Deep learning-based invoice information management method, system and readable medium
A deep learning and information management technology, applied in the field of invoice information collection and management, to achieve the effect of saving human resources and strong robustness
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Publication Date
- 2019-11-19
Smart Images

Figure 1 
Figure 2 
Figure 3
Abstract
Description
technical field
[0001] The present invention relates to the technical field of invoice information collection and management, and more specifically, relates to a deep learning-based invoice information management method, system and readable medium. Background technique
[0002] The traditional financial reimbursement system requires financial personnel to collect invoices, manually identify the types and amounts of expenses that need to be reimbursed, and enter information on the computer. Traditional manual information entry is too time-consuming and laborious, and financial personnel need to spend a lot of time on this type of repetitive work.
[0003] The invention patent of publication number CN109344838A "Invoice Information Automatic and Rapid Recognition Method, System and Device" discloses a technical solution for image and text recognition using a neural network, but it does not further extract valid information of invoices separately to form an information manageme...
Examples
Embodiment 1
[0050] Such as figure 1 As shown, the first aspect of the present invention discloses a method for managing invoice information based on deep learning, comprising the following steps:
[0051] S1. Scan the paper version of the invoice through the invoice scanning module, and collect the invoice image data converted into image format, such as figure 2 shown in;
[0052] S2. Correct the preprocessing module to identify the invoice image and perform edge detection. If the inclination angle of the edge of the invoice image is greater than the preset threshold, correct the angle offset of the invoice image; otherwise, no operation is performed;
[0053] S3. The 1,000 invoice images collected by the detection module are annotated with the tool Imglabel. According to the characteristics of the information to be extracted, the information points to be extracted are annotated with corresponding category labels, such as image 3 shown; put the marked data into the Faster-Rcnn neural ...
Embodiment 2
[0074] The second aspect of the present invention discloses an invoice information management system based on deep learning, based on the above-mentioned invoice information management method based on deep learning, including
[0075] The invoice scanning module converts the original paper version of the invoice through the scanning of the peripheral lens device into a picture format file, and uploads the picture file to the system for further processing;
[0076] Correct the preprocessing module, identify the invoice image and perform edge detection. If the inclination angle of the edge of the invoice image is greater than the preset threshold, correct the angle offset of the invoice image; otherwise, no operation is performed;