CN118799890B - A method for express information extraction and distributed representation based on deep learning - Google Patents
A method for express information extraction and distributed representation based on deep learning Download PDFInfo
- Publication number
- CN118799890B CN118799890B CN202411280328.3A CN202411280328A CN118799890B CN 118799890 B CN118799890 B CN 118799890B CN 202411280328 A CN202411280328 A CN 202411280328A CN 118799890 B CN118799890 B CN 118799890B
- Authority
- CN
- China
- Prior art keywords
- information
- express
- digital
- text
- address
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- 238000013135 deep learning Methods 0.000 title claims abstract description 21
- 238000000605 extraction Methods 0.000 title claims abstract description 15
- 238000012512 characterization method Methods 0.000 claims abstract description 41
- 238000001514 detection method Methods 0.000 claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 26
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 238000012216 screening Methods 0.000 claims abstract description 5
- 238000005516 engineering process Methods 0.000 claims description 21
- 238000002372 labelling Methods 0.000 claims description 13
- 230000003287 optical effect Effects 0.000 claims description 13
- 238000012015 optical character recognition Methods 0.000 claims description 12
- 238000012937 correction Methods 0.000 claims description 8
- 238000012360 testing method Methods 0.000 claims description 5
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 240000004282 Grewia occidentalis Species 0.000 claims description 3
- 238000013145 classification model Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 230000010365 information processing Effects 0.000 abstract description 4
- 238000010276 construction Methods 0.000 abstract 1
- 238000002360 preparation method Methods 0.000 abstract 1
- 238000011161 development Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
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
- G06V30/19—Recognition using electronic means
- G06V30/191—Design or setup of recognition systems or techniques; Extraction of features in feature space; Clustering techniques; Blind source separation
- G06V30/19173—Classification techniques
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/08—Logistics, e.g. warehousing, loading or distribution; Inventory or stock management
- G06Q10/083—Shipping
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/764—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using classification, e.g. of video objects
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06V30/18—Extraction of features or characteristics of the image
- G06V30/1801—Detecting partial patterns, e.g. edges or contours, or configurations, e.g. loops, corners, strokes or intersections
-
- 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
- G06V30/19—Recognition using electronic means
- G06V30/19007—Matching; Proximity measures
-
- 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
- G06V30/40—Document-oriented image-based pattern recognition
- G06V30/42—Document-oriented image-based pattern recognition based on the type of document
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Business, Economics & Management (AREA)
- Molecular Biology (AREA)
- Economics (AREA)
- Medical Informatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Databases & Information Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Marketing (AREA)
- Human Resources & Organizations (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a deep learning-based express information extraction and distributed characterization method, which comprises the following steps: data preparation, namely acquiring an express delivery face list data set required by training through data enhancement and preprocessing; model construction, namely training an express delivery face list area and a text area identification model by using YOLOv target detection network; identifying an express delivery face list, and cutting out an express delivery face list area; acquiring the effective information of the face list and acquiring the coordinates of a target text area; paddleOCR identifying texts, screening and identifying text information; text screening, namely classifying and digitizing the identified text information; the distributed representation is used for respectively converting address, telephone and name information into digital codes; express matching, namely matching express characteristic information with characteristic information in a client database by comparing digital codes, and binding corresponding clients. The method improves the efficiency and accuracy of express information processing and delivery, and reduces the error rate of manual operation.
Description
Technical Field
The invention relates to the field of logistics storage, computer vision and natural language processing, in particular to an express information extraction and distributed characterization method based on deep learning.
Background
With the rapid development of electronic commerce, the business volume of the express industry is rapidly increased, and how to efficiently and accurately manage and distribute express information becomes a problem to be solved urgently. The traditional express information processing method generally depends on manual operation, and is low in efficiency and easy to make mistakes. The rapid development of target detection and Optical Character Recognition (OCR) technology based on deep learning provides a new technical means for solving the problem.
Disclosure of Invention
The invention aims to provide an express information extraction and distributed characterization method based on deep learning, which realizes the efficient identification of an express bill and a target text region by using YOLOv target detection network and PaddleOCR optical character identification technology, digitizes the identified text information and forms uniform digital codes so as to realize the automatic management and efficient distribution of the express information. The traditional express information processing method generally depends on manual operation, and is low in efficiency and easy to make mistakes.
An express information extraction and distributed characterization method based on deep learning, the method comprises the following steps:
S1, acquiring an express delivery face list data set required by training, enriching the training set by a data enhancement method, preprocessing the data set and finishing data labeling;
S2, training a network aiming at express bill area identification by using YOLOv target detection network, and predicting the bill position and rotation angle; carrying out detection model reasoning, carrying out rotation target detection on the face sheet according to the detected position and angle, outputting a face sheet image and a face sheet rotation angle, and righting the cut face sheet in a single direction to obtain a complete face sheet area;
S3, training a model for detecting the position of the target text by using YOLOv target detection network, and predicting the position of the target text; performing detection model reasoning, and calculating text region coordinates according to the detected position to obtain a YOLO target information frame;
S4, identifying express bill information by using PaddleOCR optical character identification technology to obtain PaddleOCR text frames, and calculating PaddleOCR text frame coordinates and YOLO target information frame coordinates on the basis of YOLOv target identification results and PaddleOCR optical character identification technology text identification results;
S5, establishing an information digital distributed characterization method to classify the obtained text information, and storing the classified text information in an Excel table;
S6, establishing a complete customer database according to the information digital distributed characterization method, wherein the characteristic information of each customer in the database comprises a name, an address and a telephone number and is stored in a digital mode;
S7, according to an information digital distributed characterization method, the identified express characteristic information containing characters such as literal, name, address, telephone number and the like is converted into digital characteristic information, and then the express characteristic information is coded into a series of ordered codes; when the characteristic information of a certain express delivery face list is not identified, the information is marked by a single byte character 0 through coding;
S8, the express characteristic information obtained according to the information digital distributed characterization method is a string of digital codes which are arranged according to a set sequence, the digital codes are compared with the digital codes corresponding to the customer characteristic information in the database, the digital address codes, the digital telephone codes and the digital name codes are sequentially matched, and if the digital address codes, the digital telephone codes and the digital name codes can be matched correctly, the express is bound to the customer; if the matching is not correct, the information is input into a customer database.
As a further technical scheme of the invention, the information digital distributed characterization method comprises the following steps:
a) Identifying an address, and converting each section of address (including communities, buildings and family numbers) in the literal address into a digital address, wherein the address is an 8-bit digital string;
b) Identifying a telephone number of the customer, the telephone number being an 11-digit string;
c) And identifying the customer name, wherein the customer name is a 12-bit number string, and one Chinese character corresponds to 4 digits according to the comparison of the four-corner number dictionary to form a code string.
As a further technical scheme of the invention, the data labeling of the data set in the S1 comprises labeling an express delivery face sheet area and an effective information area in the face sheet, wherein the express delivery face sheet area data set collects multiple groups of express delivery with face sheets shot at different angles, RGB images are converted into 640 multiplied by 480 resolution, the effective information data set in the face sheet collects the cut express delivery face sheet, gray images are converted into 640 multiplied by 480 resolution, and data enhancement is carried out by using a data enhancement method, which comprises rotation, cutting, overturning and noise addition; marking the express delivery face list area is to use Labelme software to carry out rectangular label marking on the express delivery face list area in RGB images of the express delivery face list shot at different angles, and the label storage format is selected from the YOLO format; the effective information area in the labeling menu is to use Labelme software to label the number of the express menu and the recipient information in the gray level image of the cut express menu by rectangular labels, and the label storage format is selected from the Yolo format.
As a further technical scheme of the invention, a YOLOv target detection network is used for training in the S2 to obtain a model, and the model can effectively detect and cut and extract overlapped surface sheets, multi-surface sheets, fuzzy surface sheets, surface sheets with incomplete information and fold surface sheets; the method comprises the steps of firstly rotating a face sheet angle, carrying out preliminary correction by adopting affine transformation of OpenCV, then judging the face sheet orientation by using an orientation classification model, carrying out final correction on the face sheet in the reverse direction and the left and right sides, and finally outputting the face sheet which is a forward angle with characters upwards; the method comprises the steps that the face sheet information extraction is based on YOLOv rotation target detection, the face sheet is subjected to rotation target detection, a complete face sheet area is cut according to a rotation rectangular frame predicted by a model, and other non-face sheet areas are not included; training the YOLOv target detection network trained in the step S3, and dividing a training set, a verification set and a test set by a sample according to the ratio of 3:1:1 to obtain a model for detecting the position of a target text; the target position detection is to select effective text area information in the picture, including the name, address, telephone number and express bill number of the customer in the bill.
As a further technical scheme of the present invention, the PaddleOCR optical character recognition technology used in S4 is an optical character recognition technology based on a convolutional neural network, characters are converted into image information through an optical input mode, the optical input mode includes scanning, the image information is converted into editable computer text by utilizing the character recognition technology, and text information is screened by calculating the coincidence degree of PaddleOCR frame coordinates and YOLO frame coordinates; and calculating the coincidence degree of PaddleOCR text boxes and YOLOv text boxes, and reserving PaddleOCR text boxes with the coincidence degree reaching 0.5, so as to achieve the aim of screening text information.
As a further technical solution of the present invention, the preset sequence of the information digital distributed characterization method in S5 is as follows: sequentially comprises communities, buildings, family numbers, telephone numbers and names from left to right; a code is formed by 31-bit single-byte characters to represent specific characteristic information, the code consists of 5 sections, namely a community code section, a building code section, a household code section, a telephone code section and a name code section, wherein the community code section consists of 2-bit single-byte characters, the building code section consists of 2-bit single-byte characters, the household code section consists of 4-bit single-byte characters, the telephone code section consists of 11-bit single-byte characters, and the name code section consists of 12-bit single-byte characters.
As a further technical solution of the present invention, the address code segment of the information digital distributed characterization method in S5 includes: a tenth address code segment, a hundredth address code segment, a thousandth address code segment and a thousandth address code segment, wherein the tenth address code segment is a first second bit character of the address code segment, the hundredth address code segment is a third fourth bit character of the address code segment, the thousandth address code segment is a fifth sixth bit character of the segment code segment, and the thousandth address code segment is a seventh eighth bit character of the address code segment.
As a further technical scheme of the invention, the complete customer database in S6 is the digital characteristic information of the community customers, the community customers enter personal information, and the community customers' characteristic information is converted into digital coded ordered codes according to an information digital distributed characterization method to obtain the characteristic digital information of the community customers.
As a further technical scheme of the invention, the express characteristic information in S7 is that when express delivery is carried out, an express bill is shot through a camera, information on the express bill is acquired by using an optical character recognition technology based on deep learning, the identified express bill information is converted into digital characteristic information according to an information digitizing method, digital information distribution carries out a principle of step-by-step distribution from front to back, a series of ordered codes are formed, and the express bill information is not existed due to external factors, and the non-existent information is converted into digital 0 when the express bill information is converted according to the information digitizing method, wherein the external factors comprise blurring, overlapping, stains and folds of the express bill.
As a further technical scheme of the invention, the express characteristic information obtained by the information digital distributed characterization method in the S8 is a string of digital codes arranged according to a set sequence, and if the express characteristic information and the client characteristic information are matched correctly, the express is bound to corresponding clients; if the matching is not right, judging that the client information of the client to which the express belongs is not in the client database, and storing the express characteristic information in the client database as the client characteristic information in the database.
Compared with the prior art, the invention has the beneficial effects that:
According to the invention, by combining YOLOv target detection network and PaddleOCR optical character recognition technology, the high-efficiency recognition of the express bill and the target text area is realized, and the recognized text information is converted into uniform digital codes by an information digitizing method, so that the automatic management and high-efficiency distribution of the express information are realized, the efficiency and accuracy of the express information processing are improved, and the error rate of manual operation is reduced.
Drawings
Fig. 1 is a flowchart of a method for extracting express information and performing distributed characterization according to an embodiment of the present invention.
Fig. 2 is a schematic structural diagram of a method for digitally representing information in an embodiment of the present invention.
Detailed Description
The technical scheme of the invention is further described in detail below with reference to the specific embodiments.
Referring to fig. 1 to 2, an embodiment of the present invention provides a deep learning-based express information extraction and distributed characterization method thereof, the method includes the following steps:
S1, acquiring an express delivery face list data set required by training, enriching the training set by a data enhancement method, preprocessing the data set and finishing data labeling;
the method comprises the steps of obtaining a large number of express delivery face list data sets required by training from different express delivery companies, wherein data samples comprise face list pictures shot under various angles and illumination conditions, and the face list data sets comprise multiple groups of express delivery with face lists shot at different angles. The RGB image is converted to 640 x 480 resolution size and data enhancement operations such as rotation, cropping, flipping, and noise addition are performed. Preprocessing a dataset, marking rectangular labels on express delivery face single areas shot at different angles by using Labelme software, selecting a Yolo format as a label storage format, and marking the marked express delivery face single areas with a sample size of about 4800. And processing the cut express bill, converting the cut express bill into a gray image with the resolution of 640 multiplied by 480, marking the effective information (such as express bill number and recipient information) in the express bill by using Labelme software with rectangular labels, selecting a Yolo format for a label storage format, and marking the marked express bill with the sample size of about 4800.
The data labeling of the data set in the S1 comprises labeling an express delivery face list area and an effective information area in the face list, wherein the labeling of the express delivery face list area is to use Labelme software to label rectangular labels of the express delivery face list area in RGB images of the express delivery face list shot at different angles, and a label storage format is selected from a YOLO format; the effective information area in the labeling menu is to use Labelme software to label the number of the express menu and the recipient information in the gray level image of the cut express menu by rectangular labels, and the label storage format is selected from the Yolo format.
S2, training a network aiming at express bill area identification by using YOLOv target detection network, and predicting the bill position and rotation angle; carrying out detection model reasoning, carrying out rotation target detection on the face sheet according to the detected position and angle, outputting a face sheet image and a face sheet rotation angle, and righting the cut face sheet in a single direction to obtain a complete face sheet area;
Training the S2 by using YOLOv target detection network to obtain a model, wherein the model can effectively detect and cut and extract overlapped surface sheets, multi-surface sheets, fuzzy surface sheets and surface sheets with incomplete information; the method comprises the steps of firstly rotating a face sheet angle, carrying out preliminary correction by adopting affine transformation of OpenCV, then judging the face sheet orientation by using an orientation classification model, carrying out final correction on the face sheet in the reverse direction and the left and right sides, and finally outputting the face sheet which is a forward angle with characters upwards; the method comprises the steps that the face sheet information extraction is based on YOLOv rotation target detection, the face sheet is subjected to rotation target detection, a complete face sheet area is cut according to a rotation rectangular frame predicted by a model, and other non-face sheet areas are not included; training the YOLOv target detection network trained in the step S3, dividing a training set, a verification set and a test set according to the ratio of 3:1:1 by using about 1440 samples for training the model, about 480 samples for verifying the model, about 480 samples for testing the model, and training the marked data set to obtain a model for detecting the position of the target text; the target position detection is to select effective text area information in the picture, including the name, address, telephone number and express bill number of the customer in the bill. The model calculates the text region coordinates to realize effective text region information frame selection.
S3, training a model for detecting the position of the target text by using YOLOv target detection network, and predicting the position of the target text; performing detection model reasoning, and calculating text region coordinates according to the detected position to obtain a YOLO target information frame;
Inputting an original express image into a trained model for detecting the position of a target text, carrying out detection model reasoning on a test set, outputting a face list image and a rotation angle, carrying out preliminary correction by using affine transformation of OpenCV, carrying out final correction on the face list in the reverse direction, the left and right directions, obtaining a forward face list image with characters upwards, identifying and cutting out an express face list area, and obtaining an independent express face list image.
S4, identifying express bill information by using PaddleOCR optical character identification technology to obtain PaddleOCR text frames, and calculating PaddleOCR text frame coordinates and YOLO target information frame coordinates on the basis of YOLOv target identification results and PaddleOCR optical character identification technology text identification results;
The PaddleOCR optical character recognition technology used in the S4 is an optical character recognition technology based on a convolutional neural network, characters are converted into image information through an optical input mode, the optical input mode comprises scanning, the image information is converted into editable computer texts through the character recognition technology, and text information is screened through calculating the coincidence degree of PaddleOCR frame coordinates and YOLO frame coordinates; and calculating the coincidence degree of PaddleOCR text boxes and YOLOv text boxes, and reserving PaddleOCR text boxes with the coincidence degree reaching 0.5, so as to achieve the aim of screening text information.
S5, establishing an information digital distributed characterization method to classify the obtained text information, and storing the classified text information in an Excel table;
the information digital distributed characterization method comprises the following steps:
a) Identifying an address, and converting each section of address (including communities, buildings and family numbers) in the literal address into a digital address, wherein the address is an 8-bit digital string;
b) Identifying a telephone number of the customer, the telephone number being an 11-digit string;
c) And identifying the customer name, wherein the customer name is a 12-bit number string, and one Chinese character corresponds to 4 digits according to the comparison of the four-corner number dictionary to form a code string.
The preset sequence of the information digital distributed characterization method in the S5 is as follows: sequentially comprises communities, buildings, family numbers, telephone numbers and names from left to right; a code is formed by 31-bit single-byte characters to represent specific characteristic information, the code consists of 5 sections, namely a community code section, a building code section, a household code section, a telephone code section and a name code section, wherein the community code section consists of 2-bit single-byte characters, the building code section consists of 2-bit single-byte characters, the household code section consists of 4-bit single-byte characters, the telephone code section consists of 11-bit single-byte characters, and the name code section consists of 12-bit single-byte characters.
The address code segment of the information digital distributed characterization method in S5 includes: a tenth address code segment, a hundredth address code segment, a thousandth address code segment and a thousandth address code segment, wherein the tenth address code segment is a first second bit character of the address code segment, the hundredth address code segment is a third fourth bit character of the address code segment, the thousandth address code segment is a fifth sixth bit character of the segment code segment, and the thousandth address code segment is a seventh eighth bit character of the address code segment.
S6, establishing a complete customer database according to the information digital distributed characterization method, wherein the characteristic information of each customer in the database comprises a name, an address and a telephone number and is stored in a digital mode;
And S6, the complete customer database is the digital characteristic information of the community customer, the community customer inputs personal information, and the community customer characteristic information is converted into a digital coded ordered code according to an information digital distributed characterization method to obtain the characteristic digital information of the community customer.
S7, according to an information digital distributed characterization method, the identified express characteristic information containing characters such as literal, name, address, telephone number and the like is converted into digital characteristic information, and then the express characteristic information is coded into a series of ordered codes; when the characteristic information of a certain express delivery face list is not identified, the information is marked by a single byte character 0 through coding;
The express characteristic information in the S7 is that when express delivery is carried out, an express bill is shot through a camera, information on the express bill is acquired by using an optical character recognition technology based on deep learning, the identified express bill information is converted into digital characteristic information according to an information digitizing method, a principle of gradual distribution from front to back is carried out on digital information distribution, a string of ordered codes is formed, the information of the express bill is not existed due to external factors, the information which is not existed is converted into digital 0 when the information is converted according to the information digitizing method, and the external factors comprise blurring, overlapping, stains and folds of the express bill.
S8, the express characteristic information obtained according to the information digital distributed characterization method is a string of digital codes which are arranged according to a set sequence, the digital codes are compared with the digital codes corresponding to the customer characteristic information in the database, the digital address codes, the digital telephone codes and the digital name codes are sequentially matched, and if the digital address codes, the digital telephone codes and the digital name codes can be matched correctly, the express is bound to the customer; if the matching is not correct, the information is input into a customer database.
The express characteristic information obtained by the information digital distributed characterization method in the S8 is a string of digital codes arranged according to a set sequence, and if the express characteristic information and the client characteristic information are matched correctly, the express is bound to corresponding clients; if the matching is not right, judging that the client information of the client to which the express belongs is not in the client database, and storing the express characteristic information in the client database as the client characteristic information in the database.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
The foregoing description is only of the preferred embodiments of the present invention, and is not intended to limit the scope of the invention, but rather is intended to cover any equivalents of the structures or equivalent processes disclosed herein or in the alternative, which may be employed directly or indirectly in other related arts.
Claims (10)
1. The express information extraction and distributed characterization method based on deep learning is characterized by comprising the following steps of:
S1, acquiring an express delivery face list data set required by training, enriching the training set by a data enhancement method, preprocessing the data set and finishing data labeling;
S2, training a network aiming at express bill area identification by using YOLOv target detection network, and predicting the bill position and rotation angle; carrying out detection model reasoning, carrying out rotation target detection on the face sheet according to the detected position and angle, outputting a face sheet image and a face sheet rotation angle, and righting the cut face sheet in a single direction to obtain a complete face sheet area;
S3, training a model for detecting the position of the target text by using YOLOv target detection network, and predicting the position of the target text; performing detection model reasoning, and calculating text region coordinates according to the detected position to obtain a YOLO target information frame;
S4, identifying express bill information by using PaddleOCR optical character identification technology to obtain PaddleOCR text frames, and calculating PaddleOCR text frame coordinates and YOLO target information frame coordinates on the basis of YOLOv target identification results and PaddleOCR optical character identification technology text identification results;
S5, establishing an information digital distributed characterization method to classify the obtained text information, and storing the classified text information in an Excel table;
S6, establishing a complete customer database according to the information digital distributed characterization method, wherein the characteristic information of each customer in the database comprises a name, an address and a telephone number and is stored in a digital mode;
S7, according to an information digital distributed characterization method, the identified express characteristic information containing the literal, name, address and telephone number is converted into digital characteristic information, and then the express characteristic information is encoded into a string of ordered codes; when the characteristic information of a certain express delivery face list is not identified, the information is marked by a single byte character 0 through coding;
S8, the express characteristic information obtained according to the information digital distributed characterization method is a string of digital codes which are arranged according to a set sequence, the digital codes are compared with the digital codes corresponding to the customer characteristic information in the database, the digital address codes, the digital telephone codes and the digital name codes are sequentially matched, and if the digital address codes, the digital telephone codes and the digital name codes can be matched correctly, the express is bound to the customer; if the matching is not correct, the information is input into a customer database.
2. The deep learning-based express information extraction and distributed characterization method according to claim 1, wherein the information digital distributed characterization method comprises the following steps:
a) Identifying an address, converting each segment of address in the literal address into a digital address, wherein the address is an 8-bit digital string;
b) Identifying a telephone number of the customer, the telephone number being an 11-digit string;
c) And identifying the customer name, wherein the customer name is a 12-bit number string, and one Chinese character corresponds to 4 digits according to the comparison of the four-corner number dictionary to form a code string.
3. The method for extracting and distributing type characterization of express information based on deep learning according to claim 1, wherein the data labeling of the data set in S1 includes labeling two kinds of express bill area and effective information area in the express bill, the express bill area data set collects multiple groups of express with bills shot at different angles, RGB images are converted into 640 x 480 resolution, the effective information data set in the express bill collects the cut express bill, gray images are converted into 640 x 480 resolution, and data enhancement is performed by using a data enhancement method, wherein the data enhancement method includes rotation, cutting, overturning and noise addition; marking the express delivery face list area is to use Labelme software to carry out rectangular label marking on the express delivery face list area in RGB images of the express delivery face list shot at different angles, and the label storage format is selected from the YOLO format; the effective information area in the labeling menu is to use Labelme software to label the number of the express menu and the recipient information in the gray level image of the cut express menu by rectangular labels, and the label storage format is selected from the Yolo format.
4. The method for extracting express information and carrying out distributed characterization on the basis of deep learning according to claim 1, wherein a YOLOv target detection network is used for training in the step S2 to obtain a model, and the model is used for effectively detecting and cutting and extracting overlapped surface sheets, multi-surface sheets, fuzzy surface sheets and fold surface sheets with incomplete information; the method comprises the steps of firstly rotating a face sheet angle, carrying out preliminary correction by adopting affine transformation of OpenCV, then judging the face sheet orientation by using an orientation classification model, carrying out final correction on the face sheet in the reverse direction and the left and right sides, and finally outputting the face sheet which is a forward angle with characters upwards; the method comprises the steps that the face sheet information extraction is based on YOLOv rotation target detection, the face sheet is subjected to rotation target detection, a complete face sheet area is cut according to a rotation rectangular frame predicted by a model, and other non-face sheet areas are not included; training the YOLOv target detection network trained in the step S3, and dividing a training set, a verification set and a test set by a sample according to the ratio of 3:1:1 to obtain a model for detecting the position of a target text; the target position detection is to select effective text area information in the picture, including the name, address, telephone number and express bill number of the customer in the bill.
5. The method for extracting and distributing the express information based on the deep learning according to claim 1, wherein PaddleOCR optical character recognition technology used in the step S4 is optical character recognition technology based on convolutional neural network, characters are converted into image information through an optical input mode, the optical input mode comprises scanning, the image information is converted into editable computer text through the character recognition technology, and text information is screened through calculating the coincidence degree of PaddleOCR frame coordinates and YOLO frame coordinates; and calculating the coincidence degree of PaddleOCR text boxes and YOLOv text boxes, and reserving PaddleOCR text boxes with the coincidence degree reaching 0.5, so as to achieve the aim of screening text information.
6. The deep learning-based express information extraction and distributed characterization method according to claim 2, wherein the preset sequence of the information digital distributed characterization method in S5 is as follows: sequentially comprises communities, buildings, family numbers, telephone numbers and names from left to right; a code is formed by 31-bit single-byte characters to represent specific characteristic information, the code consists of 5 sections, namely a community code section, a building code section, a household code section, a telephone code section and a name code section, wherein the community code section consists of 2-bit single-byte characters, the building code section consists of 2-bit single-byte characters, the household code section consists of 4-bit single-byte characters, the telephone code section consists of 11-bit single-byte characters, and the name code section consists of 12-bit single-byte characters.
7. The deep learning-based express delivery information extraction and distributed characterization method according to claim 6, wherein the address code section of the information digital distributed characterization method in S5 includes: a tenth address code segment, a hundredth address code segment, a thousandth address code segment and a thousandth address code segment, wherein the tenth address code segment is a first second bit character of the address code segment, the hundredth address code segment is a third fourth bit character of the address code segment, the thousandth address code segment is a fifth sixth bit character of the segment code segment, and the thousandth address code segment is a seventh eighth bit character of the address code segment.
8. The deep learning-based express information extraction and distributed characterization method according to claim 1, wherein the complete customer database in S6 is digital characteristic information of the community customers, the community customers enter principal information, and the community customers' characteristic information is converted into digital coded ordered codes according to the information digital distributed characterization method, so as to obtain characteristic digital information of the community customers.
9. The method for extracting and distributing the express delivery information based on the deep learning according to claim 1, wherein the express delivery characteristic information in the S7 is that when the express delivery is delivered, the express delivery bill is shot by a camera, the information on the express delivery bill is acquired by using an optical character recognition technology based on the deep learning, the recognized express delivery bill information is converted into digital characteristic information according to an information digitizing method, a principle of gradual distribution from front to back is carried out by digital information distribution, a series of ordered codes are formed, the information of the express delivery bill is not existed due to external factors, the information which does not exist is converted into digital 0 when the information is converted according to the information digitizing method, and the external factors comprise blurring, overlapping, stains and folds of the express delivery bill.
10. The method for extracting and distributing the express information based on the deep learning according to claim 1, wherein the express characteristic information obtained by the information digital distributed characterization method in the step S8 is a string of digital codes arranged according to a set sequence, and if the express characteristic information and the client characteristic information are matched correctly, the express is bound to the corresponding client; if the matching is not right, judging that the client information of the client to which the express belongs is not in the client database, and storing the express characteristic information in the client database as the client characteristic information in the database.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411280328.3A CN118799890B (en) | 2024-09-13 | 2024-09-13 | A method for express information extraction and distributed representation based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202411280328.3A CN118799890B (en) | 2024-09-13 | 2024-09-13 | A method for express information extraction and distributed representation based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118799890A CN118799890A (en) | 2024-10-18 |
| CN118799890B true CN118799890B (en) | 2024-11-12 |
Family
ID=93035266
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202411280328.3A Active CN118799890B (en) | 2024-09-13 | 2024-09-13 | A method for express information extraction and distributed representation based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118799890B (en) |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113919345A (en) * | 2021-10-15 | 2022-01-11 | 北京同城必应科技有限公司 | Intelligent information extraction method for express orders in same city |
| CN116129416A (en) * | 2023-01-31 | 2023-05-16 | 广东海洋大学 | AI algorithm-based vehicle management system with double systems and double modes |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110377686B (en) * | 2019-07-04 | 2021-09-17 | 浙江大学 | Address information feature extraction method based on deep neural network model |
| US20230134651A1 (en) * | 2021-10-28 | 2023-05-04 | Akporefe Agbamu | Synchronized Identity, Document, and Transaction Management |
| CN118155226A (en) * | 2024-03-13 | 2024-06-07 | 国网安徽省电力有限公司涡阳县供电公司 | Power consumption information input method and system based on image recognition and computer equipment |
-
2024
- 2024-09-13 CN CN202411280328.3A patent/CN118799890B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113919345A (en) * | 2021-10-15 | 2022-01-11 | 北京同城必应科技有限公司 | Intelligent information extraction method for express orders in same city |
| CN116129416A (en) * | 2023-01-31 | 2023-05-16 | 广东海洋大学 | AI algorithm-based vehicle management system with double systems and double modes |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118799890A (en) | 2024-10-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN1103087C (en) | Recognition and Correction Method of Optical Scanning Form | |
| JP5522408B2 (en) | Pattern recognition device | |
| JP5105561B2 (en) | Method for processing mail pieces using client code associated with digital imprint | |
| CN112508011A (en) | OCR (optical character recognition) method and device based on neural network | |
| CN108960223A (en) | The method for automatically generating voucher based on bill intelligent recognition | |
| US20140185933A1 (en) | Document image compression method and its application in document authentication | |
| US20140268250A1 (en) | Systems and methods for receipt-based mobile image capture | |
| CN114550158B (en) | Scene character recognition method and system | |
| CN108364037A (en) | Method, system and the equipment of Handwritten Chinese Character Recognition | |
| CN113158895A (en) | Bill identification method and device, electronic equipment and storage medium | |
| CN113887484B (en) | Card type file image identification method and device | |
| CN112149401A (en) | Document comparison identification method and system based on ocr | |
| CN112149654B (en) | Invoice text information identification method based on deep learning | |
| CN113657377A (en) | Structured recognition method for airplane ticket printing data image | |
| CN115116068B (en) | Archive intelligent archiving system based on OCR | |
| CN116311292A (en) | Document image information extraction method, device, computer equipment and storage medium | |
| CN111213157A (en) | Express information input method and system based on intelligent terminal | |
| CN118799890B (en) | A method for express information extraction and distributed representation based on deep learning | |
| CN108090728B (en) | Express information input method and system based on intelligent terminal | |
| US9679179B2 (en) | Method for processing information from a hand-held scanning device | |
| CN116756358A (en) | Electronic management method for flight manifest | |
| JP2009146245A (en) | Image collation method, image collation apparatus, and image collation program | |
| CN114241485A (en) | Information identification method, device, equipment and storage medium of property certificate | |
| JP3090070B2 (en) | Form identification method and device | |
| Kumar et al. | Optical Character Recognition (OCR) Using Opencv and Python: Implementation and Performance Analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |