Supplier intelligent account checking system and method based on block chain and OCR

Abstract

The invention relates to the technical field of financial block chains, and the method comprises the steps: carrying out the format recognition of a heterogeneous bill file according to the heterogeneous bill file submitted by a first participant, and generating a to-be-verified standardized bill data set; uploading a key verification field in the standardized bill data set to a main chain node of a block chain network, generating a corresponding first data hash value through a consensus mechanism, and storing the first data hash value; uploading the service data packet to a side chain storage node associated with the main chain node, and generating a corresponding second data hash value; and based on a purchase order database provided by the second participant, matching the standardized bill data set with the purchase order database in real time. According to the invention, an automatic and credible intelligent reconciliation system is constructed by fusing optical character recognition and block chain evidence storage technologies, and a business closed loop is formed. According to the overall scheme, full-link intelligence and credibility from data acquisition, processing and verification to dispute solution are realized, and the financial cooperation cost and risk of the supply chain are greatly reduced.

Description

Technical Field

[0001] This invention relates to the field of financial blockchain, and in particular to a smart reconciliation system for OCR suppliers. Background Technology

[0002] In traditional supply chain reconciliation, document processing and account reconciliation rely heavily on manual operations. Staff must manually enter and compare invoices provided by suppliers with the buyer's internal orders and receiving records—a tedious and highly repetitive process. This involves not only the digitization of paper documents but also the organization and data extraction of electronic files in different formats.

[0003] Currently, this field generally suffers from low levels of automation in data processing, errors in manual verification, and low efficiency. Data from different parties is isolated, lacking reliable collaborative audit leads. Disputes arising from discrepancies in amount, quantity, or category become lengthy and difficult to trace responsibility. Furthermore, the authenticity and integrity of business data lack effective technical safeguards, posing a risk of tampering and failing to meet the demands of modern enterprises for efficient, transparent, and reliable financial operations. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] In view of the problems mentioned above and / or existing intelligent reconciliation systems for suppliers based on blockchain and OCR, this invention is proposed.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a supplier intelligent reconciliation system based on blockchain and OCR, characterized in that it includes a multi-source data access module, an intelligent parsing and standardization module, a dual-chain collaborative evidence storage module and a rule-driven reconciliation engine module;

[0007] The multi-source data access module includes a function to receive structured data files from the first participant through an encrypted interface, perform feature recognition on the received files, and dynamically allocate them to the OCR recognition or structured parsing pipeline based on the recognition results.

[0008] The intelligent parsing and standardization module works in collaboration with the optical character recognition unit and the data verification and fusion unit. The optical character recognition unit performs text localization and recognition on unstructured bill images based on a convolutional recurrent neural network model. The data verification and fusion unit cleans the fields and normalizes the format of the recognition results and the directly parsed data, and performs logical verification by associating them with the transaction party's evidence storage information database to generate a standardized bill data set with timestamps.

[0009] The dual-chain collaborative evidence storage module integrates a main chain evidence storage unit and a side chain anchoring unit. The main chain evidence storage unit is used to generate an immutable main chain hash record after consensus on the key verification fields in the standardized data set. The side chain anchoring unit is used to package and store the original document, complete parsing log and attached electronic signature on the side chain, and generate a side chain hash index that is cross-verified with the main chain record.

[0010] The rule-driven reconciliation engine module configures business logic rules, performs real-time matching, cross-checking, and comprehensive consistency calculation of the standardized invoice data set and the order dataset provided by the second participating node, and outputs reconciliation results or discrepancy alarm signals.

[0011] As a preferred embodiment of the supplier intelligent reconciliation system based on blockchain and OCR described in this invention, wherein: in the intelligent parsing and standardization module, the data verification and fusion unit is specifically used for:

[0012] Receive the recognized text output by the optical character recognition unit or the data output by the structured parsing pipeline;

[0013] Based on a pre-defined knowledge base of invoice templates and semantic rules, key fields in the identified text are located, extracted, and semantically normalized to eliminate format ambiguity.

[0014] The system calls upon the transaction party identity verification information and historical transaction record hashes stored in the dual-chain collaborative evidence storage module to verify the legality of the extracted bill data and determine the continuity of transactions.

[0015] A unique verification identifier and timestamp are attached to the data items that pass the verification, and a standardized invoice data set with verification tags is generated.

[0016] As a preferred embodiment of the supplier intelligent reconciliation system based on blockchain and OCR described in this invention, the rule-driven reconciliation engine module further includes:

[0017] A dynamically configurable business rule base, wherein the rules stored in the business rule base include, but are not limited to: encoding mapping rules, historical price fluctuation range tolerance rules, transaction quantity cumulative matching rules, and pricing rules based on contract terms;

[0018] A real-time computing unit, which is used for:

[0019] Based on the business rule base, the bill data in the standardized bill data set is intelligently associated and matched with the order dataset provided by the second participating node;

[0020] Once a match is successful, the tolerance parameters agreed upon in the order are automatically retrieved, and cross-checking of the sub-items and the total item is performed based on the overall amount matching.

[0021] The supplier intelligent reconciliation system also includes a consensus-based discrepancy processing module. This module comprises a negotiation interaction unit and an arbitration execution unit. Upon receiving a discrepancy alarm, the negotiation interaction unit establishes a secure communication channel for the first and second participating nodes to exchange negotiation opinions and reach a consensus. When consensus fails, the arbitration execution unit automatically triggers and broadcasts the disputed data packet to the decentralized arbitration network. Based on the smart contract ruling, it enforces the account update and synchronizes the final state to the dual-chain collaborative evidence storage module.

[0022] The first participant is the supplier, and the second participant is the purchaser.

[0023] To address the aforementioned technical problems, this invention provides the following technical solution: a supplier intelligent reconciliation method based on blockchain and OCR, comprising the following steps:

[0024] Based on the heterogeneous invoice file submitted by the first participant, the data is received through a multi-channel data acquisition module, and the format of the heterogeneous invoice file is recognized. Based on the recognition results, the OCR text recognition engine or the structured data parsing engine is called respectively to generate initial structured invoice data.

[0025] The initial structured bill data is standardized, cleaned, and key fields are extracted. Based on the preset transaction party information and historical transaction records, a standardized bill dataset to be verified is generated.

[0026] The key verification fields in the standardized bill dataset are uploaded to the main chain node of the blockchain network, and the corresponding first data hash value is generated and stored through the consensus mechanism; at the same time, the business data packet is uploaded to the side chain storage node associated with the main chain node to generate the corresponding second data hash value.

[0027] Based on the purchase order database provided by the second participant, an automatic comparison model driven by a rule engine is constructed to match the standardized invoice dataset with the purchase order database in real time and calculate the consistency of key fields and the degree of matching of amounts.

[0028] When the matching result's consistency exceeds a preset threshold, the reconciliation is determined to be successful, a reconciliation success instruction is generated, and the final state hash value of this successful reconciliation is updated to the blockchain main chain node.

[0029] When the matching result's similarity is lower than a preset threshold, the rule engine triggers a difference alarm, automatically generates difference prompt information containing specific difference items, and pushes the difference prompt information synchronously to the first participant node and the second participant node.

[0030] The first participating node and the second participating node input their negotiation feedback in response to the difference prompt information, and then initiate the negotiation consensus module to determine the consistency of the feedback from both parties.

[0031] As a preferred embodiment of the supplier intelligent reconciliation method based on blockchain and OCR described in this invention, the standardization cleaning and key field extraction of the initial structured invoice data involves correlation verification based on preset transaction party information and historical transaction records to generate a standardized invoice dataset to be verified, specifically including:

[0032] The initial structured invoice data is processed by filtering invalid characters, unifying formats, and standardizing semantics based on a preset invoice field rule base.

[0033] Extract key verification fields from the standardized data;

[0034] The key verification fields are compared with the preset, on-chain, and stored transaction party identity information and historical performance records to verify the legality of the source of the bill data and the continuity of the transaction.

[0035] Based on the results of the correlation comparison, a standardized invoice dataset with verification identifiers is generated;

[0036] The transaction information includes information on the first participant and information on the second participant.

[0037] As a preferred embodiment of the supplier intelligent reconciliation method based on blockchain and OCR described in this invention, the method includes: uploading key verification fields from the standardized invoice dataset to the main chain node of the blockchain network, generating and storing the corresponding first data hash value through a consensus mechanism; simultaneously, uploading business data packets to the side chain storage node associated with the main chain node to generate the corresponding second data hash value, specifically including:

[0038] The key verification fields are combined with the current timestamp and the digital identity of the transacting party to form a main chain data packet. After being verified by the consensus nodes of the blockchain network, a first data hash value is generated and permanently recorded in the immutable ledger of the main chain.

[0039] The original image containing the heterogeneous ticket file, the initial structured ticket data, the complete data parsing process log, and the associated digital signature file are packaged together to form a business data package;

[0040] By using cross-chain anchoring technology, the business data packet is stored in a sidechain storage node that is uniquely associated with the main chain node, a second data hash value is generated, and the index pointer of the second data hash value is recorded in the main chain data packet.

[0041] As a preferred embodiment of the supplier intelligent reconciliation method based on blockchain and OCR described in this invention, the step of constructing an automatic comparison model driven by a rule engine, which performs real-time matching of the standardized invoice dataset with the purchase order database, and calculates the consistency of key fields and the degree of matching of amounts, specifically includes:

[0042] The rule engine has a pre-configurable matching rule library, which includes exact matching rules, fuzzy matching rules, and tolerance matching rules.

[0043] The automatic comparison model calls the matching rule library to intelligently match the information in the standardized invoice dataset with the order details in the purchase order database;

[0044] After a successful match, the quantity, unit price, and total amount are cross-checked, and the overall amount matching degree C is calculated based on the preset tolerance parameters and weighting coefficients.

[0045] The formula for calculating the overall amount matching degree C is as follows:

[0046]

[0047] Where n is the total number of comparison items, P i For the i-th amount item in the bill, O i w is the base amount item in the order. i The pre-configured weight coefficient for the i-th amount item, Ti, is the dynamic tolerance threshold for the corresponding item, and the function δ(x, T) is:

[0048]

[0049] Record the complete matching path, the rules used, various audit data, and the final calculated overall amount of consistency to generate a structured reconciliation process log;

[0050] The amount items include total items and sub-items;

[0051] The w i satisfy .

[0052] As a preferred embodiment of the supplier intelligent reconciliation method based on blockchain and OCR described in this invention, the method further includes: initiating the negotiation consensus module, determining the consistency of feedback from both parties, and then further comprising:

[0053] If both parties agree, a consensus confirmation instruction is generated. Based on the confirmed corrections, the corresponding records in the standardized invoice dataset and the purchase order database are updated, and the final state hash value of this consensus result is synchronously updated to the blockchain main chain node.

[0054] If the feedback from both parties is inconsistent, the negotiation consensus module will automatically switch to the arbitration stage and broadcast the dispute details, relevant evidence hashes, and opinions from both parties to multiple pre-authorized arbitration nodes.

[0055] The arbitration node independently adjudicates disputes based on a pre-set arbitration rule base and stores the adjudication results on the blockchain for evidence.

[0056] The system receives the majority of valid arbitration awards as the final ruling, generates an arbitration enforcement instruction, forcibly updates the relevant data records, and updates the status hash value of the final ruling to the blockchain main chain node.

[0057] This invention provides the following technical solution: an electronic device, comprising:

[0058] One or more processors;

[0059] A storage device on which one or more programs are stored;

[0060] When the one or more programs are executed by the one or more processors, the one or more processors implement a supplier smart reconciliation method based on blockchain and OCR.

[0061] This invention provides the following technical solution: an electronic device, comprising:

[0062] A computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to implement a blockchain-based and OCR-based intelligent reconciliation method for suppliers.

[0063] This invention integrates optical character recognition (OCR) and blockchain notarization technologies to construct an automated and reliable intelligent reconciliation system. The system can automatically process multi-format invoices, converting them into standardized data. Through a dual-chain structure, it ensures the immutability of key data and full-process traceability, fundamentally guaranteeing the authenticity of business data and the reliability of audits. The built-in rule engine achieves millisecond-level accurate matching and intelligent discrepancy identification, freeing manual labor from tedious verification work and significantly improving reconciliation efficiency and accuracy. When discrepancies arise, the system provides a structured online negotiation and automated arbitration mechanism, leveraging the indisputable nature of blockchain notarization to quickly resolve disputes and form a closed business loop. The overall solution achieves end-to-end intelligence and reliability from data collection, processing, verification to dispute resolution, significantly reducing the cost and risk of supply chain financial collaboration. Attached Figure Description

[0064] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0065] Figure 1 This is an operation flowchart of a supplier intelligent reconciliation system based on blockchain and OCR in Example 1.

[0066] Figure 2 This is a flowchart of the data parsing operation of a supplier intelligent reconciliation method based on blockchain and OCR in Example 2.

[0067] Figure 3 This is a flowchart illustrating the main chain and side chain operation logic of a supplier intelligent reconciliation method based on blockchain and OCR in Example 2.

[0068] Figure 4 This is a flowchart of the automatic difference marking operation based on a rule engine in a supplier intelligent reconciliation method based on blockchain and OCR in Example 2. Detailed Implementation

[0069] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0070] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0071] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0072] Example 1

[0073] Reference Figure 1 This is the first embodiment of the present invention, which provides a supplier intelligent reconciliation system based on blockchain and OCR, comprising:

[0074] Multi-source data access module 100, intelligent parsing and standardization module 200, dual-chain collaborative evidence storage module 300 and rule-driven reconciliation engine module 400;

[0075] The multi-source data access module 100 includes a function to receive structured data files from a first participant through an encrypted interface, perform feature recognition on the received files, and dynamically allocate them to an OCR recognition or structured parsing pipeline based on the recognition results.

[0076] The intelligent parsing and standardization module 200 works in collaboration with the optical character recognition unit 201 and the data verification and fusion unit 202. The optical character recognition unit 201 performs text localization and recognition on unstructured bill images based on a convolutional recurrent neural network model. The data verification and fusion unit 202 performs field cleaning and format normalization on the recognition results and directly parsed data, and performs logical verification by associating with the transaction party's evidence storage information database to generate a standardized bill data set with timestamps.

[0077] In the intelligent parsing and standardization module 200, the data verification and fusion unit 202 is specifically used for:

[0078] Receive the recognized text output by the optical character recognition unit 201 or the data output by the structured parsing pipeline;

[0079] Based on a pre-defined knowledge base of invoice templates and semantic rules, key fields in the identified text are located, extracted, and semantically normalized to eliminate format ambiguity.

[0080] The system calls upon the transaction party identity verification information and historical transaction record hashes stored in the dual-chain collaborative evidence storage module 300 to verify the legality of the extracted invoice data and determine the continuity of transactions.

[0081] A unique verification identifier and timestamp are attached to the data items that pass the verification, and a standardized invoice data set with verification tags is generated.

[0082] The dual-chain collaborative evidence storage module 300 integrates a main chain evidence storage unit 301 and a side chain anchoring unit 302. The main chain evidence storage unit 301 is used to generate an immutable main chain hash record after consensus on the key verification fields in the standardized data set. The side chain anchoring unit 302 is used to package and store the original document, complete parsing log and attached electronic signature on the side chain, and generate a side chain hash index that is cross-verified with the main chain record.

[0083] The rule-driven reconciliation engine module 400 configures business logic rules, performs real-time matching, cross-checking, and comprehensive consistency calculation of the standardized invoice data set and the order dataset provided by the second participating node, and outputs reconciliation results or difference alarm signals.

[0084] The rule-driven reconciliation engine module 400 also includes:

[0085] A dynamically configurable business rule base, wherein the rules stored in the business rule base include, but are not limited to: encoding mapping rules, historical price fluctuation range tolerance rules, transaction quantity cumulative matching rules, and pricing rules based on contract terms;

[0086] A real-time computing unit, which is used for:

[0087] Based on the business rule base, the bill data in the standardized bill data set is intelligently associated and matched with the order dataset provided by the second participating node;

[0088] Once a match is successful, the tolerance parameters agreed upon in the order are automatically retrieved, and cross-checking of the sub-items and the total item is performed based on the overall amount matching.

[0089] The supplier intelligent reconciliation system also includes a consensus-based discrepancy processing module 500. The consensus-based discrepancy processing module 500 includes a negotiation interaction unit 501 and an arbitration execution unit 502. The negotiation interaction unit 501 is used to build a secure communication channel for the first and second participating nodes to exchange negotiation opinions and reach a consensus after receiving a discrepancy alarm. When the consensus fails, the arbitration execution unit 502 automatically triggers and broadcasts the dispute data packet to the decentralized arbitration network, enforces the account update according to the smart contract ruling, and synchronizes the final state to the dual-chain collaborative evidence storage module 300.

[0090] The first participant is the supplier, and the second participant is the purchaser.

[0091] Example 2

[0092] Reference Figures 2 to 4 The second embodiment of the present invention provides a supplier intelligent reconciliation method based on blockchain and OCR, comprising the following steps:

[0093] Based on the heterogeneous invoice files submitted by the first participating party, the data is received through a multi-channel data acquisition module, and the format of the heterogeneous invoice files is recognized. Based on the recognition results, either an OCR text recognition engine or a structured data parsing engine is invoked to generate initial structured invoice data. The structured invoice data recognition method is as follows: Figure 2 As shown;

[0094] In the standardization and cleaning of the initial structured bill data and the extraction of key fields, the data is correlated and verified with preset transaction party information and historical transaction records to generate a standardized bill dataset to be verified, specifically including:

[0095] The initial structured invoice data is processed by filtering invalid characters, unifying formats, and standardizing semantics based on a preset invoice field rule base.

[0096] Extract key verification fields from the standardized data;

[0097] The key verification fields are compared with the preset, on-chain, and stored transaction party identity information and historical performance records to verify the legality of the source of the bill data and the continuity of the transaction.

[0098] Based on the results of the correlation comparison, a standardized invoice dataset with verification identifiers is generated;

[0099] The transaction information includes information on the first participant and information on the second participant.

[0100] The initial structured bill data is standardized, cleaned, and key fields are extracted. Based on the preset transaction party information and historical transaction records, a standardized bill dataset to be verified is generated.

[0101] The key verification fields in the standardized bill dataset are uploaded to the main chain node of the blockchain network, and the corresponding first data hash value is generated and stored through the consensus mechanism. Simultaneously, the business data package is uploaded to the side chain storage node associated with the main chain node to generate the corresponding second data hash value. The main chain and side chain operating logic is as follows: Figure 3 As shown;

[0102] The key verification fields in the standardized bill dataset are uploaded to the main chain node of the blockchain network. A first data hash value is generated and stored through the consensus mechanism. Simultaneously, the business data packet is uploaded to the side chain storage node associated with the main chain node to generate a corresponding second data hash value. Specifically, this includes:

[0103] The key verification fields are combined with the current timestamp and the digital identity of the transacting party to form a main chain data packet. After being verified by the consensus nodes of the blockchain network, a first data hash value is generated and permanently recorded in the immutable ledger of the main chain.

[0104] The original image containing the heterogeneous ticket file, the initial structured ticket data, the complete data parsing process log, and the associated digital signature file are packaged together to form a business data package;

[0105] By using cross-chain anchoring technology, the business data packet is stored in a sidechain storage node that is uniquely associated with the main chain node, a second data hash value is generated, and the index pointer of the second data hash value is recorded in the main chain data packet.

[0106] An automatic comparison model driven by a rules engine is constructed to perform real-time matching between the standardized invoice dataset and the purchase order database, calculating the consistency of key fields and the degree of agreement in amounts. Specifically, this includes:

[0107] The rule engine has a pre-configurable matching rule library, which includes exact matching rules, fuzzy matching rules, and tolerance matching rules.

[0108] The automatic comparison model calls the matching rule library to intelligently match the information in the standardized invoice dataset with the order details in the purchase order database;

[0109] After a successful match, the quantity, unit price, and total amount are cross-checked, and the overall amount matching degree C is calculated based on the preset tolerance parameters and weighting coefficients.

[0110] The formula for calculating the overall amount matching degree C is as follows:

[0111]

[0112] Where n is the total number of comparison items, P i For the i-th amount item in the bill, O i w is the base amount item in the order. i The pre-configured weight coefficient for the i-th amount item, Ti, is the dynamic tolerance threshold for the corresponding item, and the function δ(x, T) is:

[0113]

[0114] Record the complete matching path, the rules used, various audit data, and the final calculated overall amount of consistency to generate a structured reconciliation process log;

[0115] The amount items include total items and sub-items;

[0116] The w i satisfy .

[0117] Based on the purchase order database provided by the second participant, an automatic comparison model driven by a rule engine is constructed to match the standardized invoice dataset with the purchase order database in real time and calculate the consistency of key fields and the degree of matching of amounts.

[0118] When the matching result's consistency exceeds a preset threshold, the reconciliation is determined to be successful, a reconciliation success instruction is generated, and the final state hash value of this successful reconciliation is updated to the blockchain main chain node.

[0119] When the matching result's similarity is lower than a preset threshold, the rule engine triggers a difference alarm, automatically generates difference prompt information containing specific difference items, and pushes the difference prompt information synchronously to the first participant node and the second participant node.

[0120] The first participating node and the second participating node input their negotiation feedback in response to the difference prompt information, and then initiate the negotiation consensus module to determine the consistency of the feedback from both parties.

[0121] The consensus-building module is activated to assess the consistency of feedback from both parties. This process also includes:

[0122] If both parties agree, a consensus confirmation instruction is generated. Based on the confirmed corrections, the corresponding records in the standardized invoice dataset and the purchase order database are updated, and the final state hash value of this consensus result is synchronously updated to the blockchain main chain node.

[0123] If the feedback from both parties is inconsistent, the negotiation consensus module will automatically switch to the arbitration stage and broadcast the dispute details, relevant evidence hashes, and opinions from both parties to multiple pre-authorized arbitration nodes.

[0124] The arbitration node independently adjudicates disputes based on a pre-set arbitration rule base and stores the adjudication results on the blockchain for evidence.

[0125] The system receives the majority of valid arbitration awards as the final ruling, generates an arbitration enforcement instruction, forcibly updates relevant data records, and updates the status hash value of the final ruling to the blockchain main chain node. The arbitration operation logic is as follows: Figure 4 As shown.

[0126] Example 3

[0127] A third embodiment of the present invention provides a supplier intelligent reconciliation method based on blockchain and OCR, comprising:

[0128] To verify the effectiveness and superiority of the "Supplier Intelligent Reconciliation Method and System Based on Blockchain and OCR" described in this invention, a one-quarter field deployment and comparative test were conducted between a medium-sized machinery manufacturing enterprise (as the second participant) and its core component suppliers (as the first participant).

[0129] This enterprise needs to process about 150 invoices from this supplier every month, involving more than 300 purchase orders. The traditional pure manual reconciliation mode takes an average of 5 working days and has a high error rate.

[0130] The test environment was deployed as follows: The core module of the intelligent reconciliation system of this invention was deployed on the enterprise's internal server cluster, including an OCR recognition server, a blockchain test network node, and a reconciliation rule engine. The supplier side accesses the system through a secure API gateway for uploading bills.

[0131] The blockchain network includes a main chain and a side chain storage system anchored to it. The test selected historical transaction data that was completed in the previous quarter and had been manually reviewed and confirmed to be correct as the benchmark data set, totaling 450 invoices (including 300 scanned copies, 100 PDFs, and 50 Excel files) and the corresponding 900 purchase order records.

[0132] The test was divided into two stages:

[0133] In the first stage, this batch of data was processed using the traditional manual process, recording time, manpower, and error situations;

[0134] In the second stage, the same batch of data was processed using the system of this invention, and the indicators of each link of the system were recorded throughout the process.

[0135] During the implementation process, the system first receives the mixed-format bill file package uploaded by the supplier through the encrypted API interface of the multi-source data access module.

[0136] Based on the file MIME type and binary header characteristics, the format routing and distribution unit routes scanned copies and PDFs to the OCR processing queue and directly routes Excel files to the structured parsing queue. The intelligent parsing and standardization module starts to work:

[0137] The optical character recognition unit loads the pre-trained CRNN model to perform text localization and recognition on image bills;

[0138] The data verification and fusion unit cleans the OCR recognition text and Excel parsing data. For example, it uniformly standardizes the recognized "Year 2023" to "2023", and performs fuzzy matching between "P / N: GX-203" and the material code "GX203" in the order library.

[0139] The data verification and fusion unit calls the supplier's unified social credit code hash that has been stored on the blockchain to verify the legality of the bill header and associate it with the supplier's transaction sequence over the past three months to form a standardized bill dataset with timestamps and credibility labels.

[0140] Subsequently, the dual-chain collaborative evidence storage module was activated. The main chain evidence storage unit combined the key verification fields of each invoice (invoice number, total amount, date, supplier ID) to generate a data packet, sorted it through consensus on the blockchain test network, calculated the SHA-256 hash value (i.e., the first data hash value), and uploaded it to the chain.

[0141] Meanwhile, the sidechain anchoring unit packages the corresponding original image of the invoice, the complete JSON format parsing log, and the supplier's digital signature file, compresses them, stores them in the IPFS (InterPlanetary File System) cluster, and submits its content identifier (CID) as the second data hash value to the sidechain, and establishes a bidirectional index with the main chain hash.

[0142] After the evidence is stored, the rule-driven reconciliation engine module begins the core reconciliation operation.

[0143] Its built-in business rule base has been pre-configured with the supplier's specific contract terms, such as "unit price fluctuation tolerance of bolts (model GB-5782) ±3%" and "three batches of invoices are allowed under the same order number." The real-time calculation unit associates and matches the standardized invoice details with the purchase order database. After a successful match, the amount of each detail is audited, and the formula is applied to calculate the overall amount matching degree C.

[0144] Table 1: Efficiency Comparison of Document Processing and Standardization Stages

[0145]

[0146] Table 2: OCR recognition accuracy test (for 300 scanned documents)

[0147]

[0148] Table 3: Blockchain Evidence Storage and Data Verification Performance

[0149]

[0150] Table 4: Core Matching and Auditing Results of the Intelligent Reconciliation Engine

[0151]

[0152] Table 5: Comparison of Time Required for Difference Handling and Dispute Resolution

[0153]

[0154] Table 6: Overall Economic Benefits and Accuracy Analysis

[0155]

[0156] A systematic analysis of the six data tables above clearly demonstrates the inventiveness, novelty, and significant beneficial effects of this invention compared to existing technologies. Firstly, in the initial stages of data input and processing, the high error rate (1.8%) and low standardization (85%) of the traditional model are the root causes of the subsequent complexity in reconciliation.

[0157] As shown in Tables 1 and 2, this invention reduces the data entry error rate by an order of magnitude to 0.15% and improves the field standardization uniformity to 99.7% by integrating a dedicated OCR model with intelligent verification. This fundamental improvement stems directly from the combination of the "convolutional recurrent neural network model" and the "verification of transaction evidence database," which is not simply automation but rather endows the system with cognitive and verification capabilities, ensuring data quality from the source.

[0158] Secondly, in terms of data trust and evidence preservation, traditional centralized databases (Table 3) are subject to internal tampering risks and are difficult to audit and trace.

[0159] While the "dual-chain collaborative evidence storage mechanism" introduced in this invention slightly increases the time required for each evidence storage operation, this is a necessary and minimal cost incurred to achieve extremely strong tamper resistance and traceability. This mechanism stores the key hash on the main chain and the complete data on the side chain, with cross-indexing, enabling any subsequent data verification to be completed within 2 minutes, and ensuring the evidence is technically indisputable. This provides a solid technical foundation for resolving the "conflicting accounts" dilemma in traditional reconciliation, representing an innovation in the methodological structure.

[0160] The improved accuracy of the core reconciliation process is strongly demonstrated by the data in Table 4. Traditional manual matching has an accuracy rate of 94%, but 6% of discrepancies still require significant effort to resolve.

[0161] This invention system improves the matching accuracy to 99.33% and drastically reduces the proportion of discrepancies requiring manual intervention to 0.67%. The key to this leap forward lies in the application of a "rule-driven reconciliation engine," particularly the calculation of the overall amount matching degree C based on preset tolerance parameters and weighting coefficients. For example, in a test, the invoice amount for a certain material was 10,250 yuan, the order base amount was 10,000 yuan, and the preset tolerance threshold T... i It is 300 yuan, with a weight of w iThe value is 0.6. Substituting into the formula, we calculate: δ(∣10250−10000∣,300)=0. This difference term is within the tolerance range and has no negative impact on the overall consistency C, thus avoiding unnecessary difference alarms and demonstrating intelligent fault tolerance capability, which is not available in traditional rigid comparison.

[0162] Finally, in terms of overall process efficiency and economic benefits (Tables 5 and 6), the improvements brought about by this invention are revolutionary.

[0163] The processing time for discrepancies has been shortened from days or even weeks to hours, thanks to the closed-loop online processing flow formed by the built-in "negotiation and interaction unit" and "arbitration enforcement unit".

[0164] Ultimately, the total cost of quarterly reconciliation was significantly reduced from RMB 133,500 to RMB 15,700, and the reconciliation cycle was shortened from 5 days to 0.5 days. These data collectively demonstrate that this invention is not merely a simple digitization of existing processes, but rather a creative solution to systemic pain points in existing technologies, such as low data quality, high trust costs, slow processing efficiency, and difficulty in dispute resolution, through a complete restructuring of the entire process: "OCR intelligent parsing - blockchain trusted evidence storage - rule engine precise auditing - online consensus arbitration." This achieves a fundamental transformation of reconciliation operations from a "labor-intensive, dispute-driven" model to a "technology-intensive, preventative, and collaborative" model.

[0165] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A supplier intelligent reconciliation system based on blockchain and OCR, characterized in that, It includes a multi-source data access module (100), an intelligent parsing and standardization module (200), a dual-chain collaborative evidence storage module (300), and a rule-driven reconciliation engine module (400). The multi-source data access module (100) includes a function to receive structured data files from the first participant through an encrypted interface, perform feature recognition on the received files, and dynamically allocate them to the OCR recognition or structured parsing pipeline based on the recognition results. The intelligent parsing and standardization module (200) works in collaboration with the optical character recognition unit (201) and the data verification and fusion unit (202). The optical character recognition unit (201) performs text localization and recognition on unstructured bill images based on a convolutional recurrent neural network model. The data verification and fusion unit (202) cleans the fields and normalizes the format of the recognition results and the directly parsed data, and performs logical verification by associating with the transaction party's evidence storage information database to generate a standardized bill data set with timestamps. The dual-chain collaborative evidence storage module (300) integrates the main chain evidence storage unit (301) and the side chain anchoring unit (302). The main chain evidence storage unit (301) is used to generate an immutable main chain hash record after consensus on the key verification fields in the standardized data set. The side chain anchoring unit (302) is used to package and store the original document, complete parsing log and attached electronic signature in the side chain, and generate a side chain hash index that is cross-verified with the main chain record. The rule-driven reconciliation engine module (400) configures business logic rules, performs real-time matching, cross-checking and comprehensive consistency calculation of the standardized invoice data set and the order dataset provided by the second participating node, and outputs reconciliation results or difference alarm signals.

2. The supplier intelligent reconciliation system based on blockchain and OCR as described in claim 1, characterized in that: In the intelligent parsing and standardization module (200), the data verification and fusion unit (202) is specifically used for: Receive the recognized text output by the optical character recognition unit (201) or the data output by the structured parsing pipeline; Based on a pre-defined knowledge base of invoice templates and semantic rules, key fields in the identified text are located, extracted, and semantically normalized to eliminate format ambiguity. The system calls upon the transaction party identity verification information and historical transaction record hash stored in the dual-chain collaborative evidence storage module (300) to verify the legality of the extracted bill data and determine the continuity of transactions. A unique verification identifier and timestamp are attached to the data items that pass the verification, and a standardized invoice data set with verification tags is generated.

3. The supplier intelligent reconciliation system based on blockchain and OCR as described in claim 1, characterized in that: The rule-driven reconciliation engine module (400) also includes: A dynamically configurable business rule base, wherein the rules stored in the business rule base include, but are not limited to: encoding mapping rules, historical price fluctuation range tolerance rules, transaction quantity cumulative matching rules, and pricing rules based on contract terms; A real-time computing unit, which is used for: Based on the business rule base, the bill data in the standardized bill data set is intelligently associated and matched with the order dataset provided by the second participating node; Once a match is successful, the tolerance parameters agreed upon in the order are automatically retrieved, and cross-checking of the sub-items and the total item is performed based on the overall amount matching. The supplier intelligent reconciliation system also includes a consensus-based discrepancy processing module (500), which includes a negotiation interaction unit (501) and an arbitration execution unit (502). The negotiation interaction unit (501) is used to build a secure communication channel for the first and second participating nodes to exchange negotiation opinions and reach a consensus after receiving a discrepancy alarm. When the consensus fails, the arbitration execution unit (502) automatically triggers and broadcasts the dispute data packet to the decentralized arbitration network, enforces the account update according to the smart contract ruling, and synchronizes the final state to the dual-chain collaborative evidence storage module (300). The first participant is the supplier, and the second participant is the purchaser.

4. A supplier intelligent reconciliation method based on blockchain and OCR, characterized in that, Includes the following steps: Based on the heterogeneous invoice file submitted by the first participant, the data is received through a multi-channel data acquisition module, and the format of the heterogeneous invoice file is recognized. Based on the recognition results, the OCR text recognition engine or the structured data parsing engine is called respectively to generate initial structured invoice data. The initial structured bill data is standardized and cleaned, and key fields are extracted. Based on the preset transaction party information and historical transaction records, a standardized bill dataset to be verified is generated. The key verification fields in the standardized bill dataset are uploaded to the main chain node of the blockchain network, and the corresponding first data hash value is generated and stored through the consensus mechanism; at the same time, the business data packet is uploaded to the side chain storage node associated with the main chain node to generate the corresponding second data hash value. Based on the purchase order database provided by the second participant, an automatic comparison model driven by a rule engine is constructed to match the standardized invoice dataset with the purchase order database in real time and calculate the consistency of key fields and the degree of matching of amounts. When the matching result's consistency exceeds a preset threshold, the reconciliation is determined to be successful, a reconciliation success instruction is generated, and the final state hash value of this successful reconciliation is updated to the blockchain main chain node. When the matching result's similarity is lower than a preset threshold, the rule engine triggers a difference alarm, automatically generates difference prompt information containing specific difference items, and pushes the difference prompt information synchronously to the first participant node and the second participant node. The first participating node and the second participating node input their negotiation feedback in response to the difference prompt information, and then initiate the negotiation consensus module to determine the consistency of the feedback from both parties.

5. A supplier intelligent reconciliation method based on blockchain and OCR as described in claim 4, characterized in that: In the standardization cleaning and key field extraction of the initial structured bill data, the data is correlated and verified with preset transaction party information and historical transaction records to generate a standardized bill dataset to be verified, specifically including: The initial structured invoice data is processed by filtering invalid characters, unifying formats, and standardizing semantics based on a preset invoice field rule base. Extract key verification fields from the standardized data; The key verification fields are compared with the preset, on-chain, and stored transaction party identity information and historical performance records to verify the legality of the source of the bill data and the continuity of the transaction. Based on the results of the correlation comparison, a standardized invoice dataset with verification identifiers is generated; The transaction information includes information on the first participant and information on the second participant.

6. The supplier intelligent reconciliation method based on blockchain and OCR as described in claim 4, characterized in that: The key verification fields in the standardized bill dataset are uploaded to the main chain node of the blockchain network. A first data hash value is generated and stored through the consensus mechanism. Simultaneously, the business data packet is uploaded to the side chain storage node associated with the main chain node to generate a corresponding second data hash value. Specifically, this includes: The key verification fields are combined with the current timestamp and the digital identity of the transacting party to form a main chain data packet. After being verified by the consensus nodes of the blockchain network, a first data hash value is generated and permanently recorded in the immutable ledger of the main chain. The original image containing the heterogeneous invoice file, the initial structured invoice data, the complete data parsing process log, and the associated digital signature file are packaged together to form a business data package; By using cross-chain anchoring technology, the business data packet is stored in a sidechain storage node that is uniquely associated with the main chain node, a second data hash value is generated, and the index pointer of the second data hash value is recorded in the main chain data packet.

7. A supplier intelligent reconciliation method based on blockchain and OCR as described in claim 4, characterized in that, The automatic comparison model driven by the rule engine is constructed to perform real-time matching between the standardized invoice dataset and the purchase order database, calculating the consistency of key fields and the degree of agreement between amounts, specifically including: The rule engine has a pre-configurable matching rule library, which includes exact matching rules, fuzzy matching rules, and tolerance matching rules. The automatic comparison model calls the matching rule library to intelligently match the information in the standardized invoice dataset with the order details in the purchase order database; After a successful match, the quantity, unit price, and total amount are cross-checked, and the overall amount matching degree C is calculated based on the preset tolerance parameters and weighting coefficients. The formula for calculating the overall amount matching degree C is as follows: Where n is the total number of comparison items, P i For the i-th amount item in the bill, O i w is the base amount item in the order. i The pre-configured weight coefficient for the i-th amount item, Ti, is the dynamic tolerance threshold for the corresponding item, and the function δ(x, T) is: Record the complete matching path, the rules used, various audit data, and the final calculated overall amount of consistency to generate a structured reconciliation process log; The amount items include total items and sub-items; The w i satisfy .

8. A supplier intelligent reconciliation method based on blockchain and OCR as described in claim 4, characterized in that, The consensus-building module, which initiates the negotiation, performs a consistency assessment of the feedback from both parties, and then further includes: If both parties agree, a consensus confirmation instruction is generated. Based on the confirmed corrections, the corresponding records in the standardized invoice dataset and the purchase order database are updated, and the final state hash value of this consensus result is synchronously updated to the blockchain main chain node. If the feedback from both parties is inconsistent, the negotiation consensus module will automatically switch to the arbitration stage and broadcast the dispute details, relevant evidence data hashes, and the opinions of both parties to multiple pre-authorized arbitration nodes. The arbitration node independently adjudicates disputes based on a pre-set arbitration rule base and stores the adjudication results on the blockchain for evidence. The system receives the majority of valid arbitration awards as the final ruling, generates an arbitration enforcement instruction, forcibly updates the relevant data records, and updates the status hash value of the final ruling to the blockchain main chain node.

9. An electronic device, comprising: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the system as described in any one of claims 1-3.

10. A computer-readable storage medium having stored thereon executable instructions that, when executed by a processor, cause the processor to implement the system of any one of claims 1-3.

Images