Method for integrating financial cloud platform with OCR (optical character recognition) software

Abstract

The invention discloses a method for integrating OCR recognition software on a financial cloud platform, and relates to the technical field of OCR recognition, and the method comprises the following steps: substituting a calculated image preprocessing quality evaluation value and a character recognition prediction accuracy rate into a character recognition efficiency algorithm unit, calculating different recognition efficiencies of different image financial statements, and carrying out the calculation of the recognition efficiencies; according to the calculated different recognition efficiencies of the different image financial statements, the task with the highest execution efficiency is selected to be executed preferentially by adopting the highest efficiency priority, and the OCR character recognition work of the multiple image financial statements is carried out. According to the method, the financial statement bill with the highest character recognition efficiency is preferentially subjected to OCR character recognition, repeated recognition of financial statement bills with poor image definition quality is avoided, the recognition accuracy and recognition efficiency of financial cloud platform integrated OCR recognition software are improved, and the method is worthy of popularization and use.

Description

Technical Field

[0001] The present invention relates to the field of OCR recognition technology, and in particular to a method for integrating OCR recognition software into a financial cloud platform. Background Art

[0002] Traditional business bill recognition is mainly done manually, with financial personnel manually entering bill contents into Excel or other files or databases. The disadvantages of this method are heavy workload, easy errors, and high repetitiveness. When the business volume of the enterprise increases, the types of bills and the amount of data increase, the manpower cost is huge. Especially in the financial field, the accuracy of bill content, especially the information related to the amount, is required to be very high. Therefore, in this scenario, the idea of ​​integrating computer image recognition software with traditional financial cloud platforms is proposed, and the manual identification of bill content is handed over to computer image recognition software.

[0003] OCR text recognition software is a software that uses OCR (Optical Character Recognition) technology to directly convert the text content on images and photos into editable text. At present, OCR recognition software in the existing technology only has a simple image-to-text conversion function. When the number of financial statement bills is large, it cannot avoid the financial statement bills with poor image clarity while ensuring the recognition order of the original financial statement bills, and give priority to the financial statement bills with good clarity. Instead, it repeatedly recognizes the financial statement bills with poor clarity. On the one hand, it will reduce the overall recognition efficiency of the OCR recognition software integrated in the financial cloud platform. On the other hand, it is easy to produce garbled characters when the financial statement bill recognition is entered into Excel and other files or databases, which is not conducive to use.

[0004] In view of the above problems, the present invention provides a method for integrating OCR recognition software into a financial cloud platform. Summary of the invention

[0005] The purpose of the present invention is to provide a method for integrating OCR recognition software into a financial cloud platform to solve the problems raised in the above background technology.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for integrating OCR recognition software into a financial cloud platform, the method comprising the following steps:

[0007] Image financial statement acquisition: obtain the financial statements to be processed in the form of images from the financial cloud platform and store them in the database of the OCR recognition software;

[0008] Preprocessing: Preprocess the financial statements in the form of multiple images in the database transmitted to the OCR recognition software, and obtain the three parameter values ​​of image contrast value, binarization processing value and image stability value through the built-in image processing software and algorithm of the OCR recognition software;

[0009] Image financial statement character recognition:

[0010] Specifically, the collected parameter values ​​are transmitted and substituted into the image preprocessing quality evaluation value algorithm unit and the character recognition prediction accuracy algorithm unit to calculate the image preprocessing quality evaluation value and the character recognition prediction accuracy, two key parameters that affect the character recognition efficiency when performing character recognition on the image financial statements;

[0011] Next, the calculated image preprocessing quality evaluation value and character recognition prediction accuracy are substituted into the character recognition efficiency algorithm unit to calculate different recognition efficiencies of different image financial statements;

[0012] Afterwards, according to the different recognition efficiencies of different image financial statements, the highest efficiency is given priority, that is, the task with the highest execution efficiency is selected for priority execution, and the OCR character recognition work of multiple image financial statements is performed;

[0013] Monitoring and adjustment: Real-time monitoring of the conversion of multiple image financial statements from images to characters in the financial cloud platform, and real-time adjustment of OCR character recognition strategies based on character recognition efficiency;

[0014] Feedback and iteration: Based on the calculated recognition efficiency, the task with the highest execution efficiency is selected for priority execution. As OCR character recognition proceeds, after the character recognition efficiency of an image financial statement is calculated, the calculated character recognition efficiency value is substituted into the image preprocessing quality assessment value algorithm unit of the next image financial statement for new iterative calculation. Through iterative calculation, the image preprocessing quality assessment value of the next image financial statement is affected in real time.

[0015] Optionally, the OCR recognition software operates using an image financial statement data acquisition module, a data processing decoding module, a calculation processing module and a character recognition module.

[0016] Optionally, the preprocessing includes data cleaning and data standardization.

[0017] Optionally, the calculation processing module includes an image preprocessing quality assessment value algorithm unit, a character recognition prediction accuracy algorithm unit, and a character recognition efficiency algorithm unit.

[0018] Optionally, the image preprocessing quality assessment value algorithm unit is as follows:

[0019]

[0020] in:

[0021] Q1 represents the image preprocessing quality assessment value;

[0022] C represents the image contrast value obtained after image preprocessing;

[0023] B represents the binarization value obtained after image preprocessing;

[0024] S represents the image stability value after tilt correction obtained after image preprocessing;

[0025] E3prev represents the character recognition efficiency of the previous image;

[0026] α is the weight coefficient of the image contrast value;

[0027] β is the weight coefficient of the binary processing value;

[0028] γ is the weight coefficient of the image stabilization value;

[0029] θ is the weight coefficient of character recognition efficiency in the previous iteration.

[0030] Optionally, the character recognition prediction accuracy algorithm unit is as follows:

[0031] A2=Q1×P s ×k

[0032] in:

[0033] A2 represents the prediction accuracy of character recognition, which is a percentage value;

[0034] Q1 represents the image preprocessing quality assessment value;

[0035] P S Represents the accuracy of character segmentation;

[0036] k represents the adjustment factor.

[0037] Optionally, the character recognition efficiency algorithm unit is as follows:

[0038]

[0039] Where: E3 represents the character recognition efficiency;

[0040] A2 represents the prediction accuracy of character recognition;

[0041] T stands for processing time;

[0042] C p represents the preprocessing cost;

[0043] Q1 represents the image preprocessing quality assessment value;

[0044] μ is the adjustment factor.

[0045] Optionally, the character recognition module further includes a recognition optimization control unit, which is used to monitor the recognition efficiency of the OCR recognition software on the image financial statements in real time and perform feedback adjustment.

[0046] Compared with the prior art, the present invention has the following beneficial effects:

[0047] The present invention cooperates with each other through the three algorithm units in the calculation and processing module. It can select the financial statement bills with the highest character recognition efficiency to perform OCR character recognition work preferentially while ensuring the recognition order of the original financial statement bills, so as to avoid the financial statement bills with poor image clarity quality, and avoid the OCR recognition software from repeatedly recognizing the financial statement bills with poor clarity. On the one hand, it will reduce the overall recognition efficiency of the OCR recognition software integrated with the financial cloud platform, and on the other hand, it will easily produce garbled characters when the financial statement bill recognition is entered into files such as Excel or databases, thereby improving the recognition accuracy and recognition efficiency of the OCR recognition software integrated with the financial cloud platform, and is worthy of promotion and use. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A method flow chart of a computer network management method;

[0049] Figure 2 The figure is a schematic diagram of the overall structure of the computer network management method. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] Example 1: Please refer to Figure 1 and Figure 2 The present invention provides a technical solution: a method for integrating OCR recognition software into a financial cloud platform, the method comprising the following steps:

[0052] Image financial statement acquisition: obtain multiple financial statements in the form of images that need to be processed from the financial cloud platform and store them in the database of the OCR recognition software;

[0053] Preprocessing: Preprocess multiple financial statements in the form of images in the database transmitted to the OCR recognition software, and obtain the image contrast value, binarization value and image stability value, three key factors that affect the character recognition efficiency of the OCR recognition software when recognizing image financial statements, through the built-in image processing software and algorithm of the OCR recognition software;

[0054] Image financial statement character recognition:

[0055] Specifically, the collected image contrast value, binarization processing value and image stability value are transmitted together and substituted into the image preprocessing quality evaluation value algorithm unit and the character recognition prediction accuracy algorithm unit to calculate the image preprocessing quality evaluation value and the character recognition prediction accuracy, two key factors that affect the character recognition efficiency when performing character recognition on the image financial statement;

[0056] Next, the calculated image preprocessing quality evaluation value and character recognition prediction accuracy are substituted into the character recognition efficiency algorithm unit to calculate different recognition efficiencies of different image financial statements;

[0057] Afterwards, according to the different recognition efficiencies of different image financial statements, the highest efficiency is given priority, that is, the task with the highest execution efficiency is selected for priority execution, and the OCR character recognition work of multiple image financial statements is performed;

[0058] Monitoring and adjustment: Real-time monitoring of the conversion of multiple image financial statements from images to characters in the financial cloud platform, and real-time adjustment of OCR character recognition strategies based on character recognition efficiency;

[0059] Feedback and iteration: Based on the calculated recognition efficiency, the task with the highest execution efficiency is selected for priority execution. As OCR character recognition proceeds, after the character recognition efficiency of an image financial statement is calculated, the calculated character recognition efficiency value is substituted into the image preprocessing quality assessment value algorithm unit of the next image financial statement for new iterative calculation. Through iterative calculation, the image preprocessing quality assessment value of the next image financial statement is affected in real time, so as to help the financial cloud platform integrated OCR recognition software gradually approach the optimal solution during the iteration process, thereby improving the accuracy and overall efficiency of OCR recognition.

[0060] The financial cloud platform integrated OCR recognition method adopts image financial statement data acquisition module, data processing and decoding module, calculation processing module and character recognition module.

[0061] Preprocessing includes data cleaning and data standardization.

[0062] The character recognition module also includes a recognition optimization control unit, which is used to monitor the recognition efficiency of the OCR recognition software on the image financial statements in real time and make feedback adjustments.

[0063] In this embodiment:

[0064] The present invention forms a complete financial cloud platform integrated OCR recognition method through three algorithms, and the three algorithms all have substantial effects:

[0065] Image preprocessing quality assessment algorithm 1:

[0066] It can quantitatively evaluate the image preprocessing quality based on the image contrast value, binarization processing value and image stability value collected by the image financial statement data acquisition module, which are three key factors affecting the character recognition efficiency of the OCR recognition software when recognizing image financial statements. This helps to ensure that the image quality reaches the best state when processing bills and receipts, so as to improve the accuracy of subsequent character recognition;

[0067] Character recognition prediction accuracy algorithm 2:

[0068] It can predict the accuracy of character recognition based on the image preprocessing quality assessment value and the accuracy of character segmentation, which helps to timely discover and correct potential problems during the recognition process, thereby further improving the accuracy of the overall recognition of image financial statements;

[0069] Character recognition efficiency algorithm three: It can comprehensively consider the image preprocessing quality, character recognition accuracy, processing time and cost, and evaluate the overall efficiency and cost balance of OCR recognition. It helps to optimize processing time and cost while maintaining high recognition accuracy and improve recognition efficiency.

[0070] Moreover, the combination of the three algorithms can ensure the recognition order of the original financial statement bills while selecting the task with the highest execution efficiency to execute first, so as to avoid the financial statement bills with poor image clarity and quality and give priority to the recognition of the financial statement bills with better clarity.

[0071] It is worth noting that as OCR character recognition proceeds, after the character recognition efficiency of an image financial statement is calculated, the character recognition efficiency value calculated by the character recognition efficiency algorithm three will be substituted into the image preprocessing quality assessment value algorithm unit of the next image financial statement for new iterative calculation. Through iterative calculation, the image preprocessing quality assessment value of the next image financial statement is affected in real time, so as to help the financial cloud platform integrated OCR recognition software gradually approach the optimal solution during the iteration process, thereby improving the accuracy and overall efficiency of OCR recognition.

[0072] See also Figure 1 and Figure 2 , the image preprocessing quality assessment value algorithm unit is as follows:

[0073]

[0074] in:

[0075] Q1 represents the image preprocessing quality assessment value;

[0076] C represents the image contrast value obtained after image preprocessing;

[0077] B represents the binarization value obtained after image preprocessing;

[0078] S represents the image stability value after tilt correction obtained after image preprocessing;

[0079] E 3prev Represents the character recognition efficiency of the previous image;

[0080] α is the weight coefficient of the image contrast value;

[0081] β is the weight coefficient of the binary processing value;

[0082] γ is the weight coefficient of the image stabilization value;

[0083] θ is the weight coefficient of character recognition efficiency in the previous iteration;

[0084] By adjusting the four different weight coefficients of α, β, γ and θ, the influence of different variables on the calculated value Q1 can be balanced. In layman's terms, the weight coefficient can reflect the priority or importance of different variables in the decision-making or calculation process. For example, as the OCR character recognition work proceeds, when it is found that the image contrast value C has a high impact on the clarity of a batch of image financial statement bills, the platform will increase the value of α. The four weight coefficients of α, β, γ and θ can be self-adjusted as the financial cloud platform integrates the OCR recognition software method.

[0085] This part represents the weighted sum calculated by multiplying the image contrast value, the binarization processing value, and the image stability value after tilt correction by their weights respectively. It is a real number between 0 and 1. The closer the value is to 1, the more stable the image is, and the better the image preprocessing quality is, the more likely it will be recognized by the OCR recognition software.

[0086] In the new round of calculation of Formula 1, The calculated values ​​are used as feedback and correction items. They are combined with the weighted sum of the input variables iterated in the current new round of formula 1 calculation, namely the image contrast value, the binarization processing value, and the image stability value after tilt correction, to jointly affect the calculation results of the image preprocessing quality evaluation value.

[0087] The calculation result of this part is a real number between 0.8 and 1. Compared with the weighted sum of the image contrast value, the binarization processing value and the image stability value after tilt correction, it has less impact on the image preprocessing quality evaluation value.

[0088] In this embodiment:

[0089] Through the above algorithm formula, the image preprocessing quality can be quantitatively evaluated based on the three key factors affecting the character recognition efficiency of the OCR recognition software when recognizing image financial statements, namely the image contrast value, binarization processing value and image stability value collected by the image financial statement data acquisition module. This helps to ensure that the image quality reaches the optimal state when processing bills and receipts, so as to improve the accuracy of subsequent character recognition.

[0090] It is worth noting that, as the iteration enters a new round of calculation in Formula 1, The value is always between 0.8 and 1. Therefore, assuming that the weighted sum of the image contrast value, the binarization processing value, and the image stability value after tilt correction remains unchanged, the image preprocessing quality assessment value calculated by the new round of formula one will always be slightly smaller than the previous round, that is, algorithm one can ensure the recognition order of the original financial statement bills when the clarity of the image financial statement bills does not fluctuate much.

[0091] When an image financial statement bill has poor clarity, the weighted sum of the image contrast value, the binarization value, and the image stability value after tilt correction will be The calculated value of this part is low, and the character recognition efficiency value calculated by substituting it into formula three will also be low, and it will not be recognized as a priority. Therefore, the financial statement bill with poor clarity can be temporarily skipped for recognition, and the financial statement bill with high character recognition efficiency value and better clarity can be recognized as a priority.

[0092] See also Figure 1 and Figure 2 , the character recognition prediction accuracy algorithm unit is as follows:

[0093] A2=Q1×P s ×k

[0094] in:

[0095] A2 represents the prediction accuracy of character recognition, which is a percentage value used to calculate the probability that the OCR system can correctly recognize characters after image preprocessing and character segmentation. The higher the value, the more accurate the character recognition of the OCR system;

[0096] Q1 represents the image preprocessing quality assessment value, which is calculated by the image preprocessing quality assessment value algorithm unit;

[0097] P S Represents the accuracy of character segmentation, as a percentage. The closer the value is to 1, the more accurate the character segmentation is.

[0098] k represents an adjustment coefficient, which is used to adjust the prediction accuracy coefficient. Like the above α, β, γ, and θ, k can be self-adjusted as the OCR character recognition work proceeds.

[0099] In this embodiment: by substituting the image preprocessing quality assessment value Q1 calculated in formula one into algorithm formula two, the calculated predicted accuracy of character recognition can directly reflect the expected accuracy of the OCR recognition system for bill and note character recognition. This predicted accuracy can help the financial cloud platform better evaluate the effect of OCR recognition in practical applications, so as to take corresponding measures to improve the recognition accuracy. For example, when the A2 value is low, the platform can optimize the image preprocessing process and adjust the OCR recognition parameters to improve the recognition accuracy.

[0100] See also Figure 1 and Figure 2 , the character recognition efficiency algorithm unit is as follows:

[0101]

[0102] in:

[0103] E3 stands for character recognition efficiency, which is a comprehensive evaluation value used to measure the balance between recognition accuracy, processing time and cost in OCR recognition tasks. The higher the value, the lower the processing time and cost while maintaining a high recognition accuracy, that is, the higher the overall recognition efficiency;

[0104] A2 represents the prediction accuracy of character recognition, which is calculated by the character recognition prediction accuracy algorithm unit;

[0105] T stands for processing time, which is the time required for OCR recognition software to recognize a financial statement image. It indicates the negative impact of processing time on character recognition efficiency. The longer the processing time, the lower the overall recognition efficiency.

[0106] C p Represents the preprocessing cost, which refers to the cost required for the image preprocessing step in the OCR recognition task, including computing resource consumption and storage space occupation. As part of the denominator, it, together with the processing time, constitutes the denominator of the character recognition efficiency. The higher the preprocessing cost, the lower the overall recognition efficiency.

[0107] Q1 represents the image preprocessing quality assessment value, which is calculated by the image preprocessing quality assessment value algorithm unit;

[0108] μ is an adjustment factor, which is a coefficient used to adjust the efficiency of character recognition. It reflects the importance of balancing recognition accuracy, processing time and cost in a specific application scenario. It can also be self-adjusted as the OCR character recognition work progresses.

[0109] In this embodiment:

[0110] By substituting the image preprocessing quality evaluation value Q1 calculated in Formula 1 and the predicted accuracy of character recognition A2 calculated in Formula 2 into Formula 3, the calculated character recognition efficiency value E3 comprehensively reflects the overall efficiency and cost balance of OCR recognition, including image preprocessing quality, character recognition accuracy, processing time and preprocessing cost. According to the calculated E3 value, OCR character recognition can be given priority to financial statement bills with higher E3 values, and unclear financial statement bills with lower E3 values ​​can be skipped, thereby avoiding repeated recognition of financial statement bills with poor clarity by the OCR recognition software. On the one hand, it will reduce the overall recognition efficiency of the OCR recognition software integrated with the financial cloud platform. On the other hand, it is easy to produce garbled characters when the financial statement bill recognition is entered into files such as Excel or databases, thereby improving the recognition accuracy and efficiency of the OCR recognition software integrated with the financial cloud platform.

[0111] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for integrating OCR recognition software into a financial cloud platform, characterized in that: The method comprises the following steps: Image financial statement acquisition: The image financial statement data acquisition module is used to obtain the financial statements to be processed in the form of images from the financial cloud platform and store them in the database of the OCR recognition software; Image financial statement character recognition: Specifically include: Process multiple financial statements in the form of images in the database transmitted to the OCR recognition software, and obtain three parameters: image contrast value, binarization value, and image stability value through the built-in image processing software and algorithm of the OCR recognition software; The parameters are transmitted to the data processing and decoding module for decoding preprocessing to obtain the parameter values ​​calculated in the calculation processing module; The collected parameter values ​​are transmitted and substituted into the image preprocessing quality evaluation value algorithm unit and the character recognition prediction accuracy algorithm unit to calculate the image preprocessing quality evaluation value and the character recognition prediction accuracy, two key parameters that affect the character recognition efficiency when performing character recognition on the image financial statements; Next, the calculated image preprocessing quality evaluation value and character recognition prediction accuracy are substituted into the character recognition efficiency algorithm unit to calculate different recognition efficiencies of different image financial statements; Afterwards, according to the different recognition efficiencies of different image financial statements, the highest efficiency is given priority, that is, the task with the highest execution efficiency is selected for priority execution, and the OCR character recognition work of multiple image financial statements is performed; Monitoring and adjustment: Real-time monitoring of the conversion of multiple image financial statements from images to characters in the financial cloud platform, and real-time adjustment of OCR character recognition strategies based on character recognition efficiency; Feedback and iteration: Based on the calculated recognition efficiency, the task with the highest execution efficiency is selected for priority execution. As OCR character recognition proceeds, after the character recognition efficiency of an image financial statement is calculated, the calculated character recognition efficiency value is substituted into the image preprocessing quality assessment value algorithm unit of the next image financial statement for new iterative calculation. Through iterative calculation, the image preprocessing quality assessment value of the next image financial statement is affected in real time.

2. The method for integrating OCR recognition software into a financial cloud platform according to claim 1, characterized in that: The OCR recognition software uses an image financial statement data acquisition module, a data processing and decoding module, a calculation processing module, and a character recognition module to run.

3. The method for integrating OCR recognition software into a financial cloud platform according to claim 2, characterized in that: The data processing and decoding module includes data cleaning and data standardization, and is used to decode and convert three parameters, namely, contrast value, binarization processing value and image stabilization value, into specific parameter values ​​involved in formula calculation.

4. The method for integrating OCR recognition software into a financial cloud platform according to claim 2, characterized in that: The calculation processing module includes an image preprocessing quality evaluation value algorithm unit, a character recognition prediction accuracy algorithm unit and a character recognition efficiency algorithm unit.

5. The method for integrating OCR recognition software into a financial cloud platform according to claim 4, characterized in that: The image preprocessing quality assessment value algorithm unit is as follows: in: Q1 represents the image preprocessing quality assessment value; C represents the image contrast value obtained after image preprocessing; B represents the binarization value obtained after image preprocessing; S represents the image stability value after tilt correction obtained after image preprocessing; E 3prev Represents the character recognition efficiency of the previous image; α is the weight coefficient of the image contrast value; β is the weight coefficient of the binary processing value; γ is the weight coefficient of the image stabilization value; θ is the weight coefficient of character recognition efficiency in the previous iteration.

6. The method for integrating OCR recognition software into a financial cloud platform according to claim 4, characterized in that: The character recognition prediction accuracy algorithm unit is as follows: <h2 style=";text-align:left;direction:ltr">A2=Q1×P<h2 style=";text-align:left;direction:ltr"> s <h2 style=";text-align:left;direction:ltr"> ×k in: A2 represents the prediction accuracy of character recognition; Q1 represents the image preprocessing quality assessment value; P S Represents the accuracy of character segmentation; k represents the adjustment factor.

7. The method for integrating OCR recognition software into a financial cloud platform according to claim 4, characterized in that: The character recognition efficiency algorithm unit is as follows: Where: E3 represents the character recognition efficiency; A2 represents the prediction accuracy of character recognition; T stands for processing time; C p represents the preprocessing cost; Q1 represents the image preprocessing quality assessment value; μ is the adjustment factor.

8. The method for integrating OCR recognition software into a financial cloud platform according to claim 2, characterized in that: The character recognition module also includes a recognition optimization control unit, which is used to monitor the recognition efficiency of the OCR recognition software on the image financial statements in real time and perform feedback adjustments.

Images