Search

CN-121999270-A - Financial bill intelligent classification method and device based on artificial intelligence

CN121999270ACN 121999270 ACN121999270 ACN 121999270ACN-121999270-A

Abstract

The invention discloses an artificial intelligence-based financial bill intelligent classification method and device, wherein the method comprises the steps of receiving a financial bill image, inputting the financial bill image into a pre-trained convolutional neural network model, and outputting a financial bill category, wherein the convolutional neural network model comprises a residual feature encoder and a classifier, the residual feature encoder is used for extracting a layout topological structure and a marking pattern of the financial bill image, the classifier is used for determining the financial bill category by utilizing the layout topological structure and the marking pattern, and the marking pattern reflects the category and authenticity of the financial bill. The invention can improve the accuracy, the robustness and the efficiency of distinguishing the financial bill.

Inventors

  • WANG HAO

Assignees

  • 中国建设银行股份有限公司

Dates

Publication Date
20260508
Application Date
20251215

Claims (15)

  1. 1. An artificial intelligence based financial bill intelligent classification method is characterized by comprising the following steps: receiving a financial billing image; The method comprises the steps of inputting a financial bill image into a pre-trained convolutional neural network model and outputting a financial bill category, wherein the convolutional neural network model comprises a residual feature encoder and a classifier, the residual feature encoder is used for extracting a layout topological structure and a marked pattern of the financial bill image, the classifier is used for determining the financial bill category by utilizing the layout topological structure and the marked pattern, and the marked pattern reflects the category and the authenticity of the financial bill.
  2. 2. The method of claim 1, further comprising, prior to inputting the financial billing image into the pre-trained convolutional neural network model: Preprocessing the financial bill image to obtain a preprocessed financial bill image, wherein the preprocessing comprises image graying, color correction, image enhancement and size normalization; inputting the financial bill image into a pre-trained convolutional neural network model, comprising: And inputting the preprocessed financial bill image into a pre-trained convolutional neural network model.
  3. 3. The method of claim 2, wherein the image enhancement comprises minimum intra-class variance multi-histogram equalization and/or luminance preserving dual histogram equalization.
  4. 4. The method of claim 1, wherein the residual feature encoder is a ResNet model.
  5. 5. The method of claim 1, wherein the residual feature encoder comprises a bottleneck block and a skip connection, the bottleneck block comprising three convolutional layers, the skip connection for bypassing the convolutional layers through identity mapping, the input being directly added to the output to effect residual learning.
  6. 6. The method of claim 1, wherein the convolutional neural network model is trained as follows: Pre-training the convolutional neural network model by using a general data set in advance to obtain a pre-trained convolutional neural network model; Collecting historical financial billing images; category labeling and data enhancement are carried out on the historical financial bill images, so that a training set is formed; and training the pre-trained convolutional neural network model by using the training set to obtain a trained convolutional neural network model.
  7. 7. Financial billing intelligence sorter based on artificial intelligence, its characterized in that includes: the data receiving module is used for receiving the financial bill image; The classification recognition module is used for inputting the financial bill image into a pre-trained convolutional neural network model and outputting the financial bill category, the convolutional neural network model comprises a residual feature encoder and a classifier, the residual feature encoder is used for extracting a layout topological structure and a marking pattern of the financial bill image, the classifier is used for determining the financial bill category by utilizing the layout topological structure and the marking pattern, and the marking pattern reflects the category and authenticity of the financial bill.
  8. 8. The apparatus as recited in claim 7, further comprising: The system comprises a classification recognition module, a preprocessing module, a data processing module and a data processing module, wherein the classification recognition module is used for inputting a financial bill image into a pre-trained convolutional neural network model, and preprocessing the financial bill image to obtain a preprocessed financial bill image; the classification and identification module is specifically used for: And inputting the preprocessed financial bill image into a pre-trained convolutional neural network model.
  9. 9. The apparatus of claim 8, wherein the image enhancement comprises minimum intra-class variance multi-histogram equalization and/or luminance-preserving dual-histogram equalization.
  10. 10. The apparatus of claim 7, wherein the residual feature encoder is a ResNet model.
  11. 11. The apparatus of claim 7, wherein the residual feature encoder comprises a bottleneck block and a skip connection, the bottleneck block comprising three convolutional layers, the skip connection to bypass the convolutional layers through identity mapping to add input directly to output to effect residual learning.
  12. 12. The apparatus of claim 7, wherein the convolutional neural network model is trained as follows: Pre-training the convolutional neural network model by using a general data set in advance to obtain a pre-trained convolutional neural network model; Collecting historical financial billing images; category labeling and data enhancement are carried out on the historical financial bill images, so that a training set is formed; and training the pre-trained convolutional neural network model by using the training set to obtain a trained convolutional neural network model.
  13. 13. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method of any of claims 1 to 6 when executing the computer program.
  14. 14. A computer readable storage medium, characterized in that the computer readable storage medium stores a computer program which, when executed by a processor, implements the method of any of claims 1 to 6.
  15. 15. A computer program product, characterized in that the computer program product comprises a computer program which, when executed by a processor, implements the method of any of claims 1 to 6.

Description

Financial bill intelligent classification method and device based on artificial intelligence Technical Field The invention relates to the technical field of artificial intelligence, in particular to an intelligent classification method and device for financial billing based on artificial intelligence. Background Financial instruments are legal documents, such as invoices, checks, draft, rental tickets, and the like, used by businesses, units, or individuals in economic activities to record funds payouts, explicit entitlement obligations. The prior art generally uses simple Optical Character Recognition (OCR) technology and plain text analysis methods for type recognition, but the prior art methods suffer from the following technical drawbacks in the actual financial processing scenario: (1) The robustness is poor, the existing method is seriously dependent on pure text information of a document image after OCR analysis as a core classification characteristic, so that the performance of the method is highly sensitive to the quality (such as definition, illumination and shooting angle) of an input image and OCR capability. If the document image contains the factors such as coincident text, folding, offset or seal shielding, the OCR recognition accuracy is low, and the structural defect of poor classification effect is further caused. (2) The accuracy is low, and the invoice types (such as special invoice and common invoice) with fine visual difference are difficult to distinguish from a large number of similar bill categories with high accuracy in the prior art. (3) The rules are complex to maintain and have efficiency bottlenecks, namely plain text classification generally needs to customize a complex rule set according to the recognized text content to screen whether the invoice is of a certain type, and the rules are easily influenced by high keyword superposition rate and format diversity. And the complete OCR recognition process is long in time consumption, and the automation efficiency of financial processing is affected. (4) Model vulnerability the existing Visual Document Understanding (VDU) model, while combining text and layout information, has core decision logic that presents extremely high vulnerability to micro-antagonistic disturbances of OCR bounding boxes (BBox) and pixels, and classification accuracy can severely drop once layout information is disturbed. Disclosure of Invention The embodiment of the invention provides an artificial intelligence-based financial bill intelligent classification method for improving the accuracy, robustness and efficiency of financial bill distinguishing, which comprises the following steps: receiving a financial billing image; The method comprises the steps of inputting a financial bill image into a pre-trained convolutional neural network model and outputting a financial bill category, wherein the convolutional neural network model comprises a residual feature encoder and a classifier, the residual feature encoder is used for extracting a layout topological structure and a marked pattern of the financial bill image, the classifier is used for determining the financial bill category by utilizing the layout topological structure and the marked pattern, and the marked pattern reflects the category and the authenticity of the financial bill. The embodiment of the invention also provides an artificial intelligence-based intelligent financial bill classification device for improving the accuracy, the robustness and the efficiency of financial bill distinguishing, which comprises the following components: the data receiving module is used for receiving the financial bill image; The classification recognition module is used for inputting the financial bill image into a pre-trained convolutional neural network model and outputting the financial bill category, the convolutional neural network model comprises a residual feature encoder and a classifier, the residual feature encoder is used for extracting a layout topological structure and a marking pattern of the financial bill image, the classifier is used for determining the financial bill category by utilizing the layout topological structure and the marking pattern, and the marking pattern reflects the category and authenticity of the financial bill. The embodiment of the invention also provides computer equipment, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the intelligent classification method of the financial bill based on artificial intelligence when executing the computer program. The embodiment of the invention also provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the intelligent classification method of the financial bill based on artificial intelligence when being executed by a processor. The embodim