrated in the drawings, and are only for convenience in describing the present invention and simplifying the description, but do not indicate or imply that the referred device or element must have a specific orientation, be constructed in a specific orientation, and be operated, and thus, should not be construed as limiting the present invention; the terms "connected," "mounted," "secured," and the like are to be construed broadly and include, for example, fixed connections, removable connections, or integral connections; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present invention can be understood by those skilled in the art according to specific situations.
In the description of the present invention, the description of the terms "one embodiment," "some embodiments," "specific embodiments," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In the present invention, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A method of processing a document, comprising:
collecting a bill image of a bill;
defining image noise of the bill image;
generating a data set from the document image and the image noise;
establishing and training a confrontation network model according to the data set;
acquiring a to-be-processed bill image of a to-be-processed bill;
and inputting the bill image to be processed into the confrontation network model to obtain a denoised target bill image.
2. The method of processing a document according to claim 1, wherein the image noise comprises:
the first noise comprises any one of stamp noise, ruled line noise and original printed text noise or a combination of the stamp noise, the ruled line noise and the original printed text noise;
a second noise, wherein the second noise comprises any one of shadow noise, background shading noise, reflective noise and miscellaneous line noise or a combination thereof.
3. The method according to claim 2, wherein the step of generating a data set from the document image and the image noise comprises:
removing machine-typed characters in the bill image to generate a background picture, wherein the background picture comprises the image noise;
selecting a random position of the background picture to generate a first background picture, wherein the size of the first background picture is dxd, and d is a preset pixel value;
generating a second background image according to the first background image;
selecting a random text in a text character set;
drawing the random text on the first background image to obtain a third background image;
drawing the random text on the second background image to obtain a fourth background image;
combining the third background image and the fourth background image to obtain a noise image, wherein the size of the noise image is 2d multiplied by d;
repeating the steps for N times, and taking N noise pictures as a data set.
4. The method for processing a bill according to claim 3, wherein the step of generating a second background image from the first background image specifically comprises:
acquiring three primary color parameter values of the first background image, and calculating the mean value of the three primary color parameter values of the first background image;
generating a blank map, wherein the size of the blank map is dxd, and d is a preset pixel value;
setting the three primary color parameter values of the blank map as the average value of the three primary color parameter values of the first background map to obtain the second background map.
5. The method of processing documents according to claim 3, further comprising:
and dividing the data set into a first data set, a second data set and a third data set according to a preset proportion.
6. The method of processing a bill according to claim 5,
the preset ratio is 8:1: 1.
7. The method for processing the bill according to claim 5, wherein the step of building and training the countermeasure network model according to the data set specifically comprises:
building a confrontation network model;
using the first data set as a training set for training the confrontation network model;
using the second data set as a verification set for adjusting parameters of the confrontation network model;
and taking the third data set as a test set for evaluating the confrontation network model.
8. The processing method of tickets according to any one of claims 1 to 7, characterized in that said antagonistic network model comprises: a CGAN model, a GAN model, a DCGAN model, or a WGAN model.
9. A system for processing documents, comprising:
a memory having stored thereon a computer program operable on the processor;
a processor implementing the steps of the method of processing a ticket according to any one of claims 1 to 8 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of processing a ticket according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011220484.2A CN112446369A (en) | 2020-11-05 | 2020-11-05 | Bill processing method, system and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011220484.2A CN112446369A (en) | 2020-11-05 | 2020-11-05 | Bill processing method, system and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112446369A true CN112446369A (en) | 2021-03-05 |
Family
ID=74735816
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011220484.2A Pending CN112446369A (en) | 2020-11-05 | 2020-11-05 | Bill processing method, system and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112446369A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112949523A (en) * | 2021-03-11 | 2021-06-11 | 兴业银行股份有限公司 | Method and system for extracting key information from identity card image picture |
| CN114863416A (en) * | 2022-07-07 | 2022-08-05 | 合肥高维数据技术有限公司 | Training data generation method and system for general text OCR |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198154A (en) * | 2018-03-19 | 2018-06-22 | 中山大学 | Image de-noising method, device, equipment and storage medium |
| CN109615589A (en) * | 2018-10-31 | 2019-04-12 | 北京达佳互联信息技术有限公司 | Remove the method, apparatus and terminal device of picture noise |
| CN110288547A (en) * | 2019-06-27 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image denoising model |
| CN111461979A (en) * | 2020-03-30 | 2020-07-28 | 招商局金融科技有限公司 | Verification code image denoising and identifying method, electronic device and storage medium |
-
2020
- 2020-11-05 CN CN202011220484.2A patent/CN112446369A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108198154A (en) * | 2018-03-19 | 2018-06-22 | 中山大学 | Image de-noising method, device, equipment and storage medium |
| CN109615589A (en) * | 2018-10-31 | 2019-04-12 | 北京达佳互联信息技术有限公司 | Remove the method, apparatus and terminal device of picture noise |
| CN110288547A (en) * | 2019-06-27 | 2019-09-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating image denoising model |
| CN111461979A (en) * | 2020-03-30 | 2020-07-28 | 招商局金融科技有限公司 | Verification code image denoising and identifying method, electronic device and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 陈琦;潘伟民;: "基于自编码器的图像去噪设计与实现", 新疆师范大学学报(自然科学版), no. 02, pages 85 - 90 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112949523A (en) * | 2021-03-11 | 2021-06-11 | 兴业银行股份有限公司 | Method and system for extracting key information from identity card image picture |
| CN114863416A (en) * | 2022-07-07 | 2022-08-05 | 合肥高维数据技术有限公司 | Training data generation method and system for general text OCR |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| JP6139396B2 (en) | Method and program for compressing binary image representing document | |
| CN104794421B (en) | A kind of positioning of QR codes and recognition methods | |
| US20070253040A1 (en) | Color scanning to enhance bitonal image | |
| CN108596166A (en) | A kind of container number identification method based on convolutional neural networks classification | |
| KR101737338B1 (en) | System and method for clean document reconstruction from annotated document images | |
| US6393150B1 (en) | Region-based image binarization system | |
| EP2270746B1 (en) | Method for detecting alterations in printed document using image comparison analyses | |
| US20030198386A1 (en) | System and method for identifying and extracting character strings from captured image data | |
| CN103824373B (en) | A kind of bill images amount of money sorting technique and system | |
| CN112446369A (en) | Bill processing method, system and storage medium | |
| CN110298353B (en) | Character recognition method and system | |
| US20040042677A1 (en) | Method and apparatus to enhance digital image quality | |
| EP0949579A2 (en) | Multiple size reductions for image segmentation | |
| US8456711B2 (en) | SUSAN-based corner sharpening | |
| CN111553363B (en) | End-to-end seal identification method and system | |
| CN110598566A (en) | Image processing method, device, terminal and computer readable storage medium | |
| Gede et al. | Automatic vectorisation of old maps using qgis-tools, possibilities and challenges | |
| CN114519788A (en) | Image processing method, image processing device, electronic equipment and computer readable storage medium | |
| EP1296283A2 (en) | Half-tone dot elimination method and system thereof | |
| CN116050379A (en) | Document comparison method and storage medium | |
| CN114926829A (en) | Certificate detection method and device, electronic equipment and storage medium | |
| CN112861836B (en) | Text image processing method, text and card image quality evaluation method and device | |
| JP2005242600A (en) | Pattern recognition processing apparatus, method, and program | |
| CN202058178U (en) | Character and image correction device | |
| US7567725B2 (en) | Edge smoothing filter for character recognition |
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 |