CN112200177A - Single number identification method and device based on bill picking scanning piece big data - Google Patents
Single number identification method and device based on bill picking scanning piece big data Download PDFInfo
- Publication number
- CN112200177A CN112200177A CN202010702824.9A CN202010702824A CN112200177A CN 112200177 A CN112200177 A CN 112200177A CN 202010702824 A CN202010702824 A CN 202010702824A CN 112200177 A CN112200177 A CN 112200177A
- Authority
- CN
- China
- Prior art keywords
- bill
- classification
- scanning piece
- character
- models
- 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 37
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 42
- 238000003709 image segmentation Methods 0.000 claims abstract description 14
- 238000005516 engineering process Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 32
- 238000000605 extraction Methods 0.000 claims description 12
- 238000002372 labelling Methods 0.000 claims description 12
- 238000005520 cutting process Methods 0.000 claims description 9
- 230000011218 segmentation Effects 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 4
- 230000000877 morphologic effect Effects 0.000 claims description 4
- 238000012216 screening Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 6
- 230000009286 beneficial effect Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000010191 image analysis Methods 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/14—Image acquisition
- G06V30/148—Segmentation of character regions
- G06V30/153—Segmentation of character regions using recognition of characters or words
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
-
- 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
- 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
- G06V10/225—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 based on a marking or identifier characterising the area
Abstract
The invention discloses a single number identification method and a single number identification device based on bill picking scanning piece big data, which belong to the technical field of big data processing, and comprise a plurality of Convolutional Neural Network (CNN) models, a voting mechanism and an image segmentation technology, wherein the number of the Convolutional Neural Network (CNN) models is multiple, the plurality of Convolutional Neural Network (CNN) models establish a classification identification model based on the bill picking scanning piece big data, and the voting mechanism carries out bill picking piece classification identification. In the invention, the classification recognition of the ship company to which the bill raising scanning piece belongs is carried out by adopting multiple models, the classification result is obtained according to the majority voting rule, the classification recognition effect can be fully improved, an iterative updating mechanism of the recognition model can be established for the newly added bill raising scanning piece, the performance of the model is effectively kept to be improved, the bill raising scanning piece is accurately recognized, the renaming of the bill raising scanning piece is automatically carried out according to the recognition result, the direct calling of other services is convenient, and the bill raising scanning piece can be directly output as a file to be filed.
Description
Technical Field
The invention belongs to the technical field of big data processing, and particularly relates to a single number identification method and device based on big data of a bill of lading scanned part.
Background
With the rapid development of foreign trade exports, overseas e-commerce and cross-border logistics, the trend of increasing maritime transportation as one of the mainstream cargo transportation modes is presented, and a maritime bill of lading as an important basis of maritime transportation is a key element in links such as inquiry, payment and the like.
With the development of big data and artificial intelligence technology, a special data set is established through the big data of the shipping bill scanning piece, and the sorting and the single number recognition of the bill scanning piece are carried out based on the deep learning recognition technology, so that the traditional manual sorting and inputting link can be effectively replaced, and the working efficiency is improved.
In practical application, a shipping company generally issues a paper bill of lading file to a shipper, and then performs business processes such as scanning, inputting, filing and the like. The traditional operation mode is generally carried out manually, which needs operators with certain business experience to carry out the operations of scanning, classifying, inputting single numbers and the like of bill extraction, for different shipyards, it may also involve association processes between the single numbers, which all require significant human intervention, when a large amount of bill drawing scanning recording is faced, large-scale recording is often difficult to carry out, more time and management cost are brought, and recording of the bill drawing scanning piece at the present stage mostly depends on manual operation of a salesman, but in the face of large batch bill-drawing scanned item data, the manual operation mode has a larger bottleneck on the processing efficiency and accuracy, especially, the misoperation caused by long-time repeated work is not beneficial to the effective filing of the bill-drawing scanning piece, therefore, there is a need for a single number identification method and apparatus based on the big data of the bill drawing scan piece to solve the above problems.
Disclosure of Invention
The invention aims to: in order to solve the problems that the traditional operation mode is generally carried out manually, the operation of scanning, classifying, inputting a single number and the like of an order is carried out by an operator with certain business experience, the association processing among the single numbers may be related to different shipyards, more manual intervention is required, large-scale recording is difficult to carry out when a large number of orders are scanned and recorded, more time and management cost are brought, the recording of the order scanning piece at the current stage is mostly dependent on the manual operation of a business worker, but in the case of large batch of order-extracting scanning piece data, the manual operation mode has larger bottleneck in processing efficiency and accuracy, particularly misoperation caused by long-time repeated work is not beneficial to effective filing of the order-extracting scanning piece, and the single number identification method and the single number identification device based on the large data of the order-extracting scanning piece are provided.
In order to achieve the purpose, the invention adopts the following technical scheme:
a single number recognition device based on large data of a bill-drawing scanning piece comprises a plurality of Convolutional Neural Network (CNN) models, a voting mechanism and an image segmentation technology, wherein the number of the Convolutional Neural Network (CNN) models is multiple, a classification recognition model is established on the basis of the large data of the bill-drawing scanning piece through the plurality of Convolutional Neural Network (CNN) models, the voting mechanism carries out classification recognition on the bill-drawing scanning piece, the image segmentation technology selects a corresponding bill-drawing scanning piece template to carry out image segmentation to obtain a single number area image, the Convolutional Neural Network (CNN) models comprise CNN character recognition models, and the CNN character recognition models recognize the single number area image.
A single number identification method based on bill picking scanning piece big data comprises the following steps:
step S1: making classification labels and bill picking area template images according to the ship company to which the bill picking belongs;
step S2: acquiring big data of a scanned marine bill, cutting an area image of 1/3 in front in the vertical direction, and making a training set H;
step S3: labeling a training set H of a marine bill of lading scanning piece, wherein the training set H comprises the category of a ship company and the character content of a bill of lading;
step S4: utilizing an image segmentation technology to perform segmentation of the bill extraction character to obtain character data sets M of A-Z and 0-9;
step S5: carrying out classification training on the H data set by using ResNet50 to obtain a classification recognition model R1;
step S6: carrying out classification training on the H data set by using IncepotionV 3 to obtain a classification recognition model R2;
step S7: carrying out classification training on the H data set by using a Densenet201 to obtain a classification recognition model R3;
step S8: carrying out classification training on the M data set by using AlexNet to obtain a character recognition model T;
step S9: optimizing and iterating the ship company recognition models R1, R2 and R3 and the character recognition model T;
step S10: acquiring a new shipping bill scanned piece picture, calling models R1, R2 and R3 to identify the classification of the shipcompanies, and acquiring the classification of the shipcompanies of the bill scanned piece according to a majority voting rule;
step S11: obtaining a single number area image through a single number extracting scanning piece template, and calling a model T to perform character recognition to obtain a single number;
step S12: and renaming the bill of lading scanning piece according to the ship company and the bill of lading number, and storing according to the business rule.
As a further description of the above technical solution:
the classification label in the step S1 is set according to the ship company to which the bill of lading belongs, and the bill of lading region template image is set according to the candidate region of the bill number.
As a further description of the above technical solution:
in the step S2, the bill raising scanned piece is subjected to region clipping, and the region clipping is based on that the differences of the bill raising scanned pieces of different shipping companies are mainly concentrated on the layout, Logo and the like of the image header region, and the bill raising layout of the same shipping company is relatively fixed.
As a further description of the above technical solution:
in the step S3, the labeling of the bill of lading data set includes a category label and a bill of lading character content label, the category label corresponds to the category label of the ship company set in S1, and the bill number content is a character combination of a to Z and 0 to 9.
As a further description of the above technical solution:
in the step S4, a list extraction area template image is used for preprocessing to obtain a candidate area, and the single number is segmented by image binarization, connected domain analysis and morphological filtering methods to obtain character sets of A-Z and 0-9.
As a further description of the above technical solution:
in the step S9, the iterative process of optimizing the recognition models R1, R2, and R3 and the character recognition model T is as follows:
step S91: selecting a newly added bill drawing scanning piece, cutting out a front 1/3 area image in the vertical direction, calling recognition models R1, R2 and R3 to obtain a classification result of a shipcompany, carrying out character segmentation based on a corresponding bill drawing area template image, and calling a recognition model T to obtain a bill drawing recognition result;
step S92: manually screening and checking classification results and bill drawing identification results of the ship companies, and carrying out class marking and bill drawing character content marking on incorrect pictures according to the step S3 to obtain a picture set Z;
step S93: retraining the model of the picture set Z according to the steps S5-S8, and updating to obtain a shipcompany recognition model R1, R2, R3 and a character recognition model T;
step S94: the step S91-step S93 are repeated, and the optimization process is repeated.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
1. according to the invention, the bill of lading scanning piece is used as input, image analysis is automatically carried out, the processing mode is simple and convenient, no additional engineering cost is added, the bill of lading scanning piece is automatically classified and character recognition is carried out to obtain the contents of the ship company and the bill of lading character, renaming is carried out according to rules, result verification is carried out only manually, the labor cost is reduced, the working efficiency is improved, a plurality of models are adopted for classification recognition of the ship company to which the bill of lading scanning piece belongs, the classification result is obtained according to majority voting rules, the classification recognition effect can be fully improved, an iterative updating mechanism of a recognition model can be established for the newly added bill of lading scanning piece, the performance of the model is effectively kept improved, accurate recognition is carried out on the bill of lading scanning piece, renaming of the bill of lading scanning piece is automatically carried out according to the recognition result, direct calling of.
2. According to the method, through deep analysis of big data of the bill picking scanning piece, the significant difference of bill picking scanning pieces of different ship companies is mainly concentrated on the layout, Logo and the like of the image head area, which are typical classification bases, so that the method adopts a CNN (computer network) model to classify and recognize the scanning pieces, utilizes a template image to divide the scanning pieces to obtain a single-number area image, utilizes the CNN character recognition model to recognize single-number characters, finally obtains the information of the category and the single number of the bill picking scanning piece, can rename, store and blend other business processes according to rules, reduces the labor cost and improves the working efficiency.
3. In the invention, a plurality of Convolutional Neural Network (CNN) models are adopted to classify the bill-drawing scanning pieces, the image template positioning method is utilized to segment the single-number area, the Convolutional Neural Network (CNN) models are utilized to identify characters, the automatic classification and the single-number identification of the bill-drawing scanning pieces are realized, the renaming is carried out according to the naming rule in combination with the service requirement, the storage and analysis are unified, the whole service flow is automatically entered, and the production management cost is reduced.
Drawings
FIG. 1 is a schematic diagram of an output result in a bill number identification method and apparatus based on big data of a bill picking scanning piece according to the present invention;
FIG. 2 is a schematic diagram of the vertical clipping of the bill of lading scanned item in the bill of lading scanned item big data-based single number identification method and apparatus of the present invention;
FIG. 3 is a schematic diagram of a bill of lading scanning piece template in a bill of lading scanning piece big data-based bill of lading scanning piece identification method and apparatus according to the present invention;
FIG. 4 is a binary diagram in the method and apparatus for single number identification based on the big data of the bill picking scanning piece according to the present invention;
FIG. 5 is a character segmentation diagram in a single number identification method and apparatus based on big data of a bill-picking scanning piece according to the present invention;
fig. 6 is a schematic diagram of character splitting in a bill number identification method and apparatus based on big data of a bill of lading scan piece according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
Referring to fig. 1-6, the present invention provides a technical solution: a single number recognition device based on large data of a bill-drawing scanning piece comprises a plurality of Convolutional Neural Network (CNN) models, a voting mechanism and an image segmentation technology, wherein the number of the Convolutional Neural Network (CNN) models is multiple, a classification recognition model is established on the basis of the large data of the bill-drawing scanning piece through the plurality of Convolutional Neural Network (CNN) models, the voting mechanism carries out classification recognition on the bill-drawing scanning piece, the image segmentation technology selects a corresponding bill-drawing scanning piece template to carry out image segmentation to obtain a single number area image, the Convolutional Neural Network (CNN) models comprise CNN character recognition models, and the CNN character recognition models recognize the single number area image.
A single number identification method based on bill picking scanning piece big data comprises the following steps:
step S1: making classification labels and bill picking area template images according to the ship company to which the bill picking belongs;
step S2: acquiring big data of a scanned marine bill, cutting an area image of 1/3 in front in the vertical direction, and making a training set H;
step S3: labeling a training set H of a marine bill of lading scanning piece, wherein the training set H comprises the category of a ship company and the character content of a bill of lading;
step S4: utilizing an image segmentation technology to perform segmentation of the bill extraction character to obtain character data sets M of A-Z and 0-9;
step S5: carrying out classification training on the H data set by using ResNet50 to obtain a classification recognition model R1;
step S6: carrying out classification training on the H data set by using IncepotionV 3 to obtain a classification recognition model R2;
step S7: carrying out classification training on the H data set by using a Densenet201 to obtain a classification recognition model R3;
step S8: carrying out classification training on the M data set by using AlexNet to obtain a character recognition model T;
step S9: optimizing and iterating the ship company recognition models R1, R2 and R3 and the character recognition model T;
step S10: acquiring a new shipping bill scanned piece picture, calling models R1, R2 and R3 to identify the classification of the shipcompanies, and acquiring the classification of the shipcompanies of the bill scanned piece according to a majority voting rule;
step S11: obtaining a single number area image through a single number extracting scanning piece template, and calling a model T to perform character recognition to obtain a single number;
step S12: and renaming the bill of lading scanning piece according to the ship company and the bill of lading number, and storing according to the business rule.
Specifically, in step S1, the classification label is set according to the ship company to which the bill of lading belongs, and the bill of lading area template image is set according to the candidate area of the bill number.
Specifically, in the step S2, the pick-up order scanned piece is subjected to region clipping, and the region clipping is based on that the differences of the pick-up order scanned pieces of different shipping companies are mainly concentrated on the layout, Logo, and the like of the image header region, and the pick-up layout of the same shipping company is relatively fixed.
Specifically, the labeling of the bill of lading data set in the step S3 includes category labeling and labeling of bill of lading character content, where the category labeling corresponds to the boat company category label set in S1, and the bill number content is a character combination of a to Z and 0 to 9.
Specifically, in the step S4, a list extraction region template image is used for preprocessing to obtain a candidate region, and the single number is segmented by image binarization, connected domain analysis and morphological filtering methods to obtain a character set of a to Z and 0 to 9.
Specifically, in the step S9, the optimization iteration process of the ship company recognition models R1, R2, and R3 and the character recognition model T is as follows:
step S91: selecting a newly added bill drawing scanning piece, cutting out a front 1/3 area image in the vertical direction, calling recognition models R1, R2 and R3 to obtain a classification result of a shipcompany, carrying out character segmentation based on a corresponding bill drawing area template image, and calling a recognition model T to obtain a bill drawing recognition result;
step S92: manually screening and checking classification results and bill drawing identification results of the ship companies, and carrying out class marking and bill drawing character content marking on incorrect pictures according to the step S3 to obtain a picture set Z;
step S93: retraining the model of the picture set Z according to the steps S5-S8, and updating to obtain a shipcompany recognition model R1, R2, R3 and a character recognition model T;
step S94: the step S91-step S93 are repeated, and the optimization process is repeated.
The embodiment takes a bill of lading scanning piece as an input to carry out recognition analysis, and comprises the following steps:
s1, making classification labels and bill drawing area template images according to the ship company to which the bill drawing belongs;
(1) taking the classification of the ship company as a classification label;
(2) analyzing the bill extraction area and formulating a bill extraction area template image;
s2, acquiring big data of a scanned marine bill, cutting a front 1/3 area image in the vertical direction, and making a training set H;
vertically cutting the bill picking scanning piece image to obtain a front 1/3 area image which comprises a single sign, Logo and other significant characteristic areas;
s3, labeling a training set H of the shipping bill scanning piece, wherein the training set H comprises the category of a ship company and the character content of the bill;
marking the category of the ship company and the content of the bill of lading characters on the data set H, wherein the classification is marked as CMA and the characters are marked as ACSA062968 as the drawing of the specification;
s4, performing bill extraction character segmentation by using an image segmentation technology to obtain character data sets M of A-Z and 0-9;
the bill extraction layout of the same ship company is fixed, a candidate area can be obtained through a bill extraction area template image, and a single number is segmented through image binaryzation, connected domain analysis and morphological filtering methods to obtain character sets of A-Z and 0-9;
the method comprises the steps that a target character image is positioned by segmenting large data of a bill picking scanning piece, and character data sets of A-Z and 0-9 are generated;
s5, carrying out classification training on the H data set by using ResNet50 to obtain a classification recognition model R1;
designing a ResNet50 output layer corresponding to a classification label set of a ship company, and training a data set H to obtain a model R1;
s6, carrying out classification training on the H data set by using IncepotionV 3 to obtain a classification recognition model R2;
designing an IncepotionV 3 output layer corresponding to a classification label set of a ship company, and training a data set H to obtain a model R2;
s7, carrying out classification training on the H data set by using the Densenet201 to obtain a classification recognition model R3;
designing a classification label set of a Densenet201 output layer corresponding to a ship company, and training a data set H to obtain a model R3;
s8, carrying out classification training on the M data set by using AlexNet to obtain a character recognition model T;
designing a AlexNet output layer corresponding to A-Z and 0-9 classification label sets, and training a data set M to obtain a model T;
s9, carrying out optimization iteration on the ship company recognition models R1, R2 and R3 and the character recognition model T;
the optimization iteration process is as follows:
s91, selecting a newly added bill raising scanning piece, cutting a front 1/3 area image in the vertical direction, calling recognition models R1, R2 and R3 to obtain a classification result of a shipcompany, carrying out character segmentation on the basis of a corresponding bill raising area template image, calling a recognition model T to obtain a bill raising recognition result;
s92, manually screening and checking classification results of the shipcompanies and bill picking identification results, and carrying out classification labeling and bill picking character content labeling on incorrect pictures according to the step S3 to obtain a picture set Z;
s93, model retraining is carried out on the picture set Z according to the steps S5-S8, and the picture set Z is updated to obtain a recognition model R1, R2 and R3 of the shipcompany and a character recognition model T;
s94, repeating S91-S93, and repeating the optimization process;
s10, acquiring new shipping bill scanned picture, calling models R1, R2 and R3 to identify the category of the ship company as c1、c2、c3And acquiring the classification of the ship company of the bill of lading scanned item according to a majority voting rule, wherein the specific rule is as follows:
c=mode(c1,c2,c3)
mode represents taking the mode of the classification label as the classification result;
s11, obtaining a single number area image through a single-number-extracting scanning piece template, and calling a model T to perform character recognition to obtain a single number;
and S12, renaming the bill of lading scanning piece according to the ship company and the bill of lading number, and storing according to the business rule.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art should be considered to be within the technical scope of the present invention, and the technical solutions and the inventive concepts thereof according to the present invention should be equivalent or changed within the scope of the present invention.
Claims (7)
1. The single number identification device based on the large data of the bill drawing scanning piece is characterized by comprising a plurality of Convolutional Neural Network (CNN) models, a voting mechanism and an image segmentation technology, wherein the Convolutional Neural Network (CNN) models are in a plurality, a classification identification model is established on the basis of the large data of the bill drawing scanning piece through the Convolutional Neural Network (CNN) models, the voting mechanism carries out classification identification on the bill drawing scanning piece, the image segmentation technology selects a corresponding bill drawing scanning piece template to carry out image segmentation to obtain a single number area image, the Convolutional Neural Network (CNN) models comprise CNN character identification models, and the CNN character identification models identify the single number area image.
2. The single number identification method based on the large data of the bill picking scanning piece according to claim 1, characterized by comprising the following steps:
step S1: making classification labels and bill picking area template images according to the ship company to which the bill picking belongs;
step S2: acquiring big data of a scanned marine bill, cutting an area image of 1/3 in front in the vertical direction, and making a training set H;
step S3: labeling a training set H of a marine bill of lading scanning piece, wherein the training set H comprises the category of a ship company and the character content of a bill of lading;
step S4: utilizing an image segmentation technology to perform segmentation of the bill extraction character to obtain character data sets M of A-Z and 0-9;
step S5: carrying out classification training on the H data set by using ResNet50 to obtain a classification recognition model R1;
step S6: carrying out classification training on the H data set by using IncepotionV 3 to obtain a classification recognition model R2;
step S7: carrying out classification training on the H data set by using a Densenet201 to obtain a classification recognition model R3;
step S8: carrying out classification training on the M data set by using AlexNet to obtain a character recognition model T;
step S9: optimizing and iterating the ship company recognition models R1, R2 and R3 and the character recognition model T;
step S10: acquiring a new shipping bill scanned piece picture, calling models R1, R2 and R3 to identify the classification of the shipcompanies, and acquiring the classification of the shipcompanies of the bill scanned piece according to a majority voting rule;
step S11: obtaining a single number area image through a single number extracting scanning piece template, and calling a model T to perform character recognition to obtain a single number;
step S12: and renaming the bill of lading scanning piece according to the ship company and the bill of lading number, and storing according to the business rule.
3. The bill number recognition method based on the large data of the bill extracted scan according to claim 2, wherein the classification label in step S1 is set according to the ship company to which the bill extracted scan belongs, and the bill extraction area template image is set according to the candidate area of the bill number.
4. The bill number recognition method based on the large data of the bill lifting scan piece according to claim 2, wherein in step S2, the bill lifting scan piece is subjected to region cropping, and the region cropping is based on that the differences of the bill lifting scan pieces of different shipping companies are mainly concentrated on the layout, Logo and the like of the image header region, and the bill lifting layout of the same shipping company is relatively fixed.
5. The bill number recognition method based on the big data of the bill scanning piece as claimed in claim 2, wherein the labeling of the bill data set in step S3 includes a category label and a bill character content label, and the category label corresponds to the ship company category label set in S1, and the bill number content is a character combination of a-Z and 0-9.
6. The method for identifying the single number based on the big data of the bill picking scanning piece according to claim 2, wherein in the step S4, a candidate region is obtained by preprocessing a bill picking region template image, and the single number is segmented by image binarization, connected domain analysis and morphological filtering methods to obtain character sets of A to Z and 0 to 9.
7. The single number recognition method based on the large data of bill of lading scanner as claimed in claim 2, wherein said step S9 is performed on the ship company recognition models R1, R2 and R3, and the character recognition model T is optimized as follows:
step S91: selecting a newly added bill drawing scanning piece, cutting out a front 1/3 area image in the vertical direction, calling recognition models R1, R2 and R3 to obtain a classification result of a shipcompany, carrying out character segmentation based on a corresponding bill drawing area template image, and calling a recognition model T to obtain a bill drawing recognition result;
step S92: manually screening and checking classification results and bill drawing identification results of the ship companies, and carrying out class marking and bill drawing character content marking on incorrect pictures according to the step S3 to obtain a picture set Z;
step S93: retraining the model of the picture set Z according to the steps S5-S8, and updating to obtain a shipcompany recognition model R1, R2, R3 and a character recognition model T;
step S94: the step S91-step S93 are repeated, and the optimization process is repeated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010702824.9A CN112200177A (en) | 2020-07-21 | 2020-07-21 | Single number identification method and device based on bill picking scanning piece big data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010702824.9A CN112200177A (en) | 2020-07-21 | 2020-07-21 | Single number identification method and device based on bill picking scanning piece big data |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112200177A true CN112200177A (en) | 2021-01-08 |
Family
ID=74006432
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010702824.9A Pending CN112200177A (en) | 2020-07-21 | 2020-07-21 | Single number identification method and device based on bill picking scanning piece big data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112200177A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115243250A (en) * | 2022-07-25 | 2022-10-25 | 每日互动股份有限公司 | Method, system and storage medium for acquiring wifi portrait |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
CN108596166A (en) * | 2018-04-13 | 2018-09-28 | 华南师范大学 | A kind of container number identification method based on convolutional neural networks classification |
CN109214382A (en) * | 2018-07-16 | 2019-01-15 | 顺丰科技有限公司 | A kind of billing information recognizer, equipment and storage medium based on CRNN |
CN109284750A (en) * | 2018-08-14 | 2019-01-29 | 北京市商汤科技开发有限公司 | Bank slip recognition method and device, electronic equipment and storage medium |
-
2020
- 2020-07-21 CN CN202010702824.9A patent/CN112200177A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106650721A (en) * | 2016-12-28 | 2017-05-10 | 吴晓军 | Industrial character identification method based on convolution neural network |
CN108596166A (en) * | 2018-04-13 | 2018-09-28 | 华南师范大学 | A kind of container number identification method based on convolutional neural networks classification |
CN109214382A (en) * | 2018-07-16 | 2019-01-15 | 顺丰科技有限公司 | A kind of billing information recognizer, equipment and storage medium based on CRNN |
CN109284750A (en) * | 2018-08-14 | 2019-01-29 | 北京市商汤科技开发有限公司 | Bank slip recognition method and device, electronic equipment and storage medium |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115243250A (en) * | 2022-07-25 | 2022-10-25 | 每日互动股份有限公司 | Method, system and storage medium for acquiring wifi portrait |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US11113557B2 (en) | System and method for generating an electronic template corresponding to an image of an evidence | |
CN107194400B (en) | Financial reimbursement full ticket image recognition processing method | |
US6014450A (en) | Method and apparatus for address block location | |
CN112508011A (en) | OCR (optical character recognition) method and device based on neural network | |
CN113780087B (en) | Postal package text detection method and equipment based on deep learning | |
Zhai et al. | A generative adversarial network based framework for unsupervised visual surface inspection | |
CN112200177A (en) | Single number identification method and device based on bill picking scanning piece big data | |
CN111091090A (en) | Bank report OCR recognition method, device, platform and terminal | |
CN112508000B (en) | Method and equipment for generating OCR image recognition model training data | |
CN114581928A (en) | Form identification method and system | |
CN111414917A (en) | Identification method of low-pixel-density text | |
CN110688445B (en) | Digital archive construction method | |
CN113743159A (en) | OCR method applied to power enterprises | |
CN111178464A (en) | Application of OCR recognition based on neural network in logistics industry express bill | |
CN115713775A (en) | Method, system and computer equipment for extracting form from document | |
CN110175563B (en) | Metal cutting tool drawing mark identification method and system | |
JP2003317034A (en) | Document classification system and program for realizing the same | |
CN109325557B (en) | Data intelligence acquisition method based on computer visual image identification | |
CN112348022A (en) | Free-form document identification method based on deep learning | |
CN111047261A (en) | Warehouse logistics order identification method and system | |
CN113657373A (en) | Automatic document cataloguing method | |
CN111401356A (en) | Express single-hand-written telephone number recognition method based on deep learning | |
CN114202761B (en) | Information batch extraction method based on picture information clustering | |
CN112686238B (en) | Deep learning-based shipping bill identification method | |
CN117711004A (en) | Form document information extraction method based on image 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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210108 |