Intelligent reimbursement full-process monitoring method and system based on multi-modal data fusion

Abstract

The invention provides an intelligent reimbursement whole-process monitoring method and system based on multi-modal data fusion. The method comprises the following steps: matching biological characteristic data of an employee with a preset identity template to obtain an identity verification result; when the result is successful, extracting structured text elements from the reimbursement document; transmitting the employee identity identifier and the structured text element in the identity verification result to a preset cloud database to generate a data association packet; combining the data association packet with the structured text elements to generate accounting registration entries; when the similarity of the comparison result of the structured text elements and the historical accounting registration items exceeds a similar threshold value and it is detected that approved reimbursement receipts which are not submitted exist in a preset period, a repeated reimbursement alarm and a non-reimbursement prompt are generated; and when a single amount in the fund flow data exceeds an amount threshold value, a large amount expenditure early warning signal is generated. According to the invention, accurate capture of key financial information is realized, and the anti-counterfeiting capability, compliance control efficiency and fund risk interception real-time performance of the reimbursement process are improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent reimbursement process monitoring technology, and in particular to an intelligent reimbursement process monitoring method and system based on multimodal data fusion. Background Technology

[0002] As enterprises deepen their digital transformation, verifying the authenticity of invoices and managing risks throughout the financial reimbursement process have become crucial for ensuring compliance. In intelligent reimbursement systems, efficiently and accurately verifying the authenticity of various electronic or paper invoices, and achieving full-process traceability monitoring and dynamic risk warnings from invoice submission to fund payment, has become an important technological requirement for improving financial management efficiency and risk control capabilities.

[0003] The current mainstream solution is an automated invoice verification scheme based on a combination of deep learning and optical character recognition (OCR) technology. This scheme extracts invoice image features through convolutional neural networks, structures the content using OCR, and relies on a centralized database to compare invoice metadata to determine validity. Simultaneously, it records the invoice's circulation status to create a tracking log for auditing. Building upon this, the new solution integrates a standardized accounting module to enhance the accurate extraction and consistency monitoring of key elements, and incorporates intelligent programs to effectively identify and intercept duplicate and missed reimbursements after online approval. However, existing solutions still have inherent limitations, including the difficulty of handling recognition errors caused by complex formats or overlapping content due to reliance on a single OCR engine; the vulnerability of invoice authenticity to forgery or internal tampering due to reliance on local database comparison; and the fact that circulation tracking is mostly post-event recording, lacking efficient full-process dynamic monitoring capabilities, especially in real-time early warning of abnormal large expenditures. Summary of the Invention

[0004] This invention provides a method and system for intelligent full-process monitoring of reimbursement based on multimodal data fusion, which solves the problems in the prior art, such as the difficulty of recognition errors caused by complex formats or superimposed content due to reliance on a single optical character recognition engine; the reliance on local database comparison, which makes the authenticity of invoices susceptible to forgery or internal tampering; and the fact that circulation tracking is mostly recorded after the fact, lacking efficient full-process dynamic monitoring capabilities, especially the difficulty in real-time early warning of abnormal large expenditures.

[0005] In a first aspect, the present invention provides an intelligent reimbursement process monitoring method based on multimodal data fusion, comprising: Obtain the employee's biometric data, match the biometric data with a preset identity template, and obtain the identity verification result; When the authentication result is passed, the physical image of the expense report is scanned using a pre-set standardized processing procedure to extract structured text elements from the physical image. The employee identification and the structured text elements in the authentication result are transmitted to a preset cloud database to generate a data association package; The data association package is combined with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries; When the similarity between the structured text element and the historical accounting entry exceeds a preset similarity threshold and an unsubmitted approved expense reimbursement document is detected within a preset period, a duplicate expense reimbursement alarm and an unreimbursed expense reminder are generated and sent to the terminal associated with the employee's identity. The system uses a pre-set early warning module to monitor the cash flow data in the accounting entries. When a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated.

[0006] Optionally, the employee's biometric data is acquired, and the biometric data is matched with a preset identity template to obtain an identity verification result, including: The biometric data of employees is collected using the contact sensing surface of a pre-installed biometric data collection device. The biometric data is divided into key regions to obtain effective feature regions; The set of topological feature points in the effective feature region is spatially mapped to the set of registration feature points in the preset identity template, and the spatial distribution overlap is calculated based on the mapping result. When the spatial distribution overlap reaches a preset overlap threshold, an identity verification result is generated.

[0007] Optionally, when the authentication result is successful, a pre-defined standardized processing procedure is used to scan the physical image of the expense report to extract structured text elements from the physical image, including: When the authentication result is successful, a pre-set standardized processing procedure is used to respond to the authentication result, and the physical image of the expense report is processed for illumination compensation to generate a uniformly illuminated image. Locate the boundaries of the text regions in the uniformly illuminated image to obtain a set of text regions; The optical characters in each independent region of the text region set are converted using preset character conversion rules to generate the original character sequence; The original character sequence is parsed based on a preset set of accounting element rules to obtain the amount and date information; The monetary value and the date information are combined to generate structured text elements.

[0008] Optionally, the employee identification identifier and the structured text elements in the authentication result are transmitted to a preset cloud database to generate a data association package, including: Extract the employee identification identifier from the authentication result; The structured text elements are encapsulated and marked to generate standardized data packets; The employee identification identifier is bidirectionally bound to the standardized data packet to generate an initial associated data unit; The initial associated data unit is transmitted to a preset cloud database using a preset wireless transmission protocol, triggering a preset timestamp generator in the preset cloud database to add a timestamp identifier to the initial associated data unit, thereby generating an initial associated data unit with the added timestamp identifier. The initial associated data units with added timestamp identifiers are packaged to generate a data association package.

[0009] Optionally, the data association package is combined with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries, including: The employee identification in the data association package, the monetary value in the structured text element, and the date information in the structured text element are mapped to generate a basic accounting unit. The basic accounting units are formatted using a pre-defined accounting standardization template to generate standardized accounting entries. Activate the historical storage engine in the preset cloud database, and input the standardized accounting entries into the storage area corresponding to the historical storage engine to generate historical accounting entries.

[0010] Optionally, when the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold and an unsubmitted approved expense reimbursement document is detected within a preset period, a duplicate reimbursement alarm and an unsubmitted expense reminder are generated and sent to the terminal associated with the employee's identity, including: Extract key fields from the structured text elements to generate the current reimbursement feature set; Retrieve the historical feature set that matches the employee's identity in the historical accounting entries, calculate the field overlap between the current reimbursement feature set and the historical feature set, and generate a duplicate reimbursement alarm when the field overlap exceeds a preset similarity threshold; The preset approvals are scanned through the expense reimbursement document database. The expense reimbursement documents in the database with the status of "not submitted" are associated with the employee identity identifiers corresponding to the expense reimbursement documents with the status of "not submitted" to generate an unsubmitted expense reminder. The duplicate expense report alarm and the unreported expense reminder are encapsulated to generate a terminal notification package, and the terminal notification package is sent to the terminal associated with the employee's identity using a preset communication interface.

[0011] Optionally, a pre-set early warning module is used to monitor the cash flow data in the accounting entries. When a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated, including: The fund flow data in the accounting entries is parsed to obtain a flow dataset containing multiple transaction records; Extract the amount field of each transaction record in the transaction data set, compare the value of each amount field with a preset amount threshold, and when the value of the amount field exceeds the preset amount threshold, mark the transaction record corresponding to the amount field as an abnormal transaction record. The abnormal transaction records are associated with the accounting entries to which they belong, generating a large expenditure warning signal.

[0012] Secondly, the present invention provides an intelligent reimbursement process monitoring system based on multimodal data fusion, comprising: The matching module is used to acquire the employee's biometric data, match the biometric data with a preset identity template, and obtain the identity verification result; The extraction module is used to scan the physical image of the expense report using a preset standardized processing program when the authentication result is passed, so as to extract structured text elements from the physical image. The transmission module is used to transmit the employee identity identifier and the structured text elements in the authentication result to a preset cloud database to generate a data association package; The combination module is used to combine the data association package with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries. The generation module is used to generate duplicate reimbursement alarms and non-reimbursement reminders and send them to the terminal associated with the employee's identity when the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold and an unsubmitted approved reimbursement document is detected within a preset period. The monitoring module is used to monitor the cash flow data in the accounting entries using a pre-set early warning module. When the amount of a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated.

[0013] Thirdly, the present invention provides a computing device, including a processor and a memory, wherein the memory stores a computer program, and the processor is configured to run the computer program to execute the intelligent reimbursement process monitoring method based on multimodal data fusion as described in any of the first aspects.

[0014] Fourthly, the present invention provides a computer storage medium storing computer program instructions thereon, wherein the computer program instructions, when executed by a processor, implement the intelligent reimbursement process monitoring method based on multimodal data fusion as described in any one of the first aspects.

[0015] This invention verifies employee identity by integrating biometric data, ensuring the legitimate authority of reimbursement operators from the outset. It intelligently scans and extracts structured text elements from the physical images of reimbursement documents through a pre-set standardized processing program, achieving precise capture of key financial information. It generates data association packages from a cloud database and combines them to create standardized accounting entries, constructing a unified and standardized financial data foundation. By comparing the similarity between structured text elements and historical accounting entries in real time, combined with status scanning of approved documents, it proactively identifies the risk of duplicate reimbursements and missed reimbursements, generating targeted alarms and reminders. A warning module dynamically monitors large expenditures in the cash flow data, triggering warning signals immediately. This end-to-end collaboration significantly improves the anti-counterfeiting capabilities, compliance control efficiency, and real-time interception of financial risks in the reimbursement process.

[0016] Furthermore, by extracting key fields from structured text elements to generate a current expense reimbursement feature set, the core financial attributes are accurately located. By retrieving historical feature sets associated with employee identities and calculating field overlap, duplicate expense reimbursement behavior is quantitatively identified, generating blocking alerts. By scanning unsubmitted documents in a pre-set approved document database and associating them with corresponding employee identity identifiers, the responsible party for missed expense reimbursements is proactively located, generating driving reminders. Alarms and reminders are packaged into terminal notification packets and pushed to employee terminals via a pre-set communication interface. This mechanism proactively intercepts duplicate expense reimbursements and accurately urges the completion of missed reimbursements, eliminating the risk of financial loss at the source of operations and strengthening process integrity.

[0017] These or other aspects of the invention will become more apparent from the following description of the embodiments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart illustrating an intelligent reimbursement process monitoring method based on multimodal data fusion, provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of an intelligent reimbursement process monitoring system based on multimodal data fusion provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a computing device provided in an embodiment of the present invention. Detailed Implementation

[0020] To enable those skilled in the art to better understand the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings.

[0021] In some of the processes described in the specification, claims, and accompanying drawings of this invention, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Figure 1 A flowchart of an intelligent reimbursement process monitoring method based on multimodal data fusion is provided in this embodiment of the invention, as shown below. Figure 1 As shown, the method includes: In scenarios involving the verification of the authenticity of receipts, traditional reimbursement systems suffer from several technical deficiencies: First, manual entry of paper receipts is prone to errors, leading to chaotic accounting records; second, identity verification relies on passwords or employee badges, which are easily misused, posing a risk of fund theft; third, decentralized systems cannot link employee identities with reimbursement behavior, resulting in frequent duplicate and missed reimbursements; and fourth, traditional early warning mechanisms can only detect large abnormal expenditures after the fact and lack real-time interception capabilities. To address these issues, the research and development approach of this invention is as follows: A multimodal data fusion-driven, end-to-end intelligent monitoring system is constructed. Biometric matching serves as the starting point of the process to ensure the authenticity of the operator's identity. Pre-set standardized processing procedures automatically scan physical images and extract structured text elements to solve the problem of manual data entry errors. Innovatively, identity identifiers are bound to text elements and transmitted to the cloud to generate data association packages, achieving data traceability that integrates the person and the document. Standardized accounting entries are generated in the cloud and stored as historical benchmarks, establishing a standardized data source for dynamic comparison. Based on the similarity analysis between structured text and historical entries, duplicate reimbursements are proactively identified. Simultaneously, scanned and approved documents that have not been submitted to the database to locate missed reimbursements. Dual-path generation of targeted alarms and reminders is pushed to the responsible person's terminal. Combined with a pre-set early warning module, real-time monitoring of fund flows is conducted, and large abnormal expenditures are immediately intercepted through threshold comparison. This method forms a closed-loop control chain from identity authentication and anti-counterfeiting, automated document processing, standardized data association, to proactive risk monitoring, completely solving the problems of data silos, delayed response, and regulatory blind spots in traditional reimbursement processes. Based on this, the present invention provides an intelligent end-to-end monitoring method for expense reimbursement based on multimodal data fusion, such as... Figure 1 ,include: Step 101: Obtain the employee's biometric data, match the biometric data with a preset identity template, and obtain the identity verification result.

[0024] In this step, biometric data refers to the raw information of human physiological characteristics collected by sensors, including digital features used for identity authentication such as fingerprint ridge distribution and iris texture; the preset identity template refers to the employee biometric baseline dataset pre-registered in the system, which includes comparison criteria such as feature point spatial coordinates; the matching operation refers to the process of calculating the spatial overlap of feature points between real-time biometric data and identity template, which is achieved through coordinate mapping; the identity verification result refers to the output data containing verification status and employee identity code, which is used for access control in subsequent processes.

[0025] In this embodiment of the invention, firstly, biometric data such as employee fingerprints or irises are acquired through a biometric acquisition device. Secondly, the acquired biometric data is matched one-to-one with the spatial distribution of feature points in the identity template pre-stored in the system. Finally, when the overlap ratio of matching points reaches a preset security threshold, an identity verification result containing employee identity identifiers and verification status is generated.

[0026] Step 102: When the authentication result is passed, the physical image of the expense report is scanned using a preset standardized processing procedure to extract structured text elements from the physical image.

[0027] In this step, the pre-set standardized processing program refers to the system's built-in image processing and text extraction module, which achieves standardized parsing of documents through illumination compensation and region segmentation; the scanning operation refers to the physical action of controlling optical equipment to photograph paper documents and digitize them; and the structured text elements refer to the formatted financial data extracted from the image, which at least includes monetary value and transaction date information.

[0028] In this embodiment of the invention, when the identity verification result is passed, a preset standardized processing program is first started to control the image scanning device to capture the physical image of the reimbursement document. Then, illumination compensation and text region segmentation processing are performed on the image. Next, structured text elements such as amount, date information and other information are extracted through optical character conversion technology. Finally, a set of elements that conforms to the accounting rules is output.

[0029] Step 103: Transmit the employee identity identifier and the structured text elements from the authentication result to a preset cloud database to generate a data association package.

[0030] In this step, the employee identification code refers to the unique digital code that binds to the employee and is used to associate the person responsible for the reimbursement behavior; the preset cloud database refers to the pre-configured remote data storage system that provides data association and historical entry storage functions; the data association package refers to the data unit that encapsulates the employee identification code and structured text elements, with an additional timestamp to ensure time sequence.

[0031] In this embodiment of the invention, the employee identity identifier is first parsed from the authentication result. Then, the structured text elements are encrypted and encapsulated with type marking. Next, the employee identity identifier is bidirectionally bound to the encapsulated data packet to generate an initial association unit. Finally, the data is transmitted to the cloud database through a wireless communication protocol and a timestamp is added to form a data association packet.

[0032] Step 104: Combine the data association package with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries.

[0033] In this step, the combined operation refers to the process of mapping identity identifiers and financial elements to fields according to accounting rules in the cloud; accounting entry refers to a formatted record that conforms to the accounting subject coding standard and contains complete accounting data including the applicant's amount and date; historical accounting entry refers to accounting entries that have been marked with historical tags and stored in the persistent storage area.

[0034] In this embodiment of the invention, firstly, the data association package is decapsulated and separated in the cloud database to extract employee identification and structured text elements. Secondly, the identification and the amount and date in the elements are mapped to generate basic accounting units. Then, the accounting standardization template is called to perform format conversion. Finally, the standardized accounting entries are written to the storage area through the historical storage engine and historical tags are added to generate historical accounting entries.

[0035] Step 105: When the similarity between the structured text element and the historical accounting entry exceeds a preset similarity threshold and an unsubmitted approved expense reimbursement document is detected within a preset period, a duplicate expense reimbursement alarm and an unreimbursed expense reminder are generated and sent to the terminal associated with the employee's identity.

[0036] In this step, the comparison result refers to the calculated result of the numerical overlap ratio of the same fields between the current reimbursement feature set and the historical feature set; the duplicate reimbursement alarm refers to the blocking instruction generated when the overlap exceeds the threshold; the non-reimbursement reminder refers to the reminder notice generated after associating the unsubmitted documents with the responsible person; and the terminal refers to the mobile communication device held by the employee, used to receive the warning information.

[0037] In this embodiment of the invention, firstly, key fields in the structured text elements are extracted to generate the current expense reimbursement feature set. Secondly, historical feature sets with the same employee identity are retrieved from historical accounting entries. Then, the numerical overlap ratio of the same fields in the two feature sets is calculated. When the overlap ratio exceeds a preset similarity threshold, a duplicate expense reimbursement alarm is generated. At the same time, the unsubmitted documents are scanned and filtered through the approval document database and associated with the corresponding employee identity to generate an unreimbursed reminder. Finally, the alarm and reminder are packaged and sent to the employee's mobile terminal.

[0038] Step 106: Use the pre-set early warning module to monitor the cash flow data in the accounting entries. When the amount of a single transaction in the cash flow data exceeds the preset amount, generate a large expenditure early warning signal.

[0039] In this step, the pre-set early warning module refers to the system's built-in fund monitoring engine, which identifies abnormal transactions through threshold comparison; fund flow data refers to the sequence of transaction amounts recorded in the accounting entries; monitoring operation refers to the process of scanning transaction data one by one and performing numerical comparison; preset amount value refers to the critical amount parameter that triggers a large amount alarm; large expenditure early warning signal refers to the abnormal transaction alarm instruction with an attached risk level label.

[0040] In this embodiment of the invention, firstly, the fund flow records in the accounting entries are parsed to obtain multiple transaction data. Secondly, the transaction amount field value is extracted one by one and compared with a preset amount threshold. When a transaction exceeding the threshold is detected, it is marked as an abnormal record. Then, a level label is generated according to the alarm level matching rule based on the threshold exceedance. Finally, the abnormal record is associated with the source accounting entry and the level label is added to generate a large expenditure alarm signal.

[0041] For example, employees submit biometric data using a fingerprint sensor. The system matches this data with pre-stored identity templates, returns an authentication result, and so on. Upon successful authentication, a standardized process controls a scanner to capture an image of the invoice. After illumination correction, the amount and date are extracted to form structured text elements. The system encrypts and binds the employee's ID to these text elements, transmitting the data wirelessly to the cloud to generate a timestamped data packet. The cloud unpacks the packet, mapping the employee's ID to the amount and date fields, converting it into standard accounting entries, and storing it in the historical database. The system compares the amount and date similarity between new and old entries in real time. When duplicate features are detected, an alert is sent to the employee's mobile phone, and an unsubmitted document database is scanned to generate a reminder. The early warning module continuously monitors cash flow, immediately adding a risk level marker and generating an early warning signal upon detecting a single expenditure exceeding a threshold.

[0042] This invention ensures the authenticity of the operator's identity from the source through biometric authentication, eliminating the risk of fraudulent claims; it automatically extracts structured financial data using standardized image processing, eliminating manual data entry errors; it constructs a data package with strong correlation between identity identifiers and financial elements in the cloud, enabling full-process data traceability; it proactively intercepts duplicate reimbursements based on historical item similarity comparison, and simultaneously scans unsubmitted documents to generate targeted reminders; it monitors fund flows in real time to trigger large transaction alerts, forming a full-chain defense system covering identity authentication, document processing, data association, and risk monitoring, significantly improving the anti-fraud capabilities, data standardization, and real-time risk response of the reimbursement process.

[0043] To address the vulnerability of traditional identity authentication to forgery, this step involves collecting biometric data via a contact-sensing surface, dividing the data into effective feature regions, and calculating the overlap degree based on topological feature point spatial mapping to achieve high-precision identity verification. This invention provides a specific embodiment: Step 101 involves acquiring the employee's biometric data, matching the biometric data against a preset identity template to obtain the identity verification result, specifically including the following steps: Step 111: Collect employees' biometric data using the contact sensor surface of the pre-set biometric data collection device.

[0044] In this step, the pre-set biometric acquisition device refers to the physical sensing device integrated into the system, which is used to capture human biometric signals through a contact sensing surface; the contact sensing surface refers to the area on the surface of the device that is in direct contact with the human body, and collects feature images through capacitive or optical sensing principles.

[0045] In this embodiment of the invention, the employee's finger pressing action is first received through the contact sensing surface of the preset biometric acquisition device, then the raw biometric data containing the fingerprint ridge distribution is collected, and finally the collected raw data is transmitted to the processing unit.

[0046] Step 112: Divide the biometric data into key regions to obtain effective feature regions.

[0047] In this step, the key region segmentation process refers to the process of identifying the effective fingerprint region based on the ridge direction, which is used to eliminate edge distortion caused by finger pressure offset; the effective feature region refers to the image block that retains the complete ridge texture after segmentation, containing biometric features for comparison.

[0048] In this embodiment of the invention, firstly, image edge detection is performed on the biometric data; secondly, the effective recognition region is segmented based on the ridge continuity feature; and finally, edge noise interference is eliminated to generate an effective feature region with a complete ridge structure.

[0049] Step 113: Map the set of topological feature points in the effective feature region to the set of registered feature points in the preset identity template, and calculate the spatial distribution overlap based on the mapping result.

[0050] In this step, the topological feature point set refers to the set of bifurcation points and endpoint coordinates extracted from the effective feature region, which constitutes the basis for fingerprint comparison; the registration feature point set refers to the baseline feature point coordinate data stored in the identity template during employee pre-registration; the spatial location mapping process refers to the operation of unifying the coordinate system of the real-time feature point set and the registration point set; the mapping result refers to the spatial location relationship data of the two feature point sets after coordinate alignment; and the spatial distribution overlap refers to the percentage of the number of coordinate overlap points after mapping out of the total number of feature points.

[0051] In this embodiment of the invention, firstly, the bifurcation points and endpoint coordinates of the topological feature point set are located within the effective feature area. Secondly, the feature point set is aligned with the registered feature point set in the preset identity template in terms of spatial coordinates. Finally, the spatial distribution overlap is generated by calculating the ratio of the number of overlapping points of the two feature point sets after alignment to the total number of points.

[0052] Step 114: When the spatial distribution overlap reaches a preset overlap threshold, generate an identity verification result.

[0053] In this step, the preset overlap threshold refers to the critical proportion parameter for determining successful identity matching.

[0054] In this embodiment of the invention, the spatial distribution overlap is first compared with a preset overlap threshold. Then, when the overlap reaches or exceeds the threshold, it is determined to be a successful match. Finally, an identity verification result containing employee identification and verification status is generated.

[0055] This invention employs a contact-sensing surface to accurately collect raw biometric data, utilizes key region segmentation to eliminate environmental interference and extract effective feature regions, achieves high-precision biometric authentication based on the spatial location mapping of topological feature points, and finally generates a reliable identity verification result by judging the overlap degree threshold. This process significantly improves the anti-counterfeiting capability and recognition accuracy of identity authentication, establishing a secure and reliable initial barrier for the reimbursement process.

[0056] To improve the accuracy of digitizing paper documents, this step performs illumination compensation to generate a uniform image in response to the identity verification result. After locating the text area, character conversion rules are applied to parse the core elements and construct structured text data. This invention provides a specific embodiment: Step 102, when the identity verification result is successful, a preset standardized processing program is used to scan the physical image of the expense report document to extract structured text elements from the physical image. This specifically includes the following steps: Step 201: When the authentication result is passed, the system uses a pre-set standardized processing procedure to respond to the authentication result and performs illumination compensation processing on the physical image of the expense report to generate a uniformly illuminated image.

[0057] In this step, illumination compensation processing refers to the technique of eliminating uneven illumination by adjusting the brightness values ​​of image pixels, which is used to solve the text recognition obstacles caused by differences in brightness in the shooting environment; uniform illumination image refers to a standardized image whose pixel brightness variance is lower than a threshold after compensation processing, and has uniform background light distribution characteristics.

[0058] In this embodiment of the invention, firstly, a pre-set standardized processing program responds to the authentication pass status and initiates a scanning command; secondly, it controls the image acquisition device to capture a physical image of the reimbursement document; then, it performs light intensity equalization processing on the image to eliminate shadow and reflection interference; and finally, it outputs a uniformly illuminated image with even light distribution.

[0059] Step 202: Locate the boundaries of the text regions in the uniformly illuminated image to obtain a set of text regions.

[0060] In this step, the text region boundary refers to the circumscribed polygonal outline of continuous text blocks in the image, a closed geometric shape generated by the edge detection algorithm; the text region set refers to the text block coordinate dataset formed after boundary localization, containing the outline coordinate information of each independent region.

[0061] In this embodiment of the invention, the color contrast distribution characteristics of a uniformly illuminated image are first analyzed, the outline boundaries of continuous text blocks are identified, adjacent text regions are then connected to form logical units, and finally a set of text regions containing all text regions is generated.

[0062] Step 203: Apply preset character conversion rules to convert the optical characters of each independent region in the text region set to generate the original character sequence.

[0063] In this step, the preset character conversion rules refer to the character recognition template library pre-stored in the system, which includes matching criteria such as standard glyph contour features; independent regions refer to a single text block with a closed boundary in the text region set, corresponding to discrete text units on the document; optical characters refer to the set of text pixels presented in the image, which need to be converted into computer-encoded characters; the conversion operation refers to the process of matching the character image with the rule library template and outputting the encoding; the original character sequence refers to the character encoding stream sorted according to the original position after conversion, preserving the spatial order of the text on the document.

[0064] In this embodiment of the invention, firstly, the glyph matching template in the preset character conversion rule library is loaded; secondly, the single character image is segmented for each independent region in the text region set; then, the character image and the template are matched for contour similarity; and finally, the recognized original character sequence is output.

[0065] Step 204: Parse the original character sequence based on the preset accounting element rule set to obtain the amount and date information.

[0066] In this step, the preset accounting element rule set refers to the rule base that defines the format of core elements such as amount and date, including features such as currency symbol position and date separator; the parsing operation refers to the process of scanning character sequences according to the rule set pattern and extracting target fields.

[0067] In this embodiment of the invention, the amount and date format pattern defined by the preset accounting element rule set is first invoked, then the original character sequence is scanned to locate the field that matches the combination of currency symbol numbers, then the continuous numerical information is extracted to verify the legality of the date, and finally the standardized amount value and date information are obtained.

[0068] Step 205: Combine the monetary value and the date information to generate structured text elements.

[0069] In this step, the combination operation refers to logically associating and encapsulating the parsed elements according to the accounting data structure.

[0070] In this embodiment of the invention, a logical relationship between monetary value and date information is first established, then the numerical value and timestamp field are encapsulated according to the accounting data structure, and finally a structured text element containing core elements is generated.

[0071] This invention improves text recognition reliability by generating standardized images through illumination compensation, accurately locating text region boundaries to ensure complete element extraction, applying pre-set character conversion rules to achieve high-precision optical character recognition, and intelligently parsing core financial elements based on an accounting rule set to ultimately generate structured text data. This process significantly improves the accuracy of document information digitization and establishes a standardized data foundation for subsequent financial processing.

[0072] To address the issues of data transmission security and time-series traceability, this step extracts the identity identifier and encapsulates the tagged text elements. After two-way binding, a timestamp is added via wireless transmission, and finally, an auditable data association package is generated. This invention provides a specific embodiment: Step 103, transmitting the employee identity identifier and the structured text elements from the authentication result to a preset cloud database to generate a data association package, specifically includes the following steps: Step 301: Extract the employee identity identifier from the authentication result.

[0073] In this step, the extraction operation refers to the process of separating the target field from the composite data structure, which is achieved by parsing the field address mapping table.

[0074] In this embodiment of the invention, the data structure of the authentication result is first parsed, the storage field of the employee identity identifier is located, and finally the identifier is extracted through a data reading operation.

[0075] Step 302: Encapsulate and mark the structured text elements to generate standardized data packets.

[0076] In this step, encapsulation tagging refers to the technology of adding metadata tags and check codes to data, including protection elements such as type tags and encryption identifiers; standardized data packets refer to data units that conform to a preset transmission format, containing a three-layer structure of payload data, type tags, and check codes.

[0077] In this embodiment of the invention, the amount and date fields of the structured text elements are first loaded, then a type marker and an encryption check code are added, and finally the data is encapsulated into a standardized data packet conforming to the transmission protocol.

[0078] Step 303: Bind the employee identification to the standardized data packet in both directions to generate an initial associated data unit.

[0079] In this step, two-way binding refers to the technology of establishing mutual indexing between two data entities, and realizing association and tracing by generating two-way pointers; the initial associated data unit refers to the transmission carrier after binding employee identity identifiers and standardized data packets, which has two-way indexing characteristics.

[0080] In this embodiment of the invention, firstly, an index relationship between employee identification and standardized data packets is established; secondly, a bidirectional reference pointer is generated; and finally, an initial associated data unit containing the bidirectional pointer is formed.

[0081] Step 304: The initial associated data unit is transmitted to the preset cloud database using a preset wireless transmission protocol, triggering the preset timestamp generator in the preset cloud database to add a timestamp identifier to the initial associated data unit, thereby generating the initial associated data unit with the added timestamp identifier.

[0082] In this step, the preset wireless transmission protocol refers to the data communication specifications pre-configured by the system, including transmission encryption mechanisms and error retransmission rules; the preset timestamp generator refers to the precision clock module built into the cloud database, which provides millisecond-level time recording function; and the timestamp identifier refers to a digital tag that accurately records the arrival time of data, in the format of year, month, day, hour, minute, second, and millisecond.

[0083] In this embodiment of the invention, the initial associated data unit is first transmitted to the cloud database via a wireless network, then the database's built-in timestamp generator is triggered, and finally, a timestamp identifier accurate to the millisecond level is written into the header of the data unit.

[0084] Step 305: Package the initial associated data units after adding the timestamp identifier to generate a data association package.

[0085] In this step, the packaging operation refers to the encapsulation process of compressing and encoding data units and adding verification fields.

[0086] In this embodiment of the invention, the data unit with added timestamp is first compressed and encoded, then a data verification field is added, and finally it is encapsulated to generate a data association packet with a complete transmission structure.

[0087] This invention ensures operational traceability by accurately extracting employee identification identifiers, guarantees data transmission integrity using tagging and encapsulation technology, establishes a strong link between employees and documents through an innovative two-way binding mechanism, and solidifies the operation sequence using timestamps, ultimately generating a standardized data association package. This process constructs a secure transmission system that binds identity, documents, and time in three dimensions, providing an auditable data foundation for cloud processing.

[0088] To eliminate differences in accounting entry formats, this step maps identity identifiers, amounts, and dates to generate basic accounting units. After converting the format using a standardized template, the historical storage engine is activated to generate persistent historical accounting entries. This invention provides a specific embodiment: Step 104, in the preset cloud database, combines the data association package with the structured text elements to generate standardized accounting entry entries and stores them as historical accounting entry entries, specifically including the following steps: Step 401: Map the employee identification in the data association package, the monetary value in the structured text element, and the date information in the structured text element to generate a basic accounting unit.

[0089] In this step, the field mapping operation refers to the process of mapping source data fields to target fields according to preset rules, including establishing matching relationships between field names and types; the basic accounting unit refers to the initial data set consisting of the applicant's expenditure amount and transaction date, which serves as the basic carrier for generating accounting entries.

[0090] In this embodiment of the invention, the employee identity identifier is first extracted by parsing the data association package. Then, the amount value and date information fields are located from the structured text elements. Next, the identity identifier is mapped to the applicant field, the amount value is mapped to the expenditure amount field, and the date information is mapped to the transaction date field. Finally, the data is combined to generate a basic accounting unit containing a triplet of the applicant's expenditure amount and transaction date.

[0091] Step 402: Apply the preset accounting standardization template to convert the format of the basic accounting unit and generate standardized accounting entry.

[0092] In this step, the preset accounting standardization template refers to the data structure specifications that define the accounting subject coding, currency format, and date specifications; the format conversion operation refers to the standardization processing performed on the data according to the template rules, including actions such as adding currency symbols and reconstructing date formats.

[0093] In this embodiment of the invention, the accounting subject coding rules in the preset accounting standardization template are first invoked, the expenditure amount in the basic accounting unit is converted into a standard currency format, the transaction date is then unified into an international date format, and finally the fields are reorganized according to the data structure defined in the template to generate standardized accounting entry.

[0094] Step 403: Activate the historical storage engine in the preset cloud database, and input the standardized accounting entries into the storage area corresponding to the historical storage engine to generate historical accounting entries.

[0095] In this step, the historical storage engine refers to the component in the cloud database responsible for persistent storage, providing data writing and version management functions; the storage area refers to the dedicated data partition managed by the engine that has historical version tags.

[0096] In this embodiment of the invention, a storage command is first sent to the cloud database to activate the historical storage engine. Then, standardized accounting entries are transmitted to the storage partition corresponding to the engine. Next, historical attribute tags are added to the entries in the storage partition. Finally, historical accounting entries with historical tags are generated.

[0097] This invention constructs standardized basic accounting units through field mapping, enforces uniform currency formats and date standards using preset templates, activates a historical storage engine to persist entries and add historical tags, forming a standardized and traceable financial data chain. This process completely eliminates the format differences caused by manual data entry, establishing a highly consistent data foundation for risk monitoring.

[0098] To proactively intercept duplicate expense reports and fix process vulnerabilities, this step extracts key fields to generate a feature set, calculates historical field overlap to trigger alarms, associates unsubmitted documents to generate reminders, and then packages and pushes them to the responsible person's terminal. This invention provides a specific embodiment: Step 105, when the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold and unsubmitted approved expense reports are detected within a preset period, a duplicate expense report alarm and an unsubmitted expense report reminder are generated and sent to the terminal associated with the employee's identity identifier, specifically including the following steps: Step 501: Extract the key fields of the structured text elements to generate the current reimbursement feature set.

[0099] In this step, key fields refer to the core data items in the expense report used for repetitive verification, including the amount and date information; the current expense feature set refers to the standardized data set composed of key fields, reflecting the core attributes of the expense report to be reviewed.

[0100] In this embodiment of the invention, the amount and date information fields are first located from the structured text elements, then the key fields used for reimbursement verification are selected, and finally the key fields are combined in a preset format to generate the current reimbursement feature set.

[0101] Step 502: Retrieve the historical feature set that matches the employee's identity in the historical accounting entries, calculate the field overlap between the current reimbursement feature set and the historical feature set, and generate a duplicate reimbursement alarm when the field overlap exceeds a preset similarity threshold.

[0102] In this step, the historical feature set refers to the set of data with the same structure extracted from historical accounting entries, which serves as the benchmark for duplication comparison; the field overlap refers to the numerical similarity index of the same field, which reflects the probability of data duplication by calculating the difference ratio.

[0103] In this embodiment of the invention, firstly, the employee's identity is used as the retrieval condition to query historical accounting entries. Secondly, the amount and date fields in the historical entries are extracted to form a historical feature set. Then, the numerical difference ratio of the same fields between the current reimbursement feature set and the historical feature set is calculated. When the difference ratio is lower than a preset threshold, a duplicate reimbursement alarm is generated.

[0104] Step 503: Scan the preset approved expense reimbursement document database, associate the expense reimbursement documents in the preset approved expense reimbursement document database with the employee identity identifiers corresponding to the expense reimbursement documents with the status of not submitted, and generate an unsubmitted expense reminder.

[0105] In this step, the preset approval through the reimbursement document database refers to the database that stores documents that have been approved but have not yet entered the reimbursement process; the association operation refers to the process of establishing a data binding between the unsubmitted document and the responsible person's identifier.

[0106] In this embodiment of the invention, the status flag field of the pre-approved expense reimbursement document database is first scanned, then expense reimbursement documents with a status value of "not submitted" are filtered, then the employee identity identifier embedded in the document is extracted, and finally the association between the unsubmitted document and the responsible person is established to generate an unsubmitted expense reminder.

[0107] Step 504: Encapsulate the duplicate expense report alarm and the unreported expense reminder to generate a terminal notification package, and send the terminal notification package to the terminal associated with the employee's identity using a preset communication interface.

[0108] In this step, encapsulation refers to the technology of reassembling alarms and reminders into transmission units according to communication protocols; terminal notification packets refer to standardized transmission data units containing alarm content, reminder information, and target addresses; and pre-built communication interfaces refer to the data transmission channels integrated into the system, supporting mobile network message push functions.

[0109] In this embodiment of the invention, the duplicate expense alarms and non-expense reminders are first reorganized according to the notification template, then the target terminal address information is added, and finally the data transmission channel of the preset communication interface is sent to the employee's mobile terminal.

[0110] This invention constructs a precise reimbursement feature set by extracting key fields, proactively identifying duplicate reimbursement behavior based on historical data comparison; it scans the database of unsubmitted documents to associate responsible parties and generates targeted reminders; finally, it encapsulates multiple types of early warning information and pushes them to the terminal in real time. This mechanism achieves proactive interception of reimbursement risks and immediate patching of process loopholes, effectively preventing financial losses and improving process integrity.

[0111] To enhance the real-time interception capability of large-sum fund risks, this step analyzes fund flow records to construct a transaction dataset, compares each transaction's amount field to mark abnormal records, and generates traceable early warning signals by associating them with the source accounting entries. This invention provides a specific embodiment: Step 106 utilizes a pre-set early warning module to monitor the fund flow data in the accounting entries. When a single transaction in the fund flow data exceeds a preset amount, a large expenditure early warning signal is generated. This specifically includes the following steps: Step 601: Parse the fund flow data in the accounting entries to obtain a flow dataset containing multiple transaction records.

[0112] In this step, the parsing operation refers to the process of decomposing the data structure of the accounting entries and identifying the boundaries of transaction records, which is achieved through delimiter identification; the transaction dataset refers to the structured set of transaction records formed after parsing, which contains multiple independent transaction data sorted by time.

[0113] In this embodiment of the invention, the fund flow data structure in the accounting entries is first parsed, then the record separator of the flow data is identified, then each independent transaction record is extracted and its integrity is verified, and finally a flow dataset containing multiple complete transaction records is generated.

[0114] Step 602: Extract the amount field of each transaction record in the transaction data set, compare the value of each amount field with a preset amount threshold, and when the value of the amount field exceeds the preset amount threshold, mark the transaction record corresponding to the amount field as an abnormal transaction record.

[0115] In this step, the amount field refers to the data unit that stores the transaction value in a single transaction record, measured in currency units; abnormal transaction records refer to special transaction data that are marked after the amount field exceeds a preset threshold, carrying an abnormal identifier.

[0116] In this embodiment of the invention, each transaction record in the transaction data set is first traversed, then the position of the amount field in the record is located, then the value of the amount field is read and compared with a preset amount threshold. When the value exceeds the threshold, an abnormal mark is added to the transaction record to generate an abnormal transaction record.

[0117] Step 603: Associate the abnormal transaction record with the accounting entry to which the abnormal transaction record belongs to generate a large expenditure warning signal.

[0118] In this step, the association operation refers to the technique of establishing a logical link between abnormal transaction records and their corresponding accounting entries, which is achieved through bidirectional pointers.

[0119] In this embodiment of the invention, the source accounting entry identifier of the abnormal transaction record is first obtained, then a bidirectional index relationship between the abnormal record and the source entry is established, and finally a risk level label is attached to generate a large expenditure warning signal.

[0120] This invention generates a complete transaction history dataset by accurately parsing accounting entries, compares the amount field of each transaction to mark abnormal transactions in real time, and innovates a correlation mechanism to trace abnormal records back to the original accounting entries, ultimately generating traceable early warning signals. This process constructs a closed-loop monitoring chain from data parsing and anomaly detection to risk tracing, significantly improving the timeliness of intercepting large-scale financial risks.

[0121] Figure 2 This invention provides a schematic diagram of the structure of an intelligent reimbursement process monitoring system based on multimodal data fusion, as shown in the embodiment of the invention. Figure 2 As shown, the system includes: The matching module 21 is used to acquire the employee's biometric data, match the biometric data with a preset identity template, and obtain the identity verification result; Extraction module 22 is used to scan the physical image of the expense report using a preset standardized processing program when the authentication result is passed, so as to extract structured text elements from the physical image. Transmission module 23 is used to transmit the employee identity identifier and the structured text elements in the authentication result to a preset cloud database to generate a data association package; The combination module 24 is used to combine the data association package with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries. The generation module 25 is used to generate duplicate reimbursement alarms and non-reimbursement reminders and send them to the terminal associated with the employee's identity when the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold and an unsubmitted approved reimbursement document is detected within a preset period. The monitoring module 26 is used to monitor the cash flow data in the accounting entries using a pre-set early warning module. When the amount of a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated.

[0122] Figure 2 The aforementioned intelligent reimbursement process monitoring system based on multimodal data fusion can execute... Figure 1 The implementation principle and technical effects of the intelligent reimbursement process monitoring method based on multimodal data fusion described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit performs operations in the intelligent reimbursement process monitoring system based on multimodal data fusion described in the above embodiments have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0123] In one possible design, Figure 2 The intelligent reimbursement process monitoring system based on multimodal data fusion, as shown in the embodiment, can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32; The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0124] The processing component 32 is used for: acquiring the employee's biometric data; matching the biometric data with a preset identity template to obtain an identity verification result; when the identity verification result is successful, scanning the physical image of the expense report using a preset standardized processing program to extract structured text elements from the physical image; transmitting the employee's identity identifier and the structured text elements from the identity verification result to a preset cloud database to generate a data association package; combining the data association package with the structured text elements in the preset cloud database to generate standardized accounting entries and storing them as historical accounting entries; when the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold and an unsubmitted approved expense report is detected within a preset period, generating duplicate expense alerts and non-approved expense reminders and sending them to the terminal associated with the employee's identity identifier; and monitoring the cash flow data in the accounting entries using a preset early warning module, generating a large expenditure early warning signal when a single amount in the cash flow data exceeds a preset amount.

[0125] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0126] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0127] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0128] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0129] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0130] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0131] This invention also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The embodiment shown is an intelligent reimbursement process monitoring method based on multimodal data fusion.

[0132] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0133] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0134] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0135] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent end-to-end monitoring of expense reimbursement based on multimodal data fusion, characterized in that, include: Obtain the employee's biometric data, match the biometric data with a preset identity template, and obtain the identity verification result; When the authentication result is passed, the physical image of the expense report is scanned using a pre-set standardized processing procedure to extract structured text elements from the physical image. The employee identification and the structured text elements in the authentication result are transmitted to a preset cloud database to generate a data association package; The data association package is combined with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries; When the similarity between the structured text element and the historical accounting entry exceeds a preset similarity threshold and an unsubmitted approved expense reimbursement document is detected within a preset period, a duplicate expense reimbursement alarm and an unreimbursed expense reminder are generated and sent to the terminal associated with the employee's identity. The system uses a pre-set early warning module to monitor the cash flow data in the accounting entries. When a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated.

2. The method according to claim 1, characterized in that, Obtain employee biometric data, match the biometric data with a preset identity template to obtain an identity verification result, including: The biometric data of employees is collected using the contact sensing surface of a pre-installed biometric data collection device. The biometric data is divided into key regions to obtain effective feature regions; The set of topological feature points in the effective feature region is spatially mapped to the set of registration feature points in the preset identity template, and the spatial distribution overlap is calculated based on the mapping result. When the spatial distribution overlap reaches a preset overlap threshold, an identity verification result is generated.

3. The method according to claim 1, characterized in that, When the authentication result is successful, a pre-defined standardized processing procedure is used to scan the physical image of the expense report to extract structured text elements from the physical image, including: When the authentication result is successful, a pre-set standardized processing procedure is used to respond to the authentication result, and the physical image of the expense report is processed for illumination compensation to generate a uniformly illuminated image. Locate the boundaries of the text regions in the uniformly illuminated image to obtain a set of text regions; The optical characters in each independent region of the text region set are converted using preset character conversion rules to generate the original character sequence; The original character sequence is parsed based on a preset set of accounting element rules to obtain the amount and date information; The monetary value and the date information are combined to generate structured text elements.

4. The method according to claim 1, characterized in that, The employee identification identifier and the structured text elements from the authentication result are transmitted to a preset cloud database to generate a data association package, including: Extract the employee identification identifier from the authentication result; The structured text elements are encapsulated and marked to generate standardized data packets; The employee identification identifier is bidirectionally bound to the standardized data packet to generate an initial associated data unit; The initial associated data unit is transmitted to a preset cloud database using a preset wireless transmission protocol, triggering a preset timestamp generator in the preset cloud database to add a timestamp identifier to the initial associated data unit, thereby generating an initial associated data unit with the added timestamp identifier. The initial associated data units with added timestamp identifiers are packaged to generate a data association package.

5. The method according to claim 1, characterized in that, The data association package is combined with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries, including: The employee identification in the data association package, the monetary value in the structured text element, and the date information in the structured text element are mapped to generate a basic accounting unit. The basic accounting units are formatted using a pre-defined accounting standardization template to generate standardized accounting entries. Activate the historical storage engine in the preset cloud database, and input the standardized accounting entries into the storage area corresponding to the historical storage engine to generate historical accounting entries.

6. The method according to claim 1, characterized in that, When the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold, and an unsubmitted approved expense reimbursement document is detected within a preset period, a duplicate reimbursement alarm and an unsubmitted expense reminder are generated and sent to the terminal associated with the employee's identity, including: Extract key fields from the structured text elements to generate the current reimbursement feature set; Retrieve the historical feature set that matches the employee's identity in the historical accounting entries, calculate the field overlap between the current reimbursement feature set and the historical feature set, and generate a duplicate reimbursement alarm when the field overlap exceeds a preset similarity threshold; The preset approvals are scanned through the expense reimbursement document database. The expense reimbursement documents in the database with the status of "not submitted" are associated with the employee identity identifiers corresponding to the expense reimbursement documents with the status of "not submitted" to generate an unsubmitted expense reminder. The duplicate expense report alarm and the unreported expense reminder are encapsulated to generate a terminal notification package, and the terminal notification package is sent to the terminal associated with the employee's identity using a preset communication interface.

7. The method according to claim 1, characterized in that, The system uses a pre-set early warning module to monitor the cash flow data in the accounting entries. When a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated, including: The fund flow data in the accounting entries is parsed to obtain a flow dataset containing multiple transaction records; Extract the amount field of each transaction record in the transaction data set, compare the value of each amount field with a preset amount threshold, and when the value of the amount field exceeds the preset amount threshold, mark the transaction record corresponding to the amount field as an abnormal transaction record. The abnormal transaction records are associated with the accounting entries to which they belong, generating a large expenditure warning signal.

8. A smart reimbursement process monitoring system based on multimodal data fusion, characterized in that, include: The matching module is used to acquire the employee's biometric data, match the biometric data with a preset identity template, and obtain the identity verification result; The extraction module is used to scan the physical image of the expense report using a preset standardized processing program when the authentication result is passed, so as to extract structured text elements from the physical image. The transmission module is used to transmit the employee identity identifier and the structured text elements in the authentication result to a preset cloud database to generate a data association package; The combination module is used to combine the data association package with the structured text elements in the preset cloud database to generate standardized accounting entries and store them as historical accounting entries. The generation module is used to generate duplicate reimbursement alarms and non-reimbursement reminders and send them to the terminal associated with the employee's identity when the similarity between the structured text elements and the historical accounting entries exceeds a preset similarity threshold and an unsubmitted approved reimbursement document is detected within a preset period. The monitoring module is used to monitor the cash flow data in the accounting entries using a pre-set early warning module. When the amount of a single transaction in the cash flow data exceeds a preset amount, a large expenditure early warning signal is generated.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the intelligent reimbursement process monitoring method based on multimodal data fusion as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The device contains a computer program that, when executed by a computer, implements a method for intelligent full-process monitoring of reimbursement based on multimodal data fusion as described in any one of claims 1 to 7.

Images