Logistics order invoice registration identification method, apparatus and device, and storage medium

Abstract

The invention relates to the technical field of image recognition processing, in particular to a logistics order invoice registration recognition method and device, equipment and a storage medium. The logistics order invoice registration identification method comprises the following steps: obtaining an unregistered invoice picture, and carrying out the preprocessing of the unregistered invoice picture, so as to obtain a preprocessed picture; constructing a character recognition model, and processing the preprocessed picture by adopting the character recognition model to extract character information in the preprocessed picture; and establishing a verification rule, verifying the character information according to the verification rule, marking the character information which does not pass the verification, and associating the character information which passes the verification with the logistics order. According to the logistics order invoice registration identification method, invoice registration can be automatically completed, invoice with problems can be accurately identified, invoice registration efficiency can be improved, errors caused by manual registration can be reduced, financial risks can be effectively reduced, and problem invoices can be avoided.

Description

Technical Field

[0001] The present invention relates to the field of image recognition processing technology, and in particular to a logistics order invoice registration and recognition method, device, equipment and storage medium. Background Art

[0002] The traditional invoice registration method mainly relies on manual input, and the staff needs to manually input the information on the invoice into the system one by one. This method has many disadvantages: first, there is a problem of low efficiency, especially when facing a large number of invoices, manual input requires a lot of time and manpower costs; second, there is a problem of easy errors. Due to the influence of human factors, input errors and missing information may occur, resulting in inaccurate invoice information; third, it is difficult to ensure the timeliness of data. Manual input often has a certain lag, and the invoice information cannot be registered in the system in time, affecting the company's financial management and decision-making. With the continuous increase in logistics business volume and the rapid development of information technology, the traditional invoice registration method can no longer meet the needs of modern logistics companies. On the one hand, logistics companies need a more efficient and accurate invoice registration method to improve work efficiency, reduce costs, and ensure the accuracy and completeness of invoice information. On the other hand, with the increasingly strict tax management, logistics companies need a more standardized and scientific invoice management method to avoid tax risks. Summary of the invention

[0003] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a logistics order invoice registration and identification method, device, equipment and storage medium, aiming to solve the technical problems of low efficiency and high error rate in the traditional invoice entry method in the prior art.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] The first aspect of the present invention provides a logistics order invoice registration and recognition method, comprising the following steps: obtaining an unregistered invoice image, preprocessing the unregistered invoice image to obtain a preprocessed image; constructing a character recognition model, using the character recognition model to process the preprocessed image to extract character information in the preprocessed image; establishing verification rules, verifying the character information according to the verification rules, marking the character information that fails the verification, and associating the character information that passes the verification with the logistics order.

[0006] Optionally, in a first implementation manner of the first aspect of the present invention, the obtaining of the unregistered invoice picture and preprocessing the unregistered invoice picture to obtain a preprocessed picture specifically includes: obtaining the unregistered invoice picture and denoising the unregistered invoice picture using a filtering algorithm to obtain a first picture; processing the first picture using a histogram equalization method to enhance the contrast of the first picture to obtain a second picture; processing the second picture using a threshold method to convert the second picture into a black and white binary image to obtain a preprocessed picture.

[0007] Optionally, in a second implementation of the first aspect of the present invention, the character recognition model is constructed and the preprocessed image is processed using the character recognition model to extract character information in the preprocessed image, specifically including: obtaining a sample invoice image, and annotating the characters in the sample invoice image to obtain training data; taking a convolutional neural network model as a basic model, and training the basic model using a deep learning framework and training data to obtain a preliminary model; verifying the accuracy of the preliminary model, and continuously optimizing the preliminary model based on the verification result to obtain a character recognition model.

[0008] Optionally, in a third implementation of the first aspect of the present invention, establishing verification rules, verifying character information according to the verification rules, marking character information that fails the verification, and associating character information that passes the verification with a logistics order, specifically includes: obtaining the format specifications and business rules of the invoice, and establishing verification rules according to the format specifications and business rules of the invoice; verifying the character information according to the verification rules and a pre-built enterprise database to obtain a verification result; if the verification result is that the character information does not comply with the verification rules or cannot be matched with the enterprise database, marking it as an exception; if the verification result is that the character information complies with the verification rules and can be matched with the enterprise database, it is deemed to have passed the verification, and the character information that passes the verification is associated with the logistics order.

[0009] Optionally, in a fourth implementation method of the first aspect of the present invention, the character information is verified according to the verification rules and a pre-built enterprise database to obtain a verification result, specifically including: obtaining different types of paper invoice images, and constructing a clarity judgment model according to the different types of paper invoice images; judging whether the invoice is a paper invoice according to the character information, and if the invoice is a paper invoice, using the clarity judgment model to judge the unregistered invoice image to obtain a judgment result; if the judgment result is failure, marking the corresponding unregistered invoice image; if the judgment result is pass, verifying the character information according to the verification rules and a pre-built enterprise database to obtain a verification result.

[0010] Optionally, in a fifth implementation manner of the first aspect of the present invention, the obtaining of different types of paper invoice images and constructing a clarity judgment model based on the different types of paper invoice images specifically includes: obtaining different types of paper invoice images, marking whether the clarity of the paper invoice images meets the requirements, and forming a training set with the paper invoice images and their marking results; using a machine learning algorithm as a basic model, and using the training set to train the basic model to obtain a clarity judgment model; formulating an automatic optimization iteration strategy, automatically updating the training set according to the automatic optimization iteration strategy, and using the updated training set to optimize and iterate the clarity judgment model.

[0011] Optionally, in a sixth implementation of the first aspect of the present invention, if the judgment result is through, the character information is verified according to the verification rules and a pre-built enterprise database to obtain a verification result, specifically including: if the judgment result is through, obtaining key information of the enterprise and its affiliated companies, the key information including taxpayer identification number and enterprise name, and building an enterprise database with the key information; verifying the character information according to the verification rules and the enterprise database, the character information including: taxpayer information, invoicing date, amount and tax rate, matching the taxpayer information in the character information with the enterprise database to obtain a first verification result; verifying the invoicing date, amount or tax rate in the character information according to the verification rules to obtain a second verification result.

[0012] The second aspect of the present invention provides a logistics order invoice registration and identification device, including: a preprocessing module, used to obtain an unregistered invoice image, and preprocess the unregistered invoice image to obtain a preprocessed image; an extraction module, used to build a character recognition model, and use the character recognition model to process the preprocessed image to extract character information in the preprocessed image; a verification module, used to establish verification rules, verify the character information according to the verification rules, mark the character information that fails the verification, and associate the character information that passes the verification with the logistics order.

[0013] Optionally, in a first implementation manner of the second aspect of the present invention, the preprocessing module includes: a denoising unit, used to obtain an unregistered invoice image, and to perform denoising on the unregistered invoice image using a filtering algorithm to obtain a first image; an adjustment unit, used to process the first image using a histogram equalization method to enhance the contrast of the first image to obtain a second image; and a conversion unit, used to process the second image using a threshold method to convert the second image into a black and white binary image to obtain a preprocessed image.

[0014] Optionally, in a second implementation of the second aspect of the present invention, the extraction module includes: an acquisition unit, used to acquire a sample invoice image, and annotate the characters in the sample invoice image to obtain training data; a training unit, used to train the basic model using a convolutional neural network model as a basic model, using a deep learning framework and training data to obtain a preliminary model; an optimization unit, used to verify the accuracy of the preliminary model, and continuously optimize the preliminary model based on the verification result to obtain a character recognition model.

[0015] Optionally, in a third implementation of the second aspect of the present invention, the verification module includes: an establishment sub-module, used to obtain the format specifications and business rules of the invoice, and establish verification rules based on the format specifications and business rules of the invoice; a verification sub-module, used to verify the character information according to the verification rules and a pre-built enterprise database to obtain a verification result; a marking sub-module, used to mark as abnormal if the verification result is that the character information does not comply with the verification rules or cannot be matched with the enterprise database; an association sub-module, used to deem the verification as passed if the verification result is that the character information complies with the verification rules and can be matched with the enterprise database, and associate the verified character information with the logistics order.

[0016] Optionally, in a fourth implementation of the second aspect of the present invention, the verification submodule includes: a construction unit, used to obtain different types of paper invoice images, and construct a clarity judgment model based on the different types of paper invoice images; a judgment unit, used to judge whether the invoice is a paper invoice based on the character information, and if the invoice is a paper invoice, use the clarity judgment model to judge the unregistered invoice image to obtain a judgment result; a marking unit, used to mark the corresponding unregistered invoice image if the judgment result is failure; and a verification unit, used to verify the character information according to the verification rules and a pre-constructed enterprise database to obtain a verification result if the judgment result is a pass.

[0017] Optionally, in a fifth implementation of the second aspect of the present invention, the construction unit includes: an acquisition subunit, used to acquire different types of paper invoice images, mark whether the clarity of the paper invoice images meets the requirements, and form a training set with the paper invoice images and their annotation results; a training subunit, used to train the basic model with the training set using the machine learning algorithm as the basic model to obtain a clarity judgment model; an optimization subunit, used to formulate an automatic optimization iteration strategy, automatically update the training set according to the automatic optimization iteration strategy, and use the updated training set to optimize and iterate the clarity judgment model.

[0018] Optionally, in a sixth implementation of the second aspect of the present invention, the verification unit includes: a construction subunit, which is used to obtain key information of the enterprise and its affiliated companies if the judgment result is passed, and the key information includes the taxpayer identification number and the enterprise name, and to build an enterprise database with the key information; a first verification subunit, which is used to verify the character information according to the verification rules and the enterprise database, and the character information includes: taxpayer information, invoice date, amount and tax rate, and match the taxpayer information in the character information with the enterprise database to obtain a first verification result; a second verification subunit, which is used to verify the invoice date, amount or tax rate in the character information according to the verification rules to obtain a second verification result.

[0019] The third aspect of the present invention provides a logistics order invoice registration and identification device, including a memory and at least one processor, wherein the memory stores computer-readable instructions; the at least one processor calls the computer-readable instructions in the memory to execute the various steps of the logistics order invoice registration and identification method as described above.

[0020] A fourth aspect of the present invention provides a computer-readable storage medium having computer-readable instructions stored thereon, which, when executed by a processor, implement the various steps of the logistics order invoice registration and identification method as described above.

[0021] Beneficial effect: The present invention provides a logistics order invoice registration and recognition method, which first obtains an unregistered invoice picture, pre-processes the unregistered invoice picture, obtains a pre-processed picture, thereby improving the quality of the picture and making the processing result of the subsequent model more accurate, and then processes the pre-processed picture by adopting a character recognition model to extract character information in the pre-processed picture without manual operation, with higher efficiency and lower error rate, and finally establishes a verification rule, verifies the character information according to the verification rule, marks the character information that fails the verification, associates the character information that passes the verification with the logistics order, automatically verifies and identifies abnormal invoices, and automatically associates the invoices that have successfully been verified, effectively improving work efficiency and the timeliness of information collection. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A first flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0023] Figure 2 A second flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0024] Figure 3A third flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0025] Figure 4 A fourth flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0026] Figure 5 A fifth flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0027] Figure 6 A sixth flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0028] Figure 7 A seventh flow chart of the logistics order invoice registration and identification method provided by an embodiment of the present invention;

[0029] Figure 8 A schematic diagram of the structure of a logistics order invoice registration and identification device provided by an embodiment of the present invention;

[0030] Fig. 9 Another structural schematic diagram of the logistics order invoice registration and identification device provided by an embodiment of the present invention;

[0031] Fig.10 A schematic diagram of the structure of a logistics order invoice registration and identification device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0032] The present invention provides a method, device, equipment and storage medium for registering and identifying a logistics order invoice. The present invention obtains an unregistered invoice picture first, pre-processes the unregistered invoice picture, and obtains a pre-processed picture; then constructs a character recognition model, uses the character recognition model to process the pre-processed picture, and extracts character information in the pre-processed picture; finally, establishes a verification rule, verifies the character information according to the verification rule, marks the character information that fails the verification, and associates the character information that passes the verification with the logistics order. The present invention adopts a method of automatically extracting and verifying the information in the invoice, which effectively improves the registration efficiency of the invoice. In order to improve the accuracy of character recognition, the unregistered invoice picture is further pre-processed, and the character recognition model is used to extract the character information, which effectively avoids the problem of errors and omissions in the extraction of character information, and ensures that the logistics order invoice registration process is carried out quickly and efficiently.

[0033] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] For ease of understanding, the specific process of the embodiment of the present invention is described below. It should be noted that the relevant steps of obtaining user personal information in the present invention are all performed with the user's authorization.

[0035] See also Figure 1 The first embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0036] S101. Obtain an unregistered invoice image, and pre-process the unregistered invoice image to obtain a pre-processed image;

[0037] Specifically, each outlet of the logistics enterprise can directly upload invoice images to the system at their respective business locations, so that the system can obtain unregistered invoice images. The quality of unregistered invoice images taken by different personnel or different devices varies greatly. In order to facilitate the subsequent character recognition model to accurately extract the character information, the unregistered invoice images need to be preprocessed. There are many ways to preprocess, such as denoising, contrast enhancement, brightness adjustment, conversion to black and white images, etc. Various preprocessing methods can be used in combination to improve the preprocessing effect.

[0038] S102. Constructing a character recognition model, and using the character recognition model to process the preprocessed image to extract character information in the preprocessed image;

[0039] The character recognition model can automatically process a large number of images and extract the character information from them, which significantly improves work efficiency and reduces the need for manual intervention. Trained character recognition models usually have a high recognition accuracy, especially when processing clear and standardized character images. This helps reduce the errors that may occur in manual recognition. The character recognition model can also handle character images in various complex scenarios, such as blur, rotation, distortion, etc. This makes the model more flexible in practical applications.

[0040] The character information extracted from the pre-processed image includes various key information on the invoice, such as: company name, taxpayer identification number, amount, tax amount, invoicing items, etc. The extracted character information can also be further used for verification analysis, such as verifying whether the invoice is valid or forged.

[0041] S103. Establish verification rules, verify the character information according to the verification rules, mark the character information that fails the verification, and associate the character information that passes the verification with the logistics order.

[0042] There are strict requirements on the format of each character on the invoice, such as the font format, date writing format, invoicing items and tax amount, etc. By establishing verification rules, the system can verify each content obtained from the invoice image one by one according to the verification rules to ensure that the content of the invoice is accurate and meets the requirements.

[0043] When it is found that the invoice content does not meet the requirements, the system will mark the abnormal character content and output it to the interface for display to the user. If the character information meets the verification rules, the character information will be associated with the logistics order to complete the registration and identification of the logistics order invoice.

[0044] Through the above method, the efficiency of invoice registration can be improved. During the specific implementation, the outlets can directly upload the invoice attachments, and the corresponding invoicing information can be automatically brought out through the background interface, and the invoice month can be selected to reduce the invoice registration operation time.

[0045] In addition, the present invention can also improve the registration accuracy and effectiveness: during verification, the authenticity of the verification and the return of the invoice information can be identified through a three-party interface to avoid manual input of invoice information errors. The automated invoice registration and recognition method reduces the workload of manual entry and improves the efficiency of invoice registration. At the same time, the image recognition technology and data verification algorithm are used to ensure the accuracy and completeness of the invoice information. Accurately registered invoice information can provide reliable data support for the tax declaration and financial management of logistics companies, helping companies comply with tax laws and regulations and reduce tax risks.

[0046] See also Figure 2 The second embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0047] S201. Obtain an unregistered invoice image, and perform denoising on the unregistered invoice image using a filtering algorithm to obtain a first image;

[0048] Specifically, during the denoising process, algorithms such as median filtering and Gaussian filtering can be used to remove noise points in the invoice image. Median filtering is a nonlinear filtering method that effectively removes salt and pepper noise by replacing the grayscale value of each pixel in the image with the median of the grayscale values ​​of all pixels in the neighborhood of the pixel. Gaussian filtering is a linear filtering method that removes noise by weighted averaging the image and has a good effect in removing Gaussian noise.

[0049] S202. Processing the first image using a histogram equalization method to enhance the contrast of the first image to obtain a second image;

[0050] Histogram equalization is a method to enhance the contrast of an image by adjusting the grayscale histogram of the image. It stretches the grayscale range of the image so that the dark areas in the image become darker and the bright areas become brighter, thereby improving the readability of the image.

[0051] S203. Process the second image using a threshold method to convert the second image into a black and white binary image to obtain a preprocessed image.

[0052] Convert the preprocessed image into a black and white binary image for subsequent character recognition. Binarization can be done using a fixed threshold method or an adaptive threshold method. The fixed threshold method sets the pixels in the image whose grayscale value is greater than a fixed threshold to white, and sets the pixels whose grayscale value is less than the threshold to black. The adaptive threshold method automatically determines the threshold based on the local features of the image, and has a better processing effect on invoice images with uneven lighting.

[0053] See also Figure 3 The third embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0054] S301. Obtain a sample invoice image and annotate the characters in the sample invoice image to obtain training data; manual or automatic annotation methods may be used to ensure the accuracy and completeness of the annotation.

[0055] S302. Taking the convolutional neural network (CNN) model as the basic model, the basic model is trained using the deep learning framework and training data to obtain a preliminary model;

[0056] Deep learning frameworks such as TensorFlow and PyTorch can be used. During the training process, it is necessary to continuously adjust the model parameters, such as learning rate, batch size, network structure, etc., to improve the accuracy of character recognition. At the same time, cross-validation, early stopping and other techniques can be used to prevent model overfitting.

[0057] S303. Verify the accuracy of the preliminary model, and continuously optimize the preliminary model according to the verification result to obtain a character recognition model.

[0058] After training is completed, the character recognition model can be deployed to the invoice registration and recognition system to realize automatic character recognition of invoice images. In actual application, the model can be further optimized and adjusted according to the characteristics of the invoice and the recognition effect.

[0059] Furthermore, the recognized characters can be post-processed to remove some erroneous recognition results. For example, the recognized characters can be checked and corrected by dictionary query, grammar analysis, etc. to ensure the accuracy of the recognition results.

[0060] In addition to using the above methods to build a character recognition model, you can also choose an optical character recognition (OCR) engine with excellent performance, such as Tesseract OCR, Baidu OCR, etc. These OCR engines have been trained and optimized a lot and can accurately recognize characters in various fonts and languages.

[0061] In specific applications, parameters can be adjusted and optimized according to the characteristics of the invoice. For example, for specific fonts, font sizes, colors, etc. on the invoice, the recognition parameters of the OCR engine can be adjusted to improve the accuracy of character recognition. At the same time, the invoice image can be segmented and the characters in different areas can be recognized separately to improve the accuracy and efficiency of recognition.

[0062] See also Figure 4 The fourth embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0063] S401. Obtain the format specification and business rules of the invoice, and establish verification rules according to the format specification and business rules of the invoice;

[0064] Invoices have specific format specifications. For example, an invoice should include the invoice name, invoice code and number, copies and purpose, customer name, bank account and account number, product name or business item, unit of measurement, quantity, unit price, amount in both uppercase and lowercase letters, tax rate (levy rate), tax amount, issuer, invoice date, name (seal) of the issuing unit (individual), etc.

[0065] In addition, invoices must be issued truthfully in all copies at once in accordance with the prescribed time limit, sequence, and columns, and must be stamped with a special invoice stamp. The use of text must be standardized, and Chinese should be used; the amount in uppercase and lowercase letters must be standardized; the invoice date must also be filled in in a standardized manner. The columns such as the name of the purchasing unit, the name of the goods or service items, specifications, units, quantities, and unit prices must be filled in in a standardized manner. The items must be filled in completely and the handwriting must be clear. All copies should be copied or printed at once and filled in in numerical order.

[0066] By establishing verification rules based on the invoice format specifications and business rules, the system can perform verification according to relevant regulations to ensure that the verification results are correct.

[0067] S402. Verify the character information according to the verification rules and the pre-built enterprise database to obtain a verification result;

[0068] S403. If the verification result is that the character information does not meet the verification rules or cannot match the enterprise database, it is marked as abnormal;

[0069] When an anomaly is found, the corresponding content can be marked to prompt manual processing in time. If it is found through manual inspection that the problem is still the unregistered invoice image itself, the user will be required to provide a new invoice. If the model itself identifies a problem, it can be verified through manual channels.

[0070] S404. If the verification result is that the character information complies with the verification rules and can be matched with the enterprise database, it is considered to have passed the inspection, and the verified character information is associated with the logistics order.

[0071] After association, the invoice information can be stored in a database or saved in the form of an electronic document. When registering, the invoice information should be complete and accurate, and the registration time, registration personnel and other information of the invoice should be recorded for subsequent query and management. In order to facilitate query and management, the invoice information can be classified and archived. For example, the invoice information can be classified according to the invoice type, invoice date, taxpayer identification number, etc., and stored in different folders or database tables. At the same time, an index can be established to facilitate and quickly query specific invoice information.

[0072] See also Figure 5 The fifth embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0073] S501. Obtain different types of paper invoice images, and construct a clarity judgment model based on different types of paper invoice images;

[0074] Compared with electronic invoices, paper invoices may have many printing problems or improper storage problems. For example, if there are stains on the paper invoice, the key information cannot be viewed, and the system cannot recognize it and can only provide a new invoice. For another example, paper invoices are printed by a printer, and there may be problems with unclear printing. Electronic invoices are directly generated electronic files, and there is no such problem of blurred content.

[0075] S502. Determine whether the invoice is a paper invoice based on the character information. If the invoice is a paper invoice, use a clarity judgment model to judge the unregistered invoice image to obtain a judgment result;

[0076] The character information includes all the character contents on the invoice, including the invoice header, which accurately indicates whether the invoice is an electronic invoice. By determining whether the character information contains information about the electronic invoice, it is possible to determine whether the invoice is a paper invoice.

[0077] S503. If the judgment result is not passed, the corresponding unregistered invoice image is marked;

[0078] S504. If the judgment result is passed, the character information is verified according to the verification rules and the pre-built enterprise database to obtain a verification result.

[0079] In this embodiment, if the clarity judgment model believes that the clarity of the paper invoice image is too low, it will be marked. At this time, it will remind human intervention to check whether it meets the requirements. If it is a problem with the photo, the user will be required to provide a new photo. If it is a problem with the paper invoice itself, the user will be required to provide a new invoice.

[0080] Since it is not necessary to verify all character information during verification, passing the verification does not ensure that the paper invoice itself does not have other problems. However, in this embodiment, when verifying the character information, other problems of the paper invoice have been eliminated in advance through the clarity judgment model, which can ensure that the physical paper invoice received later meets the financial requirements.

[0081] See also Figure 6 The sixth embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0082] S601. Obtain different types of paper invoice images, mark whether the clarity of the paper invoice images meets the requirements, and form a training set with the paper invoice images and their marking results;

[0083] Specifically, in order to form a training set with more complete data, it is necessary to obtain images of paper invoices with different defects, such as invoices with stains that make some key information unrecognizable, damaged invoices, unclear printing, invoices with incomplete content, etc. By manually marking which types or levels of invoices are acceptable in advance and then forming the corresponding training set, the obtained clarity judgment model can have a higher judgment accuracy.

[0084] S602. Based on the machine learning algorithm, the basic model is trained using the training set to obtain a clarity judgment model;

[0085] S603. Formulate an automatic optimization iteration strategy, automatically update the training set according to the automatic optimization iteration strategy, and use the updated training set to optimize and iterate the clarity judgment model.

[0086] Specifically, a new batch of paper invoice images can be obtained through the system, and added to the training set to expand the amount of training data, thereby improving the accuracy of the clarity judgment model.

[0087] See also Figure 7 The seventh embodiment of the logistics order invoice registration and identification method in the embodiment of the present invention includes:

[0088] S701. If the judgment result is passed, obtain key information of the enterprise and its affiliated companies, the key information including taxpayer identification number and enterprise name, and build an enterprise database with the key information;

[0089] In order to reduce the company's system development costs, all subsidiaries or affiliated companies can use the same invoice registration and identification system. The company database stores relevant information data of all subsidiaries or affiliated companies. As long as the company name and its taxpayer identification number meet the requirements, they can be matched and found in the company database.

[0090] S702. Verify the character information according to the verification rules and the enterprise database, the character information includes: taxpayer information, invoice date, amount and tax rate, and match the taxpayer information in the character information with the enterprise database to obtain a first verification result;

[0091] Specifically, the character information may also include key information such as the name of goods or taxable services, service, unit, quantity, unit price, tax amount, etc. Taxpayer information includes information such as company name and taxpayer identification number. If any one of them cannot be correctly matched, it will be deemed as a failure of verification.

[0092] S703. Verify the invoicing date, amount or tax rate in the character information according to the verification rule to obtain a second verification result.

[0093] For example: if the taxpayer information in the character information cannot be matched with the enterprise database, the verification fails; if the invoicing date in the character information does not conform to the date format, the verification fails; if the amount in the character information is not a positive number, the verification fails; if the tax rate in the character information does not match the name of the goods or taxable labor or service, the verification fails.

[0094] The above describes the logistics order invoice registration and identification method in the embodiment of the present invention. The following describes the logistics order invoice registration and identification device in the embodiment of the present invention. Figure 8In one embodiment of the present invention, a device for registering and identifying a logistics order invoice includes:

[0095] A preprocessing module 10 is used to obtain an unregistered invoice image and preprocess the unregistered invoice image to obtain a preprocessed image;

[0096] An extraction module 20, for constructing a character recognition model, and using the character recognition model to process the preprocessed image to extract character information in the preprocessed image;

[0097] The verification module 30 is used to establish verification rules, verify the character information according to the verification rules, mark the character information that fails the verification, and associate the character information that passes the verification with the logistics order.

[0098] See also Fig. 9 In one embodiment of the present invention, a device for registering and identifying a logistics order invoice includes:

[0099] A preprocessing module 10 is used to obtain an unregistered invoice image and preprocess the unregistered invoice image to obtain a preprocessed image;

[0100] An extraction module 20, for constructing a character recognition model, and using the character recognition model to process the preprocessed image to extract character information in the preprocessed image;

[0101] A verification module 30, for establishing verification rules, verifying character information according to the verification rules, marking character information that fails the verification, and associating character information that passes the verification with a logistics order;

[0102] In this embodiment, the preprocessing module 10 includes:

[0103] A denoising unit 11 is used to obtain an unregistered invoice image and perform denoising on the unregistered invoice image using a filtering algorithm to obtain a first image;

[0104] An adjustment unit 12, configured to process the first image using a histogram equalization method to enhance the contrast of the first image to obtain a second image;

[0105] The conversion unit 13 is used to process the second image by using a threshold method to convert the second image into a black and white binary image to obtain a preprocessed image;

[0106] In this embodiment, the extraction module 20 includes:

[0107] An acquisition unit 21 is used to acquire a sample invoice image and annotate characters in the sample invoice image to obtain training data;

[0108] The training unit 22 is used to train the basic model using the convolutional neural network model as a basic model using a deep learning framework and training data to obtain a preliminary model;

[0109] The optimization unit 23 is used to verify the accuracy of the preliminary model and continuously optimize the preliminary model according to the verification result to obtain a character recognition model;

[0110] In this embodiment, the verification module 30 includes:

[0111] Establishing a submodule 31 for obtaining the format specification and business rules of the invoice, and establishing verification rules according to the format specification and business rules of the invoice;

[0112] A verification submodule 32, used to verify the character information according to the verification rules and a pre-built enterprise database to obtain a verification result;

[0113] A marking submodule 33 is used to mark as abnormal if the verification result is that the character information does not meet the verification rules or cannot be matched with the enterprise database;

[0114] The association submodule 34 is used to associate the character information that has passed the verification with the logistics order if the verification result shows that the character information meets the verification rules and can be matched with the enterprise database;

[0115] In this embodiment, the check submodule 32 includes:

[0116] A construction unit 321 is used to obtain different types of paper invoice images and construct a clarity judgment model according to the different types of paper invoice images;

[0117] A judgment unit 322 is used to judge whether the invoice is a paper invoice according to the character information. If the invoice is a paper invoice, a definition judgment model is used to judge the unregistered invoice image to obtain a judgment result;

[0118] The marking unit 323 is used to mark the corresponding unregistered invoice image if the judgment result is failure;

[0119] A verification unit 324, configured to verify the character information according to the verification rule and a pre-built enterprise database to obtain a verification result if the judgment result is passed;

[0120] In this embodiment, the construction unit 321 includes:

[0121] The acquisition subunit 3211 is used to acquire different types of paper invoice images, mark whether the clarity of the paper invoice images meets the requirements, and form a training set with the paper invoice images and their marking results;

[0122] The training subunit 3212 is used to train the basic model using a training set based on the machine learning algorithm to obtain a clarity judgment model;

[0123] The optimization subunit 3213 is used to formulate an automatic optimization iteration strategy, automatically update the training set according to the automatic optimization iteration strategy, and use the updated training set to optimize and iterate the clarity judgment model;

[0124] In this embodiment, the verification unit 324 includes:

[0125] The construction subunit 3241 is used to obtain key information of the enterprise and its affiliated companies if the judgment result is passed, the key information includes the taxpayer identification number and the enterprise name, and construct an enterprise database with the key information;

[0126] A first verification subunit 3242 is used to verify the character information according to the verification rule and the enterprise database, the character information including: taxpayer information, invoice date, amount and tax rate, and match the taxpayer information in the character information with the enterprise database to obtain a first verification result;

[0127] The second verification subunit 3243 is used to verify the invoicing date, amount or tax rate in the character information according to the verification rule to obtain a second verification result.

[0128] The logistics order invoice registration and recognition device of the present invention improves the quality of the unregistered invoice pictures by automatically pre-processing the acquired unregistered invoice pictures, and then accurately extracts the character information in the pre-processed pictures by constructing a character recognition model and using the character recognition model to process the pre-processed pictures. The device is highly efficient and does not require manual processing. The registration of the invoice content can be completed quickly. Finally, verification rules are established and the character information is verified according to the verification rules. If the character information verification fails, it is marked. The character information that passes the verification is associated with the logistics order to complete the storage of the invoice information. Through the above process, the occurrence of human errors can be effectively reduced, and problematic invoices can be automatically eliminated to reduce financial risks.

[0129] The above is a detailed description of the logistics order invoice registration and identification device in the embodiment of the present invention from the perspective of modular functional entities. The following is a detailed description of the logistics order invoice registration and identification device in the embodiment of the present invention from the perspective of hardware processing.

[0130] Fig.10A schematic diagram of the structure of a logistics order invoice registration and identification device provided in an embodiment of the present invention. The logistics order invoice registration and identification device 900 may have relatively large differences due to different configurations or performances, and may include one or more processors (central processing units, CPU) 910 (for example, one or more processors) and a memory 920, and one or more storage media 930 (for example, one or more mass storage devices) storing application programs 933 or data 932. Among them, the memory 920 and the storage medium 930 can be temporary storage or permanent storage. The program stored in the storage medium 930 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations in the logistics order invoice registration and identification device 900. Furthermore, the processor 910 can be configured to communicate with the storage medium 930, and execute a series of instruction operations in the storage medium 930 on the logistics order invoice registration and identification device 900 to implement the steps of the logistics order invoice registration and identification method provided in the above-mentioned method embodiments.

[0131] The logistics order invoice registration and identification device 900 may also include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input and output interfaces 960, and / or, one or more operating systems 931, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will appreciate that Fig.10 The structure of the logistics order invoice registration and identification device shown does not constitute a limitation of the logistics order invoice registration and identification device, and may include more or fewer components than shown in the figure, or a combination of certain components, or a different arrangement of components.

[0132] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions are executed on a computer, the computer executes the steps of the logistics order invoice registration and identification method.

[0133] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the equipment or device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0134] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0135] It is understandable that those skilled in the art can make equivalent substitutions or changes based on the technical solution and inventive concept of the present invention, and all these changes or substitutions should fall within the protection scope of the claims attached to the present invention.

Claims

1. A logistics order invoice registration and identification method, characterized in that: The steps include: Acquire an unregistered invoice image, and pre-process the unregistered invoice image to obtain a pre-processed image; Constructing a character recognition model, and using the character recognition model to process the preprocessed image to extract character information in the preprocessed image; Verification rules are established, character information is verified according to the verification rules, character information that fails the verification is marked, and character information that passes the verification is associated with the logistics order.

2. The logistics order invoice registration and identification method according to claim 1 is characterized in that: The obtaining of the unregistered invoice image and preprocessing the unregistered invoice image to obtain the preprocessed image specifically includes: Obtain an unregistered invoice image, and perform denoising on the unregistered invoice image using a filtering algorithm to obtain a first image; Processing the first image by using a histogram equalization method to enhance the contrast of the first image, thereby obtaining a second image; The second image is processed by using a threshold method to convert the second image into a black and white binary image to obtain a preprocessed image.

3. The logistics order invoice registration and identification method according to claim 1 is characterized in that: The constructing of the character recognition model and using the character recognition model to process the preprocessed image to extract character information in the preprocessed image specifically includes: Obtaining sample invoice images and annotating characters in the sample invoice images to obtain training data; Taking the convolutional neural network model as the basic model, the deep learning framework and training data are used to train the basic model to obtain a preliminary model; The accuracy of the preliminary model is verified, and the preliminary model is continuously optimized according to the verification result to obtain a character recognition model.

4. The logistics order invoice registration and identification method according to claim 1 is characterized in that: The establishing of the verification rules, verifying the character information according to the verification rules, marking the character information that fails the verification, and associating the character information that passes the verification with the logistics order specifically includes: Obtaining the format specifications and business rules of the invoice, and establishing verification rules based on the format specifications and business rules of the invoice; Verify the character information according to the verification rules and the pre-built enterprise database to obtain a verification result; If the verification result is that the character information does not meet the verification rules or cannot be matched with the enterprise database, it is marked as abnormal; If the verification result is that the character information complies with the verification rules and can be matched with the enterprise database, it is considered to have passed the inspection, and the verified character information will be associated with the logistics order.

5. The logistics order invoice registration and identification method according to claim 4 is characterized in that: The verifying the character information according to the verification rule and the pre-built enterprise database to obtain the verification result specifically includes: Obtain different types of paper invoice images, and build a clarity judgment model based on the different types of paper invoice images; Determine whether the invoice is a paper invoice based on the character information. If the invoice is a paper invoice, use the clarity judgment model to judge the unregistered invoice image to obtain a judgment result; If the judgment result is failure, the corresponding unregistered invoice image will be marked; If the judgment result is passed, the character information is verified according to the verification rule and the pre-built enterprise database to obtain a verification result.

6. The logistics order invoice registration and identification method according to claim 5 is characterized in that: The method of obtaining different types of paper invoice images and constructing a clarity judgment model according to the different types of paper invoice images specifically includes: Obtain different types of paper invoice images, mark whether the clarity of the paper invoice images meets the requirements, and form a training set with the paper invoice images and their marking results; Taking the machine learning algorithm as the basic model, the basic model is trained with the training set to obtain the clarity judgment model; An automatic optimization iteration strategy is formulated, the training set is automatically updated according to the automatic optimization iteration strategy, and the clarity judgment model is optimized and iterated using the updated training set.

7. The logistics order invoice registration and identification method according to claim 5 is characterized in that: If the judgment result is passed, the character information is verified according to the verification rules and the pre-built enterprise database to obtain a verification result, which specifically includes: If the judgment result is passed, key information of the enterprise and its affiliated companies is obtained, the key information includes the taxpayer identification number and the enterprise name, and the enterprise database is constructed with the key information; Verifying the character information according to the verification rule and the enterprise database, the character information including: taxpayer information, invoice date, amount and tax rate, matching the taxpayer information in the character information with the enterprise database to obtain a first verification result; The invoicing date, amount or tax rate in the character information is verified according to the verification rule to obtain a second verification result.

8. A logistics order invoice registration and identification device, characterized in that: include: A preprocessing module, used for obtaining an unregistered invoice image, and preprocessing the unregistered invoice image to obtain a preprocessed image; An extraction module, used to construct a character recognition model, and use the character recognition model to process the pre-processed image to extract character information in the pre-processed image; The verification module is used to establish verification rules, verify the character information according to the verification rules, mark the character information that fails the verification, and associate the character information that passes the verification with the logistics order.

9. A logistics order invoice registration and identification device, characterized in that: comprising a memory and at least one processor, wherein the memory has computer-readable instructions stored therein; The at least one processor calls the computer-readable instructions in the memory to execute the various steps of the logistics order invoice registration and identification method as described in any one of claims 1-7.

10. A computer-readable storage medium having computer-readable instructions stored thereon, characterized in that: When the computer-readable instructions are executed by the processor, the various steps of the logistics order invoice registration and identification method as described in any one of claims 1-7 are implemented.

Images