CN107292823A - Electronic installation, the method for invoice classification and computer-readable recording medium - Google Patents
Electronic installation, the method for invoice classification and computer-readable recording medium Download PDFInfo
- Publication number
- CN107292823A CN107292823A CN201710715451.7A CN201710715451A CN107292823A CN 107292823 A CN107292823 A CN 107292823A CN 201710715451 A CN201710715451 A CN 201710715451A CN 107292823 A CN107292823 A CN 107292823A
- Authority
- CN
- China
- Prior art keywords
- invoice
- picture
- classification
- training
- sample
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000009434 installation Methods 0.000 title claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 77
- 238000012937 correction Methods 0.000 claims abstract description 37
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000013527 convolutional neural network Methods 0.000 claims description 38
- 230000017105 transposition Effects 0.000 claims description 16
- 238000001514 detection method Methods 0.000 claims description 10
- 230000009466 transformation Effects 0.000 claims description 8
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 description 16
- 238000004891 communication Methods 0.000 description 6
- 230000001537 neural effect Effects 0.000 description 4
- 238000010586 diagram Methods 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000006870 function Effects 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 238000004140 cleaning Methods 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000003475 lamination Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000008878 coupling Effects 0.000 description 1
- 238000010168 coupling process Methods 0.000 description 1
- 238000005859 coupling reaction Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/60—Rotation of whole images or parts thereof
- G06T3/608—Rotation of whole images or parts thereof by skew deformation, e.g. two-pass or three-pass rotation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
Abstract
The method and computer-readable recording medium classified the present invention relates to a kind of electronic installation, invoice, the electronic installation includes memory and the processor being connected with the memory, be stored with the identifying system that can be run on the processor in the memory, and following steps are realized when the identifying system is by the computing device：After pending invoice picture is received, Slant Rectify is carried out to the invoice picture using predetermined correction rule；The invoice picture recognition model generated using training in advance carries out classification identification to the invoice picture after Slant Rectify, and exports classification recognition result.The present invention quickly can carry out angle correction and classification to the invoice picture of batch, time saving and energy saving, improve the efficiency of processing.
Description
Technical field
The present invention relates to communication technical field, more particularly to the method and computer of the classification of a kind of electronic installation, invoice can
Read storage medium.
Background technology
At present, for needing the operation of the centralized financial data of progress, for example, life insurance Claims Resolution, employee's expense reimbursement
Deng before the invoice picture uploaded to batch carries out centralized traffic processing, it usually needs manual type is in advance to invoice picture
Carry out angle correction and invoice picture is classified, for carrying out centralized traffic processing, centralized traffic processing includes being sent out
Ticket is adjusted, invoice information typing etc., existing use manual type the invoice picture that batch is uploaded is carried out angle correction and
The scheme of classification wastes time and energy, inefficiency.
The content of the invention
The method and computer-readable recording medium classified it is an object of the invention to provide a kind of electronic installation, invoice,
It is intended to the quick invoice picture to batch and carries out angle correction and classification, improves treatment effeciency.
To achieve the above object, the present invention provides a kind of electronic installation, the electronic installation include memory and with it is described
Be stored with the identifying system that can be run on the processor, the identification in the processor of memory connection, the memory
Following steps are realized when system is by the computing device：
S1, after pending invoice picture is received, is inclined using predetermined correction rule to the invoice picture
Tiltedly correction；
S2, the invoice picture recognition model generated using training in advance carries out classification knowledge to the invoice picture after Slant Rectify
Not, and classification recognition result is exported；
The predetermined correction rule includes：
Isolated using Hough transformation Hough probabilistic algorithm in the invoice picture less than or equal to the first preset length
Short straight line section；
Based on respectively isolate short straight line section x coordinate or y-coordinate by the short straight line respectively isolated section divide into several classes；
Of a sort short straight line section will be belonged to obtain and each mesh as a target class straight line, and using least square method
Mark the similar long straightway of class straight line；
Calculate the slope of each long straightway, and each long straightway slope median and average, compare calculating
The median of the slope gone out and the size of average adjust invoice picture to determine smaller according to the smaller determined
Inclination angle.
Preferably, the invoice picture recognition model is depth convolutional neural networks model, the depth convolutional Neural net
Network model includes algorithm of target detection.
Preferably, when the identifying system is by the computing device, following steps are also realized：
Prepare the invoice for being labeled with corresponding classification of the first predetermined number respectively for each default invoice picture classification
Picture sample, the invoice picture sample is divided into the training subset of the first ratio and the checking subset of the second ratio；
The corresponding second predetermined number certificate picture sample of each default invoice picture classification is obtained, by the certificate figure
Piece sample is divided into the training subset of the first ratio and the checking subset of the second ratio；
Picture sample in all training subsets is mixed to obtain training set, by the picture in all checking subsets
Sample is mixed to be verified collection；
The depth convolutional neural networks model is trained using the training set；
Utilize the accuracy rate of the depth convolutional neural networks model after the checking collection checking training；
If the accuracy rate is more than or equal to default accuracy rate, training terminates, with the depth convolutional Neural after training
Network model as the invoice picture recognition model in the step S2, or, if the accuracy rate is less than default accuracy rate,
Increase the quantity of the corresponding certificate picture sample of each default invoice picture classification, to re-start training.
Preferably, when the identifying system is by the computing device, following steps are also realized：
Before the depth convolutional neural networks model is trained, the picture sample of analyzing and training collection and the picture sample of checking collection
This markup information, the picture sample of markup information mistake is cleared up；
The transposition of remaining picture sample after being cleared up according to the position analysis of the depth-width ratio information of invoice picture and seal
Situation, and to occurring the carry out upset adjustment of transposition.
To achieve the above object, the present invention also provides a kind of method of invoice classification, and the method for the invoice classification includes：
S1, after pending invoice picture is received, is inclined using predetermined correction rule to the invoice picture
Tiltedly correction；
S2, the invoice picture recognition model generated using training in advance carries out classification knowledge to the invoice picture after Slant Rectify
Not, and classification recognition result is exported；
The predetermined correction rule includes：
Isolated using Hough transformation Hough probabilistic algorithm in the invoice picture less than or equal to the first preset length
Short straight line section；
Based on respectively isolate short straight line section x coordinate or y-coordinate by the short straight line respectively isolated section divide into several classes；
Of a sort short straight line section will be belonged to obtain and each mesh as a target class straight line, and using least square method
Mark the similar long straightway of class straight line；
Calculate the slope of each long straightway, and each long straightway slope median and average, compare calculating
The median of the slope gone out and the size of average adjust invoice picture to determine smaller according to the smaller determined
Inclination angle.
Preferably, the invoice picture recognition model is depth convolutional neural networks model, the depth convolutional Neural net
Network model includes algorithm of target detection.
Preferably, also include before the step S2：
Prepare the invoice for being labeled with corresponding classification of the first predetermined number respectively for each default invoice picture classification
Picture sample, the invoice picture sample is divided into the training subset of the first ratio and the checking subset of the second ratio；
The corresponding second predetermined number certificate picture sample of each default invoice picture classification is obtained, by the certificate figure
Piece sample is divided into the training subset of the first ratio and the checking subset of the second ratio；
Picture sample in all training subsets is mixed to obtain training set, by the picture in all checking subsets
Sample is mixed to be verified collection；
The depth convolutional neural networks model is trained using the training set；
Utilize the accuracy rate of the depth convolutional neural networks model after the checking collection checking training；
If the accuracy rate is more than or equal to default accuracy rate, training terminates, with the depth convolutional Neural after training
Network model as the invoice picture recognition model in the step S2, or, if the accuracy rate is less than default accuracy rate,
Increase the quantity of the corresponding certificate picture sample of each default invoice picture classification, to re-start training.
Preferably, the method for the invoice classification also includes：
Before the depth convolutional neural networks model is trained, the picture sample of analyzing and training collection and the picture sample of checking collection
This markup information, the picture sample of markup information mistake is cleared up；
The transposition of remaining picture sample after being cleared up according to the position analysis of the depth-width ratio information of invoice picture and seal
Situation, and to occurring the carry out upset adjustment of transposition.
The present invention also provides the identification that is stored with a kind of computer-readable recording medium, the computer-readable recording medium
System, the identifying system realizes the method for above-mentioned invoice classification when being executed by processor the step of.
The beneficial effects of the invention are as follows：The present invention is advised after pending invoice picture is received using predetermined correction
Then to the invoice picture carry out Slant Rectify, afterwards, using the invoice picture recognition model of training in advance to Slant Rectify after
Invoice picture carry out classification identification, and export classification recognition result, batch uploaded compared to existing use manual type
Invoice picture carry out the scheme of angle correction and classification, the present invention can be quickly to the invoice picture of batch progress angle correction
And classification, it is time saving and energy saving, improve the efficiency of business processing.
Brief description of the drawings
Fig. 1 is the optional application environment schematic diagram of each embodiment one of the invention；
Fig. 2 is the schematic flow sheet for the embodiment of method one that invoice of the present invention is classified；
Fig. 3 is the schematic diagram of correction rule predetermined shown in Fig. 2.
Embodiment
In order to make the purpose , technical scheme and advantage of the present invention be clearer, it is right below in conjunction with drawings and Examples
The present invention is further elaborated.It should be appreciated that specific embodiment described herein is only to explain the present invention, not
For limiting the present invention.Based on the embodiment in the present invention, those of ordinary skill in the art are not before creative work is made
The every other embodiment obtained is put, the scope of protection of the invention is belonged to.
It should be noted that the description for being related to " first ", " second " etc. in the present invention is only used for describing purpose, and can not
It is interpreted as indicating or implies its relative importance or the implicit quantity for indicating indicated technical characteristic.Thus, define " the
One ", at least one this feature can be expressed or be implicitly included to the feature of " second ".In addition, the skill between each embodiment
Art scheme can be combined with each other, but must can be implemented as basis with those of ordinary skill in the art, when technical scheme
With reference to occur it is conflicting or when can not realize it will be understood that the combination of this technical scheme is not present, also not in application claims
Protection domain within.
As shown in fig.1, being the application environment schematic diagram of the preferred embodiment of the method for invoice classification of the present invention.The application
Environment schematic includes electronic installation 1 and terminal device 2.What electronic installation 1 can be adapted to by network, near-field communication technology etc.
Technology carries out data interaction with terminal device 2.
The terminal device 2, which includes, but not limited to any one, to pass through keyboard, mouse, remote control, touch with user
The mode such as plate or voice-operated device carries out the electronic product of man-machine interaction, for example, personal computer, tablet personal computer, smart mobile phone,
Personal digital assistant (Personal Digital Assistant, PDA), game machine, IPTV (Internet
Protocol Television, IPTV), intellectual Wearable, the movable equipment of guider etc., or such as
The fixed terminal of digital TV, desktop computer, notebook, server etc..
The electronic installation 1 be it is a kind of can according to the instruction for being previously set or storing, it is automatic carry out numerical computations and/
Or the equipment of information processing.The electronic installation 1 can be computer, can also be single network server, multiple networks clothes
The server group of business device composition or the cloud being made up of a large amount of main frames or the webserver based on cloud computing, wherein cloud computing
It is one kind of Distributed Calculation, a super virtual computer being made up of the computer collection of a group loose couplings.
In the present embodiment, electronic installation 1 may include, but be not limited only to, and can be in communication with each other connection by system bus
Memory 11, processor 12, network interface 13, memory 11 are stored with the identifying system that can be run on the processor 12.Need
, it is noted that Fig. 1 illustrate only the electronic installation 1 with component 11-13, it should be understood that being not required for implementing all
The component shown, what can be substituted implements more or less components.
Wherein, storage device 11 includes internal memory and the readable storage medium storing program for executing of at least one type.Inside save as electronic installation 1
Operation provides caching；Readable storage medium storing program for executing can be if flash memory, hard disk, multimedia card, card-type memory are (for example, SD or DX storages
Device etc.), random access storage device (RAM), static random-access memory (SRAM), read-only storage (ROM), electric erasable can
Program read-only memory (EEPROM), programmable read only memory (PROM), magnetic storage, disk, CD etc. it is non-volatile
Property storage medium.In certain embodiments, readable storage medium storing program for executing can be the internal storage unit of electronic installation 1, for example the electricity
The hard disk of sub-device 1；In further embodiments, the non-volatile memory medium can also be the external storage of electronic installation 1
The plug-in type hard disk being equipped with equipment, such as electronic installation 1, intelligent memory card (Smart Media Card, SMC), safe number
Word (Secure Digital, SD) blocks, flash card (Flash Card) etc..In the present embodiment, the readable storage of storage device 11
Medium is generally used for storing in the operating system for being installed on electronic installation 1 and types of applications software, such as one embodiment of the invention
The program code of identifying system etc..In addition, storage device 11 can be also used for temporarily storing and export or will be defeated
The Various types of data gone out.
The processor 12 can be in certain embodiments central processing unit (Central ProcessingUnit,
CPU), controller, microcontroller, microprocessor or other data processing chips.The processor 12 is generally used for controlling the electricity
The overall operation of sub-device 1, for example, perform the control and processing related to the terminal device 2 progress data interaction or communication
Deng.In the present embodiment, the processor 12 is used to run the program code stored in the memory 11 or processing data, example
Such as run identifying system.
The network interface 13 may include radio network interface or wired network interface, and the network interface 13 is generally used for
Communication connection is set up between the electronic installation 1 and other electronic equipments.In the present embodiment, network interface 13 is mainly used in electricity
Sub-device 1 is connected with one or more terminal devices 2, and data are set up between electronic installation 1 and one or more terminal devices 2
Transmission channel and communication connection.
The identifying system is stored in memory 11, including at least one be stored in it is computer-readable in memory 11
Instruction, at least one computer-readable instruction can be performed by processor device 12, the method to realize each embodiment of the application；With
And, at least one computer-readable instruction is different according to the function that its each several part is realized, can be divided into different logic moulds
Block.
In one embodiment, following steps are realized when above-mentioned identifying system is performed by the processor 12：
Step S1, after pending invoice picture is received, is entered using predetermined correction rule to the invoice picture
Line tilt is corrected；
In the present embodiment, after the pending invoice picture of batch is received, using predetermined correction rule to invoice
Picture carries out Slant Rectify, wherein predetermined correction rule is including a variety of：
In one embodiment, predetermined correction rule can be：The inclined angle of invoice picture is obtained, it is inclined based on this
The angle correction invoice picture；
In another embodiment, in order to more accurately be corrected to invoice picture, predetermined correction rule can be：Profit
The short straight line for being less than or equal to the first preset length in the invoice picture is isolated with the probabilistic algorithm of Hough transformation (Hough)
Section, wherein, the probabilistic algorithm of Hough transformation can from the black white image of invoice picture detection of straight lines (line segment), the first default length
Degree for example, 3mm, the short straight line section isolated is as more as possible.X coordinate or y-coordinate based on the short straight line section respectively isolated will be each
The short straight line section divide into several classes isolated, specifically, determines that level inclination is less than or equal to from the short straight line section isolated
The short straight line section of predetermined angle (such as predetermined angle is 5 degree or 10 degree), x coordinate value difference in the straightway determined is less than
Short straight line section equal to predetermined threshold value (such as predetermined threshold value be 0.5mm) is divided into a class, until by all short straight lines isolated
If section is divided into Ganlei, or, the level inclination determined is less than or equal to y-coordinate value in the short straight line of predetermined angle section and differed
Short straight line section less than or equal to predetermined threshold value (such as predetermined threshold value be 0.5mm) is divided into a class, until by it is all isolate it is short
If straightway is divided into Ganlei.Of a sort short straight line section will be belonged to as a target class straight line, and obtained using least square method
The long straightway similar to each target class straight line is taken, wherein, least square method finds each by minimizing the quadratic sum of error
The optimal function matching (i.e. long straightway) of individual target class straight line.Calculate the slope of each long straightway, and each long straight line
Section slope median and average, compare the median of the slope calculated and the size of average, with determine median and
The less slope of average, and the inclination angle of invoice picture is adjusted according to the less slope determined, in other embodiments,
The minimum corresponding slope of median or the minimum corresponding slope of average can be determined, with the minimum corresponding slope of median or
The minimum corresponding slope of person's average adjusts the inclination angle of invoice picture, in addition, when adjusting the inclination angle of invoice picture according to slope, obtaining
The angle at the corresponding inclination angle of the slope is taken, then invoice picture opposite direction is adjusted to the angle at the inclination angle.
Step S2, the invoice picture recognition model generated using training in advance carries out class to the invoice picture after Slant Rectify
Do not recognize, and export classification recognition result.
In the present embodiment, the invoice picture recognition model of training in advance generation is depth convolutional neural networks model, wherein,
When carrying out classification identification to the invoice picture after Slant Rectify using depth convolutional neural networks model, it is preferable that using can be with
The algorithm of target detection based on depth convolutional neural networks chosen in the environment of CaffeNet is to the invoice after Slant Rectify
Picture carries out classification identification, it is of course also possible to classification identification is carried out to the invoice picture after Slant Rectify using other algorithms, this
Excessive restriction is not done in place.In addition, the classification of invoice is more, for example the classification for the invoice of hospital includes outpatient service invoice and is in hospital
Invoice etc., after the invoice picture after to Slant Rectify carries out classification identification, exports the corresponding classification of each invoice picture.
Preferably, algorithm of target detection includes architecture and auxiliary framework, specifically, including 1 input layer, 13 volumes
Lamination, 5 pond layers, 2 full articulamentums, 1 classification layer is as shown in table 1 below：
| Layer Name | Batch Size | Kernel Size | Stride Size | Pad Size |
| Input | 128 | N/A | N/A | N/A |
| Conv1 | 128 | 3 | 1 | 1 |
| Conv2 | 128 | 3 | 1 | 1 |
| MaxPool1 | 128 | 2 | 2 | 0 |
| Conv3 | 128 | 3 | 1 | 1 |
| Conv4 | 128 | 3 | 1 | 1 |
| MaxPool2 | 128 | 2 | 2 | 0 |
| Conv5 | 128 | 3 | 1 | 1 |
| Conv6 | 128 | 3 | 1 | 1 |
| Conv7 | 128 | 3 | 1 | 1 |
| MaxPool3 | 128 | 2 | 2 | 0 |
| Conv8 | 128 | 3 | 1 | 1 |
| Conv9 | 128 | 3 | 1 | 1 |
| Conv10 | 128 | 3 | 1 | 1 |
| MaxPool4 | 128 | 2 | 2 | 0 |
| Conv11 | 128 | 3 | 1 | 1 |
| Conv12 | 128 | 3 | 1 | 1 |
| Conv13 | 128 | 3 | 1 | 1 |
| MaxPool5 | 128 | 2 | 2 | 0 |
| Fc1 | 4096 | 1 | 1 | 0 |
| Fc2 | 2048 | 1 | 1 | 0 |
| Softmax | 3 | N/A | N/A | N/A |
Table 1
Wherein, Layer Name row represent each layer of title, and Batch Size represent the input picture number of current layer,
Kernel Size represent that the yardstick of current layer convolution kernel (for example, Kernel Size can be equal to 3, represents the yardstick of convolution kernel
For 3x 3), Stride Size represent the moving step length of convolution kernel, that is, finish and next convolution position is moved to after a convolution
The distance put, Pad Size represent the size to the image completion among current network layer.Input table shows input layer, Conv tables
Show convolutional layer, Conv1 represents the 1st convolutional layer, and MaxPool represents maximum pond layer, and MaxPool1 represents the 1st maximum
Pond layer, Fc represents full articulamentum, and Fc1 represents the 1st full articulamentum, and Softmax represents Softmax graders.
Compared with prior art, the present embodiment utilizes predetermined correction rule after pending invoice picture is received
To the invoice picture carry out Slant Rectify, afterwards, using the invoice picture recognition model of training in advance to Slant Rectify after
Invoice picture carries out classification identification, and exports classification recognition result, and batch is uploaded compared to existing use manual type
Invoice picture carries out the scheme of angle correction and classification, and the present embodiment quickly can carry out angle correction to the invoice picture of batch
And classification, it is time saving and energy saving, improve the efficiency of business processing.
In a preferred embodiment, on the basis of above-described embodiment, the identifying system is held by the processor 12
Before the identification of row classification, following steps are also realized：
Prepare the invoice for being labeled with corresponding classification of the first predetermined number respectively for each default invoice picture classification
Picture sample, the invoice picture sample is divided into the training subset of the first ratio and the checking subset of the second ratio, wherein, in advance
If invoice picture classification include it is a variety of, such as including outpatient service class invoice and class invoice in hospital, the first predetermined number is, for example,
1000, the first ratio is, for example, 75%, and the second ratio is, for example, 25%, wherein, the first ratio is less than with the second ratio sum
Equal to 1.
The corresponding second predetermined number certificate picture sample of each default invoice picture classification is obtained, by the certificate figure
Piece sample is divided into the training subset of the first ratio and the checking subset of the second ratio, wherein, each default invoice picture classification
Corresponding certificate picture sample is the invoice picture of the corresponding standard of invoice picture classification, and the invoice picture of the standard is just to put
, the invoice picture that markup information does not go wrong, the second predetermined number is, for example, 1000, and the first ratio is, for example, 75%,
Second ratio is, for example, 25%, wherein, the first ratio is less than or equal to 1 with the second ratio sum.
Picture sample in all training subsets is mixed to obtain training set, by the picture in all checking subsets
Sample is mixed to be verified collection, and the depth convolutional neural networks model is trained using the training set, using described
The accuracy rate of depth convolutional neural networks model after checking collection checking training, if the accuracy rate is more than or equal to default standard
True rate (default accuracy rate is, for example, 0.98), then training terminates, using the depth convolutional neural networks model after training as described
Invoice picture recognition model in step S2, or, if the accuracy rate is less than default accuracy rate, increase each default hair
The quantity of the corresponding certificate picture sample of ticket picture classification, to re-start training.
In a preferred embodiment, on the basis of above-described embodiment, in order to improve training depth convolutional neural networks
The efficiency of model, the identifying system is performed by the processor 12 before training depth convolutional neural networks model, is also realized
Following steps：
Before the depth convolutional neural networks model is trained, the picture sample of analyzing and training collection and the picture sample of checking collection
This markup information, the picture sample of markup information mistake is cleared up；
The transposition of remaining picture sample after being cleared up according to the position analysis of the depth-width ratio information of invoice picture and seal
Situation, and to occurring the carry out upset adjustment of transposition.
Wherein, before the depth convolutional neural networks model is trained, the picture sample of analyzing and training collection and checking collection
Whether the markup information of picture sample, the key position information for for example analyzing picture sample lacks or beyond whole pictures scope,
And whether seal labeling position is located at the data of the obvious marking error such as invoice center, if there is the picture sample of above mentioned problem
This, then cleared up or abandoned to it, to ensure that the markup information of picture sample is accurate.
For remaining picture sample after cleaning, according to its depth-width ratio information and the position judgment picture sample of seal
Transposition situation, and upset adjustment is done to the picture sample for occurring transposition：When depth-width ratio is more than 1, illustrate the high wide top of invoice picture
, if seal position does rotated ninety degrees clockwise processing to picture sample, if seal position exists on the left of hair picture sample
On the right side of invoice picture, then 90 degree of processing of rotate counterclockwise are done to invoice image；When depth-width ratio is less than 1, illustrate picture sample
High width is not overturned, if seal position is on the downside of invoice picture, turn clockwise 180 degree of processing to invoice image, if
Seal position is not processed then on the upside of invoice picture.
In addition, being modified to the labeled data of the picture sample by overturning adjustment, the mark number of each picture sample
According to the positional information for the rectangle frame for referring to outlining this picture sample, with the top left co-ordinate of this rectangle frame (xmin,
Ymin) represented with bottom right angular coordinate (xmax, ymax) four number, if xmax<Xmin, then overturn the two position, to y-coordinate
Same processing is done, to ensure max>min.
As shown in Fig. 2 Fig. 2 is the schematic flow sheet for the embodiment of method one that invoice of the present invention is classified, invoice classification
Method comprises the following steps：
Step S1, after pending invoice picture is received, is entered using predetermined correction rule to the invoice picture
Line tilt is corrected；
In the present embodiment, after the pending invoice picture of batch is received, using predetermined correction rule to invoice
Picture carries out Slant Rectify, wherein predetermined correction rule is including a variety of：
In one embodiment, predetermined correction rule can be：The inclined angle of invoice picture is obtained, it is inclined based on this
The angle correction invoice picture；
In another embodiment, in order to more accurately be corrected to invoice picture, with reference to refering to Fig. 3, predetermined correction
Rule can be：Isolated using Hough transformation Hough probabilistic algorithm and be less than or equal to the first default length in the invoice picture
The short straight line section of degree, wherein, the probabilistic algorithm of Hough transformation can from the black white image of invoice picture detection of straight lines (line segment),
First preset length is, for example, 3mm, and the short straight line section isolated is as more as possible.X coordinate based on the short straight line section respectively isolated
Or the short straight line respectively isolated section divide into several classes specifically, is determined that level is inclined by y-coordinate from the short straight line section isolated
Angle is less than or equal to the short straight line section of predetermined angle (such as predetermined angle is 5 degree or 10 degree), by x coordinate in the straightway determined
The short straight line section that value difference is less than or equal to predetermined threshold value (such as predetermined threshold value be 0.5mm) is divided into a class, until by all separation
If the short straight line section gone out is divided into Ganlei, or, the level inclination determined is less than or equal to y in the short straight line of predetermined angle section
The short straight line section that coordinate value difference is less than or equal to predetermined threshold value (such as predetermined threshold value is 0.5mm) is divided into a class, until will be all
If the short straight line section isolated is divided into Ganlei.Of a sort short straight line section will be belonged to as a target class straight line, and using most
Small square law obtains the long straightway similar to each target class straight line, wherein, least square method is by minimizing the flat of error
Side and the optimal function matching (i.e. long straightway) for finding each target class straight line.The slope of each long straightway is calculated, and
The median and average of the slope of each long straightway, compare the median of the slope calculated and the size of average, to determine
Go out the less slope of median and average, and the inclination angle of invoice picture is adjusted according to the less slope determined, at other
In embodiment, the minimum corresponding slope of median or the minimum corresponding slope of average can also be determined, it is minimum with median
Corresponding slope or the minimum corresponding slope of average adjust the inclination angle of invoice picture, in addition, adjusting invoice figure according to slope
During the inclination angle of piece, the angle at the corresponding inclination angle of the slope is obtained, invoice picture opposite direction is then adjusted into the angle at the inclination angle i.e.
Can.
Step S2, the invoice picture recognition model generated using training in advance carries out class to the invoice picture after Slant Rectify
Do not recognize, and export classification recognition result.
In the present embodiment, the invoice picture recognition model of training in advance generation is depth convolutional neural networks model, wherein,
When carrying out classification identification to the invoice picture after Slant Rectify using depth convolutional neural networks model, it is preferable that using can be with
The algorithm of target detection based on depth convolutional neural networks chosen in the environment of CaffeNet is to the invoice after Slant Rectify
Picture carries out classification identification, it is of course also possible to classification identification is carried out to the invoice picture after Slant Rectify using other algorithms, this
Excessive restriction is not done in place.In addition, the classification of invoice is more, for example the classification for the invoice of hospital includes outpatient service invoice and is in hospital
Invoice etc., after the invoice picture after to Slant Rectify carries out classification identification, exports the corresponding classification of each invoice picture.
Preferably, algorithm of target detection includes architecture and auxiliary framework, specifically, including 1 input layer, 13 volumes
Lamination, 5 pond layers, 2 full articulamentums, 1 classification layer, as shown in Table 1 above, here is omitted.
Wherein, Layer Name row represent each layer of title, and Batch Size represent the input picture number of current layer,
Kernel Size represent that the yardstick of current layer convolution kernel (for example, Kernel Size can be equal to 3, represents the yardstick of convolution kernel
For 3x 3), Stride Size represent the moving step length of convolution kernel, that is, finish and next convolution position is moved to after a convolution
The distance put, Pad Size represent the size to the image completion among current network layer.Input table shows input layer, Conv tables
Show convolutional layer, Conv1 represents the 1st convolutional layer, and MaxPool represents maximum pond layer, and MaxPool1 represents the 1st maximum
Pond layer, Fc represents full articulamentum, and Fc1 represents the 1st full articulamentum, and Softmax represents Softmax graders.
The present embodiment is entered after pending invoice picture is received using predetermined correction rule to the invoice picture
Line tilt is corrected, afterwards, and classification is carried out to the invoice picture after Slant Rectify using the invoice picture recognition model of training in advance
Identification, and classification recognition result is exported, angle is carried out to the invoice picture that batch is uploaded compared to existing use manual type
Correction and the scheme of classification, the present embodiment quickly can carry out angle correction and classification to the invoice picture of batch, time saving and energy saving,
Improve the efficiency of business processing.
In a preferred embodiment, on the basis of above-mentioned Fig. 2 embodiment, also include before the step S2：
Prepare the invoice for being labeled with corresponding classification of the first predetermined number respectively for each default invoice picture classification
Picture sample, the invoice picture sample is divided into the training subset of the first ratio and the checking subset of the second ratio, wherein, in advance
If invoice picture classification include it is a variety of, such as including outpatient service class invoice and class invoice in hospital, the first predetermined number is, for example,
1000, the first ratio is, for example, 75%, and the second ratio is, for example, 25%, wherein, the first ratio is less than with the second ratio sum
Equal to 1.
The corresponding second predetermined number certificate picture sample of each default invoice picture classification is obtained, by the certificate figure
Piece sample is divided into the training subset of the first ratio and the checking subset of the second ratio, wherein, each default invoice picture classification
Corresponding certificate picture sample is the invoice picture of the corresponding standard of invoice picture classification, and the invoice picture of the standard is just to put
, the invoice picture that markup information does not go wrong, the second predetermined number is, for example, 1000, and the first ratio is, for example, 75%,
Second ratio is, for example, 25%, wherein, the first ratio is less than or equal to 1 with the second ratio sum.
Picture sample in all training subsets is mixed to obtain training set, by the picture in all checking subsets
Sample is mixed to be verified collection, and the depth convolutional neural networks model is trained using the training set, using described
The accuracy rate of depth convolutional neural networks model after checking collection checking training, if the accuracy rate is more than or equal to default standard
True rate (default accuracy rate is, for example, 0.98), then training terminates, using the depth convolutional neural networks model after training as described
Invoice picture recognition model in step S2, or, if the accuracy rate is less than default accuracy rate, increase each default hair
The quantity of the corresponding certificate picture sample of ticket picture classification, to re-start training.
In a preferred embodiment, on the basis of above-described embodiment, in order to improve training depth convolutional neural networks
The efficiency of model, the method for the invoice classification also includes：
Before the depth convolutional neural networks model is trained, the picture sample of analyzing and training collection and the picture sample of checking collection
This markup information, the picture sample of markup information mistake is cleared up；
The transposition of remaining picture sample after being cleared up according to the position analysis of the depth-width ratio information of invoice picture and seal
Situation, and to occurring the carry out upset adjustment of transposition.
Wherein, before the depth convolutional neural networks model is trained, the picture sample of analyzing and training collection and checking collection
Whether the markup information of picture sample, the key position information for for example analyzing picture sample lacks or beyond whole pictures scope,
And whether seal labeling position is located at the data of the obvious marking error such as invoice center, if there is the picture sample of above mentioned problem
This, then cleared up or abandoned to it, to ensure that the markup information of picture sample is accurate.
For remaining picture sample after cleaning, according to its depth-width ratio information and the position judgment picture sample of seal
Transposition situation, and upset adjustment is done to the picture sample for occurring transposition：When depth-width ratio is more than 1, illustrate the high wide top of invoice picture
, if seal position does rotated ninety degrees clockwise processing to picture sample, if seal position exists on the left of hair picture sample
On the right side of invoice picture, then 90 degree of processing of rotate counterclockwise are done to invoice image；When depth-width ratio is less than 1, illustrate picture sample
High width is not overturned, if seal position is on the downside of invoice picture, turn clockwise 180 degree of processing to invoice image, if
Seal position is not processed then on the upside of invoice picture.
In addition, being modified to the labeled data of the picture sample by overturning adjustment, the mark number of each picture sample
According to the positional information for the rectangle frame for referring to outlining this picture sample, with the top left co-ordinate of this rectangle frame (xmin,
Ymin) represented with bottom right angular coordinate (xmax, ymax) four number, if xmax<Xmin, then overturn the two position, to y-coordinate
Same processing is done, to ensure max>min.
The present invention also provides the identification that is stored with a kind of computer-readable recording medium, the computer-readable recording medium
System, the identifying system realizes the method for above-mentioned invoice classification when being executed by processor the step of.
The embodiments of the present invention are for illustration only, and the quality of embodiment is not represented.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can add the mode of required general hardware platform to realize by software, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Understood based on such, technical scheme is substantially done to prior art in other words
Going out the part of contribution can be embodied in the form of software product, and the computer software product is stored in a storage medium
In (such as ROM/RAM, magnetic disc, CD), including some instructions are to cause a station terminal equipment (can be mobile phone, computer, clothes
It is engaged in device, air conditioner, or network equipment etc.) perform method described in each embodiment of the invention.
The preferred embodiments of the present invention are these are only, are not intended to limit the scope of the invention, it is every to utilize this hair
Equivalent structure or equivalent flow conversion that bright specification and accompanying drawing content are made, or directly or indirectly it is used in other related skills
Art field, is included within the scope of the present invention.
Claims (9)
1. a kind of electronic installation, it is characterised in that the electronic installation includes memory and the processing being connected with the memory
Be stored with the identifying system that can be run on the processor in device, the memory, and the identifying system is by the processor
Following steps are realized during execution：
S1, after pending invoice picture is received, enters line tilt to the invoice picture using predetermined correction rule and rectifys
Just；
S2, the invoice picture recognition model generated using training in advance carries out classification identification to the invoice picture after Slant Rectify,
And export classification recognition result；
The predetermined correction rule includes：
Isolated using Hough transformation Hough probabilistic algorithm short straight less than or equal to the first preset length in the invoice picture
Line segment；
Based on respectively isolate short straight line section x coordinate or y-coordinate by the short straight line respectively isolated section divide into several classes；
Of a sort short straight line section will be belonged to obtain and each target class as a target class straight line, and using least square method
The similar long straightway of straight line；
Calculate the slope of each long straightway, and each long straightway slope median and average, compare what is calculated
The median of slope and the size of average adjust inclining for invoice picture to determine smaller according to the smaller determined
Angle.
2. electronic installation according to claim 1, it is characterised in that the invoice picture recognition model is depth convolution god
Through network model, the depth convolutional neural networks model includes algorithm of target detection.
3. electronic installation according to claim 2, it is characterised in that when the identifying system is by the computing device,
Also realize following steps：
Prepare the invoice picture for being labeled with corresponding classification of the first predetermined number respectively for each default invoice picture classification
Sample, the invoice picture sample is divided into the training subset of the first ratio and the checking subset of the second ratio；
The corresponding second predetermined number certificate picture sample of each default invoice picture classification is obtained, by the certificate picture sample
Originally it is divided into the training subset of the first ratio and the checking subset of the second ratio；
Picture sample in all training subsets is mixed to obtain training set, by the picture sample in all checking subsets
Mixed to be verified collection；
The depth convolutional neural networks model is trained using the training set；
Utilize the accuracy rate of the depth convolutional neural networks model after the checking collection checking training；
If the accuracy rate is more than or equal to default accuracy rate, training terminates, with the depth convolutional neural networks after training
Model as the invoice picture recognition model in the step S2, or, if the accuracy rate is less than default accuracy rate, increase
The quantity of the corresponding certificate picture sample of each default invoice picture classification, to re-start training.
4. electronic installation according to claim 3, it is characterised in that when the identifying system is by the computing device,
Also realize following steps：
Before the depth convolutional neural networks model is trained, the picture sample that the picture sample of analyzing and training collection and checking collect
Markup information, the picture sample of markup information mistake is cleared up；
The transposition situation of remaining picture sample after being cleared up according to the position analysis of the depth-width ratio information of invoice picture and seal,
And to occurring the carry out upset adjustment of transposition.
5. a kind of method of invoice classification, it is characterised in that the method for the invoice classification includes：
S1, after pending invoice picture is received, enters line tilt to the invoice picture using predetermined correction rule and rectifys
Just；
S2, the invoice picture recognition model generated using training in advance carries out classification identification to the invoice picture after Slant Rectify,
And export classification recognition result；
The predetermined correction rule includes：
Isolated using Hough transformation Hough probabilistic algorithm short straight less than or equal to the first preset length in the invoice picture
Line segment；
Based on respectively isolate short straight line section x coordinate or y-coordinate by the short straight line respectively isolated section divide into several classes；
Of a sort short straight line section will be belonged to obtain and each target class as a target class straight line, and using least square method
The similar long straightway of straight line；
Calculate the slope of each long straightway, and each long straightway slope median and average, compare what is calculated
The median of slope and the size of average adjust inclining for invoice picture to determine smaller according to the smaller determined
Angle.
6. the method for invoice classification according to claim 5, it is characterised in that the invoice picture recognition model is depth
Convolutional neural networks model, the depth convolutional neural networks model includes algorithm of target detection.
7. the method for invoice classification according to claim 6, it is characterised in that also include before the step S2：
Prepare the invoice picture for being labeled with corresponding classification of the first predetermined number respectively for each default invoice picture classification
Sample, the invoice picture sample is divided into the training subset of the first ratio and the checking subset of the second ratio；
The corresponding second predetermined number certificate picture sample of each default invoice picture classification is obtained, by the certificate picture sample
Originally it is divided into the training subset of the first ratio and the checking subset of the second ratio；
Picture sample in all training subsets is mixed to obtain training set, by the picture sample in all checking subsets
Mixed to be verified collection；
The depth convolutional neural networks model is trained using the training set；
Utilize the accuracy rate of the depth convolutional neural networks model after the checking collection checking training；
If the accuracy rate is more than or equal to default accuracy rate, training terminates, with the depth convolutional neural networks after training
Model as the invoice picture recognition model in the step S2, or, if the accuracy rate is less than default accuracy rate, increase
The quantity of the corresponding certificate picture sample of each default invoice picture classification, to re-start training.
8. the method for invoice classification according to claim 7, it is characterised in that the method for the invoice classification also includes：
Before the depth convolutional neural networks model is trained, the picture sample that the picture sample of analyzing and training collection and checking collect
Markup information, the picture sample of markup information mistake is cleared up；
The transposition situation of remaining picture sample after being cleared up according to the position analysis of the depth-width ratio information of invoice picture and seal,
And to occurring the carry out upset adjustment of transposition.
9. a kind of computer-readable recording medium, it is characterised in that be stored with identification system on the computer-readable recording medium
System, the method that the invoice as any one of claim 5 to 8 is classified is realized when the identifying system is executed by processor
Step.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710715451.7A CN107292823A (en) | 2017-08-20 | 2017-08-20 | Electronic installation, the method for invoice classification and computer-readable recording medium |
| PCT/CN2017/108762 WO2019037259A1 (en) | 2017-08-20 | 2017-10-31 | Electronic device, method and system for categorizing invoices, and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710715451.7A CN107292823A (en) | 2017-08-20 | 2017-08-20 | Electronic installation, the method for invoice classification and computer-readable recording medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107292823A true CN107292823A (en) | 2017-10-24 |
Family
ID=60106382
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710715451.7A Pending CN107292823A (en) | 2017-08-20 | 2017-08-20 | Electronic installation, the method for invoice classification and computer-readable recording medium |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN107292823A (en) |
| WO (1) | WO2019037259A1 (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108229299A (en) * | 2017-10-31 | 2018-06-29 | 北京市商汤科技开发有限公司 | The recognition methods of certificate and device, electronic equipment, computer storage media |
| CN108805285A (en) * | 2018-05-30 | 2018-11-13 | 济南浪潮高新科技投资发展有限公司 | A kind of convolutional neural networks pond unit design method |
| WO2019037259A1 (en) * | 2017-08-20 | 2019-02-28 | 平安科技（深圳）有限公司 | Electronic device, method and system for categorizing invoices, and computer-readable storage medium |
| CN109657694A (en) * | 2018-10-26 | 2019-04-19 | 平安科技（深圳）有限公司 | Picture automatic classification method, device and computer readable storage medium |
| CN109740417A (en) * | 2018-10-29 | 2019-05-10 | 深圳壹账通智能科技有限公司 | Invoice type recognition methods, device, storage medium and computer equipment |
| CN109815954A (en) * | 2019-01-31 | 2019-05-28 | 科大讯飞股份有限公司 | Correction for direction method, apparatus, equipment and the storage medium of VAT invoice image |
| CN110348441A (en) * | 2019-07-10 | 2019-10-18 | 深圳市华云中盛科技有限公司 | VAT invoice recognition methods, device, computer equipment and storage medium |
| CN110490267A (en) * | 2019-08-23 | 2019-11-22 | 四川长虹电器股份有限公司 | A kind of bill method for sorting based on deep learning |
| CN111178345A (en) * | 2019-05-20 | 2020-05-19 | 京东方科技集团股份有限公司 | Bill analysis method, bill analysis device, computer equipment and medium |
| CN111241966A (en) * | 2020-01-06 | 2020-06-05 | 广东工业大学 | Invoice target area positioning method, device, equipment and medium |
| CN111507181A (en) * | 2020-03-11 | 2020-08-07 | 平安科技（深圳）有限公司 | Bill image correction method and device and computer equipment |
| CN111639516A (en) * | 2019-03-01 | 2020-09-08 | 埃森哲环球解决方案有限公司 | Analysis platform based on machine learning |
| CN112686255A (en) * | 2020-12-31 | 2021-04-20 | 深圳赛安特技术服务有限公司 | Seal identification method, device, equipment and storage medium based on Hough transform |
| CN113011249A (en) * | 2021-01-29 | 2021-06-22 | 招商银行股份有限公司 | Bill auditing method, device, equipment and storage medium |
| CN113011249B (en) * | 2021-01-29 | 2024-05-28 | 招商银行股份有限公司 | Bill auditing method, device, equipment and storage medium |
Families Citing this family (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059689B (en) * | 2019-03-19 | 2024-05-03 | 平安科技（深圳）有限公司 | Sample set construction method, device, computer equipment and storage medium |
| CN111753842B (en) * | 2020-06-28 | 2024-02-27 | 中国银行股份有限公司 | Method and device for detecting text region of bill |
| CN112396048B (en) * | 2020-11-17 | 2023-09-29 | 中国平安人寿保险股份有限公司 | Picture information extraction method and device, computer equipment and storage medium |
| CN112287898A (en) * | 2020-11-26 | 2021-01-29 | 深源恒际科技有限公司 | Method and system for evaluating text detection quality of image |
| CN113158895B (en) * | 2021-04-20 | 2023-11-14 | 北京中科江南信息技术股份有限公司 | Bill identification method and device, electronic equipment and storage medium |
| CN113538291B (en) * | 2021-08-02 | 2024-05-14 | 广州广电运通金融电子股份有限公司 | Card image inclination correction method, device, computer equipment and storage medium |
| CN113963147B (en) * | 2021-09-26 | 2023-09-15 | 西安交通大学 | Key information extraction method and system based on semantic segmentation |
| CN116204769B (en) * | 2023-03-06 | 2023-12-05 | 深圳市乐易网络股份有限公司 | Data cleaning method, system and storage medium based on data classification and identification |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101625760A (en) * | 2009-07-28 | 2010-01-13 | 谭洪舟 | Method for correcting certificate image inclination |
| CN106909882A (en) * | 2017-01-16 | 2017-06-30 | 广东工业大学 | A kind of face identification system and method for being applied to security robot |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103955669B (en) * | 2014-04-10 | 2017-02-15 | 西安理工大学 | Invoice date positioning method based on segmentation Hough transformation straight line detection |
| CN105808610B (en) * | 2014-12-31 | 2019-12-20 | 中国科学院深圳先进技术研究院 | Internet picture filtering method and device |
| CN107292823A (en) * | 2017-08-20 | 2017-10-24 | 平安科技（深圳）有限公司 | Electronic installation, the method for invoice classification and computer-readable recording medium |
-
2017
- 2017-08-20 CN CN201710715451.7A patent/CN107292823A/en active Pending
- 2017-10-31 WO PCT/CN2017/108762 patent/WO2019037259A1/en active Application Filing
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101625760A (en) * | 2009-07-28 | 2010-01-13 | 谭洪舟 | Method for correcting certificate image inclination |
| CN106909882A (en) * | 2017-01-16 | 2017-06-30 | 广东工业大学 | A kind of face identification system and method for being applied to security robot |
Non-Patent Citations (5)
| Title |
|---|
| 吕刚 等: "一种改进的深度神经网络在小图像分类中的应用研究", 《计算机应用与软件》 * |
| 李琥: "银行单据类型识别研究与实现", 《中国优秀博硕士学位论文全文数据库（硕士） 信息科技辑》 * |
| 杜刚: "银行票据识别系统研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 郭斯羽 等: "结合Hough变换与改进最小二乘法的直线检测", 《计算机科学》 * |
| 韩梦迪: "基于BP神经网络的银行票据识别", 《信息通信》 * |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019037259A1 (en) * | 2017-08-20 | 2019-02-28 | 平安科技（深圳）有限公司 | Electronic device, method and system for categorizing invoices, and computer-readable storage medium |
| CN108229299A (en) * | 2017-10-31 | 2018-06-29 | 北京市商汤科技开发有限公司 | The recognition methods of certificate and device, electronic equipment, computer storage media |
| CN108229299B (en) * | 2017-10-31 | 2021-02-26 | 北京市商汤科技开发有限公司 | Certificate identification method and device, electronic equipment and computer storage medium |
| CN108805285A (en) * | 2018-05-30 | 2018-11-13 | 济南浪潮高新科技投资发展有限公司 | A kind of convolutional neural networks pond unit design method |
| CN109657694A (en) * | 2018-10-26 | 2019-04-19 | 平安科技（深圳）有限公司 | Picture automatic classification method, device and computer readable storage medium |
| CN109740417A (en) * | 2018-10-29 | 2019-05-10 | 深圳壹账通智能科技有限公司 | Invoice type recognition methods, device, storage medium and computer equipment |
| CN109740417B (en) * | 2018-10-29 | 2023-05-16 | 深圳壹账通智能科技有限公司 | Invoice type identification method, invoice type identification device, storage medium and computer equipment |
| CN109815954A (en) * | 2019-01-31 | 2019-05-28 | 科大讯飞股份有限公司 | Correction for direction method, apparatus, equipment and the storage medium of VAT invoice image |
| CN111639516A (en) * | 2019-03-01 | 2020-09-08 | 埃森哲环球解决方案有限公司 | Analysis platform based on machine learning |
| CN111639516B (en) * | 2019-03-01 | 2023-09-15 | 埃森哲环球解决方案有限公司 | Analysis platform based on machine learning |
| CN111178345A (en) * | 2019-05-20 | 2020-05-19 | 京东方科技集团股份有限公司 | Bill analysis method, bill analysis device, computer equipment and medium |
| CN110348441B (en) * | 2019-07-10 | 2021-08-17 | 深圳市华云中盛科技股份有限公司 | Value-added tax invoice identification method and device, computer equipment and storage medium |
| CN110348441A (en) * | 2019-07-10 | 2019-10-18 | 深圳市华云中盛科技有限公司 | VAT invoice recognition methods, device, computer equipment and storage medium |
| CN110490267A (en) * | 2019-08-23 | 2019-11-22 | 四川长虹电器股份有限公司 | A kind of bill method for sorting based on deep learning |
| CN111241966A (en) * | 2020-01-06 | 2020-06-05 | 广东工业大学 | Invoice target area positioning method, device, equipment and medium |
| CN111507181A (en) * | 2020-03-11 | 2020-08-07 | 平安科技（深圳）有限公司 | Bill image correction method and device and computer equipment |
| CN111507181B (en) * | 2020-03-11 | 2023-05-26 | 平安科技（深圳）有限公司 | Correction method and device for bill image and computer equipment |
| CN112686255A (en) * | 2020-12-31 | 2021-04-20 | 深圳赛安特技术服务有限公司 | Seal identification method, device, equipment and storage medium based on Hough transform |
| CN113011249A (en) * | 2021-01-29 | 2021-06-22 | 招商银行股份有限公司 | Bill auditing method, device, equipment and storage medium |
| CN113011249B (en) * | 2021-01-29 | 2024-05-28 | 招商银行股份有限公司 | Bill auditing method, device, equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2019037259A1 (en) | 2019-02-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107292823A (en) | Electronic installation, the method for invoice classification and computer-readable recording medium | |
| CN107798299B (en) | Bill information identification method, electronic device and readable storage medium | |
| CN107194398A (en) | Car damages recognition methods and the system at position | |
| CN110188755B (en) | Image recognition method and device and computer readable storage medium | |
| CN112016438B (en) | Method and system for identifying certificate based on graph neural network | |
| WO2019104879A1 (en) | Information recognition method for form-type image, electronic device and readable storage medium | |
| CN108229299A (en) | The recognition methods of certificate and device, electronic equipment, computer storage media | |
| CN103975342A (en) | Systems and methods for mobile image capture and processing | |
| US10133859B2 (en) | Managing registration of user identity using handwriting | |
| JP5154292B2 (en) | Information management system, form definition management server, and information management method | |
| CN103544475A (en) | Method and system for recognizing layout types | |
| CN111178357B (en) | License plate recognition method, system, device and storage medium | |
| CN113011144A (en) | Form information acquisition method and device and server | |
| CN105512657A (en) | Character recognition method and apparatus | |
| CN106293412A (en) | Icon adjustment method, icon adjust system and electronic installation | |
| CN113343740A (en) | Table detection method, device, equipment and storage medium | |
| CN110598826A (en) | Electronic certificate classified display method, device and system and computer equipment | |
| CN113673519A (en) | Character recognition method based on character detection model and related equipment thereof | |
| CN113012075A (en) | Image correction method and device, computer equipment and storage medium | |
| CN112418206B (en) | Picture classification method based on position detection model and related equipment thereof | |
| CN113255767A (en) | Bill classification method, device, equipment and storage medium | |
| CN109584165A (en) | A kind of antidote of digital picture, device, medium and electronic equipment | |
| Lu et al. | A partition approach for the restoration of camera images of planar and curled document | |
| CN112651399A (en) | Method for detecting same-line characters in oblique image and related equipment thereof | |
| CN112835497A (en) | Method and device for quickly entering electronic whiteboard and 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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20171024 |