Associated transaction business collaboration method and system based on business financial fare data

Abstract

The invention discloses an associated transaction business collaboration method and system based on financial fare data, and relates to the technical field of associated transaction businesses, and the method comprises the following steps: converting a paper bill into an electronic bill image through a high-resolution scanner; after the electronic bill is generated, the image is compressed according to the complexity of bill information; the compressed electronic bill is transmitted to an OCR system for image recognition, and in the OCR recognition process, image quality features of the electronic bill are obtained. According to the method, the image quality features of the electronic bills are evaluated through machine learning, the excessive compression risk is intelligently recognized, the compression proportion is automatically adjusted to improve the OCR recognition accuracy, financial errors are reduced, the system can dynamically adjust the compression strategy, high-quality and low-quality bills can be quickly recognized, manual operation is reduced, the processing efficiency is improved, and if problems are not compressed, the system can be used for solving the problem. And the system automatically gives an alarm to prompt troubleshooting, so that efficient and accurate processing of the electronic bill is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of related transaction business, and in particular to a method and system for related transaction business collaboration based on business, finance, bill and payment data. Background Art

[0002] The business collaboration system for related transactions based on business, finance and bill data refers to the integration of the business data, financial data and bill data of the enterprise to establish a unified management system to achieve the automation and collaborative management of the entire process of related transactions. The system can achieve seamless connection between business and financial operations, automatically collect and process relevant data, and ensure that each related transaction can be carried out accurately and efficiently from business initiation to financial settlement to bill management, thereby improving the operational efficiency of the enterprise, reducing risks, and achieving more comprehensive financial supervision and decision support.

[0003] In the processing of the bill management system, OCR (Optical Character Recognition) technology is used to identify key information in paper or electronic bills. When processing electronic bills, in order to save storage space, the existing technology usually dynamically adjusts the compression ratio of the bill image according to the complexity of the bill image. Specifically, when the content of the bill image is more complex (such as containing a lot of details, small fonts or more image elements), the system will automatically select a lower compression ratio to retain more details, thereby ensuring that the OCR technology can accurately identify key information. When the image content is simpler, the system will use a higher compression ratio to reduce storage requirements. This dynamic adjustment of compression can find a balance between image quality and storage space, which does not affect the recognition accuracy of OCR and optimizes storage efficiency.

[0004] The prior art has the following deficiencies:

[0005] When the compression ratio of the bill image is adjusted too large, the details in the image (especially key information such as text and numbers) may be significantly lost or become blurred. OCR technology relies on clear edges and character shapes for accurate recognition. When the image quality deteriorates, subtle differences in characters may disappear, resulting in recognition errors or unrecognizable information. If OCR cannot correctly identify key information such as invoice amount, date, invoice number, etc., the system may record incorrect financial data. These errors may not be obvious in the short term, but over time, they will gradually accumulate in financial statements, resulting in income, expenditure and tax data that are inconsistent with the actual situation. Ultimately, this accounting confusion may lead to financial risks and legal issues, seriously affecting the financial health and continued operations of the company.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for collaborative business of associated transactions based on business and financial bill data. By acquiring the image quality features of electronic bills and using a pre-trained machine learning model for evaluation, bills with the risk of over-compression can be intelligently identified, and the compression ratio can be automatically adjusted to improve the image quality, thereby improving OCR recognition accuracy, reducing financial data errors, and ensuring the reliability of financial information. In addition, the invention can dynamically adjust the image compression strategy, quickly identify high-quality and low-quality bills, reduce manual intervention, automatically optimize the compression settings of low-quality bills, and improve system efficiency. If the problem is not caused by compression, the system will automatically sound an alarm to troubleshoot, allowing enterprises to process large quantities of electronic bills more efficiently and accurately to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above-mentioned purpose, the present invention provides the following technical solution: a method for coordinating business of related transactions based on business, finance and bill payment data, comprising the following steps:

[0009] Convert paper receipts into electronic receipt images through high-resolution scanners;

[0010] After the electronic receipt is generated, the image is compressed according to the complexity of the receipt information;

[0011] The compressed electronic receipt is transmitted to the OCR system for image recognition. During the OCR recognition process, the image quality features of the electronic receipt are obtained, and the extracted image quality features are evaluated using a pre-trained machine learning model.

[0012] Based on the output of the machine learning model, the compressed electronic bills are divided into low-quality bills and high-quality bills;

[0013] For the bills classified as high quality, continue to perform image compression according to the pre-set electronic bill image compression strategy;

[0014] For low-quality bills, the compression strategy is adjusted based on the image quality of the electronic bill image, and the image quality problem caused by over-compression is improved by reducing the compression ratio.

[0015] Preferably, during the OCR recognition process, the image quality features of the electronic invoice obtained include image compression artifacts and contrast changes after image compression. After analyzing the image compression artifacts and contrast changes after image compression, artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are generated respectively. The artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are used as key features for evaluating the image quality of the electronic invoice.

[0016] Preferably, image compression artifacts refer to unreal textures, noise or distorted areas generated in the electronic receipt image due to data loss or the influence of the compression algorithm during the image compression process, and the artifacts will affect the visual quality and information clarity of the image;

[0017] The contrast change after image compression refers to the reduction of the brightness difference of the electronic receipt image during the compression process, which leads to the weakening of the distinction between elements in the image, affecting the readability and recognition effect of the electronic receipt image.

[0018] Preferably, the acquired artifact metric evaluation value and contrast loss metric evaluation value are input into a pre-trained machine learning model to generate an image quality coefficient, and the image quality of the compressed electronic receipt is intelligently predicted and evaluated through the image quality coefficient.

[0019] Preferably, the image quality coefficient generated after the compressed electronic receipt image is predicted by the machine learning model is compared and analyzed with a preset image quality coefficient reference threshold, and the electronic receipt is divided. The specific division steps are:

[0020] If the image quality coefficient is greater than or equal to the image quality coefficient reference threshold, the compressed electronic receipt is classified as a high-quality receipt;

[0021] If the image quality coefficient is less than the image quality coefficient reference threshold, the compressed electronic receipt is classified as a low-quality receipt.

[0022] Preferably, the specific steps of obtaining compression artifact information of the compressed electronic receipt and analyzing the image compression artifact to generate the artifact metric evaluation value are as follows:

[0023] The Sobel operator image processing algorithm is used to calculate the gradient of the compressed electronic receipt image to extract the edge information in the electronic receipt image. The calculation expression of the electronic receipt image gradient is:

[0024] ,

[0025] In the formula, G(x, y) is the gradient of the electronic bill image at pixel x, y, which reflects the clarity of the edge of the electronic bill image. I(x, y) is the pixel value of the electronic bill image at pixel x, y. and They are the gradient changes in the horizontal and vertical directions of the electronic receipt image respectively;

[0026] The difference analysis is performed on adjacent pixels in the electronic receipt gradient image to capture the transition anomaly in the electronic receipt image artifact area. The calculation expression is: D(x, y) = |G(x, y)-G(x+1, y)|+|G(x, y)-G(x, y+1)|, where D(x, y) represents the gradient change difference at pixels x and y;

[0027] The overall artifact metric evaluation value is generated by weighted integration of the gradient differences of all pixels. The artifact metric evaluation value calculation expression is:

[0028] ,

[0029] In the formula, AI artifact is the artifact metric evaluation value, which represents the overall artifact intensity of the electronic receipt image. W(x, y) is the weight function, which dynamically assigns weights according to the importance of the electronic receipt image area. Z is the normalization factor, which is used to balance the impact of different electronic receipt image sizes and complexities on the artifact metric evaluation value W. τ Indicates the total number of pixels of the electronic receipt image in the horizontal direction, H τ Indicates the total number of pixels of the electronic receipt image in the vertical direction.

[0030] Preferably, the specific steps of obtaining contrast change information of the compressed electronic receipt image and generating a contrast loss metric evaluation value after analyzing the contrast change of the compressed electronic receipt image are as follows:

[0031] Calculate the local brightness difference between the original electronic receipt image and the compressed electronic receipt image. The calculation expression is:

[0032] ,

[0033] Where D local (i, j) is the local brightness difference between the original electronic receipt image and the compressed electronic receipt image at pixel position (i, j), L orig (i, j) and L comp (i, j) are the brightness values ​​of the original electronic receipt image and the compressed electronic receipt image at the pixel position (i, j), respectively. α and β are parameters for adjusting the sensitivity of brightness difference, which control the weight of the local difference of the electronic receipt image in the calculation;

[0034] The global contrast change factor is calculated to quantify the degree of change in global contrast after the electronic receipt image is compressed. The calculation expression of the global contrast change factor is:

[0035] ,

[0036] In the formula, C globalis the global contrast change factor, W(i, j) is the weight function of the electronic receipt image at position (i, j), which gives each pixel a different weight to distinguish between important and unimportant areas max(L comp ) and min(L comp ) are respectively the maximum brightness value and the minimum brightness value in the compressed electronic receipt image;

[0037] Calculate the contrast loss metric evaluation value, which is used to quantify the contrast loss of the compressed electronic receipt image. The calculation expression is:

[0038] ,

[0039] In the formula, Contrast loss is the contrast loss metric evaluation value, λ is the parameter that adjusts the effect of global brightness change on local contrast loss, and is used to suppress the amplification effect of extreme changes in brightness distribution after compression on contrast loss. It is the global brightness reference value of the compressed electronic receipt image, which is calculated based on the function form of complex brightness changes.

[0040] Preferably, for low-quality bills, the compression strategy is adjusted based on the image quality of the electronic bill image to reduce the compression ratio. The specific steps are as follows:

[0041] Calculate the image compression abundance index, the calculation expression is:

[0042] ,

[0043] In the formula, Δ quality is the image quality difference measure, τ ref is the preset image quality reference threshold, C quality It is the image quality coefficient output by the machine learning model when the image quality of the current electronic receipt image is evaluated;

[0044] Calculate the new compression adjustment factor. The calculation expression is:

[0045] ,

[0046] In the formula, k adjust is the compression adjustment coefficient, which determines the adjustment amplitude of the compression ratio, and γ is the adjustment parameter that controls the sensitivity of the compression adjustment;

[0047] Adjust the compression ratio. The expression for adjustment is: P new =P current ·k adjust , where P new is the new compression ratio. The adjusted compression ratio will be optimized based on the current image quality.current is the current compression ratio.

[0048] Preferably, the electronic receipt image is compressed again using a new compression ratio that has been readjusted. After the compression is completed, the compressed electronic receipt image is evaluated again, and the image quality coefficient is output based on the machine learning model. The image quality coefficient is compared and analyzed again with a preset image quality coefficient reference threshold. If the image quality coefficient is greater than or equal to the image quality coefficient reference threshold, it indicates that the deterioration of the electronic receipt image is caused by excessive compression. If the image quality coefficient is less than the image quality coefficient reference threshold, it indicates that the deterioration of the electronic receipt image is not caused by excessive compression. At this time, an alarm needs to be issued to prompt the technician to conduct further troubleshooting to find out other reasons for the deterioration of the electronic receipt image and repair them.

[0049] The associated transaction business collaboration system based on business and financial bill data includes an electronic bill acquisition module, an image compression module, an image quality assessment module, a quality classification module, a high-quality bill processing module, and a low-quality bill optimization module;

[0050] The electronic receipt acquisition module converts paper receipts into electronic receipt images through a high-resolution scanner;

[0051] An image compression module, after generating an electronic receipt, compresses the image according to the complexity of the receipt information;

[0052] The image quality assessment module transmits the compressed electronic receipt to the OCR system for image recognition. During the OCR recognition process, the image quality features of the electronic receipt are obtained and the extracted image quality features are evaluated using a pre-trained machine learning model.

[0053] The quality classification module classifies the compressed electronic bills into low-quality bills and high-quality bills based on the output of the machine learning model;

[0054] The high-quality bill processing module continues to compress the image of the bill classified as high-quality using a preset electronic bill image compression strategy;

[0055] The low-quality bill optimization module adjusts the compression strategy for low-quality bills based on the image quality of the electronic bill image, and improves the image quality problem caused by over-compression by reducing the compression ratio.

[0056] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0057] The present invention obtains the image quality features of electronic receipts, uses a pre-trained machine learning model to evaluate the extracted image quality features, intelligently identifies electronic receipts that are at risk of over-compression, and improves the image quality of electronic receipts by adjusting the compression ratio. This intelligent prediction and adjustment mechanism ensures that the system can promptly detect and correct image blur or artifact problems caused by over-compression, thereby improving the accuracy of OCR recognition, reducing financial data errors, and ensuring the reliability of financial information.

[0058] The present invention can dynamically adjust the image compression strategy, which not only improves the quality of electronic invoice processing, but also greatly improves the operating efficiency of the system. Through the automated machine learning evaluation and classification mechanism, the system can quickly identify high-quality and low-quality electronic invoices, reducing unnecessary manual intervention and repeated operations. For high-quality invoices, the system can continue to process directly, and for low-quality invoices, the quality is automatically optimized by reducing the compression ratio, avoiding the time waste caused by manual adjustment. In addition, if the image quality problem is not caused by excessive compression, the system will automatically issue an alarm to prompt the technician to conduct more in-depth troubleshooting. This automation-based intelligent system significantly improves the efficiency of electronic invoice processing, reduces the error rate, and reduces the need to rely on manual troubleshooting, enabling enterprises to process large quantities of electronic invoices in a more efficient and accurate manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0060] Figure 1 It is a method flow chart of the collaborative method of related transaction business based on business and financial bill data of the present invention.

[0061] Figure 2 It is a module diagram of the collaborative system of associated transaction business based on business, finance and bill payment data of the present invention. DETAILED DESCRIPTION

[0062] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0063] The present invention provides Figure 1The method for collaborative business of associated transactions based on business and financial bill payment data shown includes the following steps:

[0064] Convert paper receipts into electronic receipt images through high-resolution scanners;

[0065] The key to this step is to ensure that every detail of the paper bill can be clearly captured, including the text, numbers, tables and graphics on the bill. The scanning resolution is usually set at a level sufficient to meet the requirements of subsequent OCR recognition. In the existing technology, the use of high-quality scanning equipment and scanning modes adapted to different types of paper bills (such as color, grayscale or black and white scanning) can effectively improve image clarity.

[0066] After the electronic receipt is generated, the image is compressed according to the complexity of the receipt information;

[0067] In the prior art, the complexity of a bill can be determined by analyzing the structure of the bill image, such as text density, number of tables, presence of image elements, etc. For example, bills with high text density will be marked as complex bills, requiring a lower compression ratio to retain more details. Simpler bills can use a higher compression ratio to save storage space. This process is usually achieved through a preset compression algorithm, such as JPEG or PNG compression format.

[0068] The compressed electronic receipt is transmitted to the OCR system for image recognition. During the OCR recognition process, the image quality features of the electronic receipt are obtained, and the extracted image quality features are evaluated using a pre-trained machine learning model.

[0069] During the OCR recognition process, the image quality features of the electronic receipt obtained include image compression artifacts and contrast changes after image compression. Image compression artifacts refer to unreal textures, noise or distorted areas generated in the electronic receipt image due to data loss or the influence of the compression algorithm during the image compression process. These artifacts will affect the visual quality and information clarity of the image. The contrast change after image compression refers to the brightness difference of the electronic receipt image (that is, the difference between the bright and dark parts of the electronic receipt image) being reduced during the compression process, resulting in a weakened distinction between elements in the electronic receipt image (such as characters and background), affecting the readability and recognition effect of the electronic receipt image. After analyzing the image compression artifacts and the contrast changes after image compression, artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are generated respectively. The artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are used as key features for evaluating the quality of electronic receipt images.

[0070] The acquired artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are input into a pre-trained machine learning model to generate an image quality coefficient, and the image quality of the compressed electronic receipt is intelligently predicted and evaluated through the image quality coefficient.

[0071] OCR technology relies on the clarity of the image and the integrity of the character shape for recognition. In the existing technology, the OCR algorithm automatically parses the text, numbers and other information in the image according to the preset character recognition model and converts it into machine-readable text data. The trained model can learn the impact of different image qualities on OCR recognition accuracy based on a large number of annotated bill data sets. By analyzing the image quality of electronic bills, machine learning models can predict the recognition effect of bills. Commonly used models include support vector machines (SVM), decision tree models, or models based on deep learning (such as convolutional neural networks).

[0072] A large area of ​​compression artifacts in electronic receipt images usually indicates that the image is at risk of over-compression during the compression process. Compression artifacts are a direct manifestation of loss or distortion of image details, usually manifested as blur, noise, color blocks, etc. When the area of ​​artifacts increases, key elements in the image, such as text, numbers, and table lines, may become unclear or difficult to recognize, which seriously affects the recognition accuracy of OCR technology. Excessive compression not only leads to a decrease in visual quality, but may also cause important receipt information to be unable to be accurately extracted, leading to financial data errors or processing delays. Therefore, images with a large artifact area should be considered to be at risk of over-compression.

[0073] The specific steps of obtaining the compression artifact information of the compressed electronic receipt and analyzing the image compression artifact to generate the artifact metric evaluation value are as follows:

[0074] The Sobel operator image processing algorithm is used to calculate the gradient of the compressed electronic receipt image to extract the edge information in the electronic receipt image. The calculation expression of the electronic receipt image gradient is:

[0075] ,

[0076] In the formula, G(x, y) is the gradient of the electronic bill image at pixel x, y, which reflects the clarity of the edge of the electronic bill image. The higher it is, the clearer the edge is. I(x, y) is the pixel value of the electronic bill image at pixel x, y. and They are the gradient changes in the horizontal and vertical directions of the electronic receipt image respectively;

[0077] The difference analysis of adjacent pixels in the electronic receipt gradient image is performed to capture the transition anomaly of the electronic receipt image artifact area. By calculating the change of the gradient between adjacent pixels, the obvious change of the artifact area can be identified. The calculation expression is: D(x, y) = |G(x, y)-G(x+1, y)|+|G(x, y)-G(x, y+1)|, where D(x, y) represents the difference in gradient change at pixels x and y. A larger D(x, y) value indicates that there may be artifacts in the area.

[0078] The overall artifact metric evaluation value is generated by weighted integration of the gradient differences of all pixels. The artifact metric evaluation value calculation expression is:

[0079] ,

[0080] In the formula, AI artifact is the artifact metric evaluation value, which represents the overall artifact intensity of the electronic receipt image. W(x, y) is the weight function, which dynamically assigns weights according to the importance of the electronic receipt image area. The text and detail areas are higher and the background area is lower. Z is the normalization factor, which is used to balance the impact of different electronic receipt image sizes and complexities on the artifact metric evaluation value W. τ Indicates the total number of pixels of the electronic receipt image in the horizontal direction (the width of the electronic receipt image), H τ Indicates the total number of pixels of the electronic receipt image in the vertical direction (the height of the electronic receipt image);

[0081] The value of the normalization factor Z is usually dynamically adjusted according to the size and complexity of the electronic receipt image and the specific area in the electronic receipt image to ensure that the artifact metric evaluation value is universal and comparable. Specifically, Z can be taken as the sum of the gradient differences of all pixels in the electronic receipt image, or according to the width W of the electronic receipt image. τ and height H τ Calculated ratio, such as Z = W τ ·H τ The purpose of this is to eliminate the artifact intensity imbalance caused by the difference in size or local complexity of the electronic receipt images, so that the artifact metric evaluation values ​​of different electronic receipt images are consistent. For specific application scenarios, the normalization factor can also be adjusted based on experience or preset reference electronic receipt images to reflect the actual artifact sensitivity.

[0082] It can be seen from the artifact metric evaluation value that the larger the performance value of the artifact metric evaluation value generated after analyzing the compression artifacts of the electronic receipt image, the larger the area of ​​the compression artifacts in the electronic receipt image, which indicates that the electronic receipt image may have suffered the risk of over-compression during the compression process. Specifically, the artifact metric evaluation value is a comprehensive indicator obtained by analyzing the characteristics of the gradient difference of adjacent pixels and the degree of edge blur in the electronic receipt image. Larger artifact metric evaluation values ​​usually reflect increased discontinuities in the electronic receipt image, such as the appearance of obvious compression artifacts, noise or distorted areas. These artifacts will destroy the details in the electronic receipt image, especially have an adverse effect on the clarity of key information such as text and numbers, resulting in the inability of OCR technology to accurately identify the receipt information. When the artifact metric evaluation value is high, the system should be alert to the possibility of over-compression of the electronic receipt image.

[0083] A reduction in contrast differences after compression of an electronic receipt image usually indicates that the image is at risk of being over-compressed. A reduction in contrast differences means that the distinction between bright and dark parts in the electronic receipt image is reduced, and the boundary between characters and background becomes blurred, especially for smaller text or numbers on the receipt. This phenomenon makes it difficult for OCR technology to accurately recognize information. When the contrast of an electronic receipt image decreases significantly, important details may be lost or confused, which increases the possibility of recognition errors. Therefore, a reduction in contrast is an important sign of over-compression, which directly affects the readability of electronic receipts and the accuracy of data processing.

[0084] The specific steps of obtaining contrast change information of the compressed electronic receipt image and analyzing the contrast change of the compressed electronic receipt image to generate a contrast loss evaluation value are as follows:

[0085] Calculate the local brightness difference between the original electronic receipt image and the compressed electronic receipt image. The calculation expression is:

[0086] ,

[0087] Where D local (i, j) is the local brightness difference between the original electronic receipt image and the compressed electronic receipt image at pixel position (i, j), L orig (i, j) and L comp (i, j) are the brightness values ​​of the original electronic receipt image and the compressed electronic receipt image at the pixel position (i, j), respectively. α and β are parameters for adjusting the sensitivity of brightness difference, which control the weight of the local difference of the electronic receipt image in the calculation, ensuring that the difference is enhanced in places with large contrast difference;

[0088] When setting the parameters α and β, they should be optimized according to the contrast characteristics of the electronic receipt image and the specific requirements of compression:

[0089] α controls the sensitivity of local brightness differences. For electronic receipt image areas with large differences, the value of α should be relatively small (such as between 0.1 and 0.5) to avoid over-smoothing and preserve large contrast changes; for electronic receipt image areas with small differences, the α value can be increased (such as 0.5 to 1) to reduce the amplification of these subtle differences.

[0090] β is used to adjust the nonlinear amplification effect of the difference in the calculation. Usually, β can be set between 2 and 5 to amplify the local contrast difference when the details of the electronic receipt image are complex. If the electronic receipt image content is simple or less noisy, the β value can be slightly reduced to avoid noise amplification.

[0091] By adjusting α and β, the contrast difference in different electronic receipt image areas can be flexibly enhanced or weakened, ensuring that sensitivity to details in important areas is maintained during the compression process.

[0092] The global contrast change factor is calculated to quantify the degree of change in global contrast after the electronic receipt image is compressed. The calculation expression of the global contrast change factor is:

[0093] ,

[0094] In the formula, C global is the global contrast change factor, W(i, j) is the weight function of the electronic receipt image at position (i, j), which gives each pixel a different weight to distinguish important areas from unimportant areas (W(i, j) has the same meaning as W(x, y) above, and will not be described in detail here). max(L comp ) and min(L comp ) are respectively the maximum brightness value and the minimum brightness value in the compressed electronic receipt image;

[0095] This formula is used to extract the overall brightness range of the compressed electronic receipt image and its changes in different areas, reflecting the distribution characteristics of the contrast changes of the electronic receipt image after compression.

[0096] Calculate the contrast loss metric evaluation value, which is used to quantify the contrast loss of the compressed electronic receipt image. The calculation expression is:

[0097] ,

[0098] In the formula, Contrast loss is the contrast loss metric evaluation value, λ is the parameter that adjusts the effect of global brightness change on local contrast loss, and is used to suppress the amplification effect of extreme changes in brightness distribution after compression on contrast loss. It is the global brightness reference value of the compressed electronic receipt image, which is calculated based on the function form of complex brightness changes instead of using the mean value;

[0099] The brightness evaluation is not simply measured by the average value, but by a more complex function to capture the nonlinear changes and detail differences in brightness in the electronic receipt image. The function of complex brightness changes can be expressed by nonlinear transformation, such as gradient changes based on the local brightness of the electronic receipt image, brightness contrast enhancement function, or polynomial-based curve fitting. These functions can capture the local features and details of brightness changes, making the calculation more flexible and accurate, especially suitable for the diverse brightness distribution in complex electronic receipt images, instead of only providing an overall brightness level reference like the mean. The global brightness reference value is not specifically limited here, but is flexibly set according to actual needs and scenarios. Specifically, for different electronic receipt image processing tasks, the global brightness reference value can be adjusted according to the characteristics of the electronic receipt image, contrast requirements, and recognition accuracy requirements. For example, in a high-contrast electronic receipt image scene, the global brightness reference value can select brighter or darker pixel values ​​to enhance the contrast of a specific area; in an image with uniform brightness distribution, the reference value can be set based on a nonlinear brightness function or a local brightness gradient. This flexible setting can better adapt to different application needs and improve the effect of electronic receipt image processing.

[0100] The performance value of the contrast loss metric evaluation value can be used to judge the compression quality of the electronic receipt image. The larger the performance value of the contrast loss metric evaluation value generated after analyzing the contrast change of the electronic receipt image after compression, the greater the contrast change of the electronic receipt image during the compression process, especially the difference between the light and dark areas in the electronic receipt image is greatly weakened. This means that key text, numbers and other information may become blurred or difficult to distinguish due to the reduction in contrast, thereby increasing the risk of OCR technology recognition failure. Therefore, a higher contrast loss metric evaluation value reflects the potential risk of over-compression of the electronic receipt image, which may lead to the loss of important details, and ultimately affect the accuracy of financial data, processing efficiency and the overall reliability of the system.

[0101] The machine learning model is not specifically limited here, and can realize the AI ​​​​of artifact metric evaluation value artifact And contrast loss metric evaluation value Contrast loss Perform comprehensive analysis to generate image quality coefficient Image Quality In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; the image quality coefficient Image Quality The resulting calculation formula is:

[0102] ,

[0103] In the formula, They are respectively the artifact metric evaluation value AI artifact And contrast loss metric evaluation value Contrast loss The preset scaling factor of Both are greater than 0.

[0104] It can be seen from the image quality coefficient that the greater the performance value of the artifact metric evaluation value generated after analyzing the image compression artifact, the greater the performance value of the contrast loss metric evaluation value generated after analyzing the contrast change after compression of the electronic receipt image. In other words, the smaller the performance value of the image quality coefficient generated after the compressed electronic receipt image is predicted by the machine learning model, the worse the quality of the compressed electronic receipt image, and vice versa.

[0105] Based on the output of the machine learning model, the compressed electronic bills are divided into low-quality bills and high-quality bills;

[0106] The image quality coefficient generated by predicting the compressed electronic receipt image through the machine learning model is compared and analyzed with the preset image quality coefficient reference threshold to divide the electronic receipt. The specific division steps are as follows:

[0107] If the image quality coefficient is greater than or equal to the image quality coefficient reference threshold, the compressed electronic receipt is classified as a high-quality receipt; a high-quality receipt usually means a higher image clarity and better OCR recognition accuracy, and the system can directly use the standard compression and recognition process.

[0108] If the image quality coefficient is less than the image quality coefficient reference threshold, the compressed electronic receipt is classified as a low-quality receipt; a low-quality receipt indicates that the image has problems such as blur and compression artifacts, which may affect the OCR recognition effect and require further processing. This classification can help the system adopt different optimization strategies for different image qualities during the processing process.

[0109] For the bills classified as high quality, continue to perform image compression according to the pre-set electronic bill image compression strategy;

[0110] At this time, due to the high image quality, the system can appropriately increase the compression ratio to maximize storage space savings without significantly affecting the accuracy of OCR recognition. In the prior art, lossy compression in JPEG format is a commonly used method. By adjusting the compression ratio, the file size can be effectively reduced while maintaining good image quality.

[0111] For low-quality bills, the compression strategy is adjusted based on the image quality of the electronic bill image, and the image quality problem caused by over-compression is improved by reducing the compression ratio;

[0112] Specifically, the system will reduce the compression ratio, which can significantly improve image quality problems caused by over-compression. The adjusted image compression strategy may use a lossless compression format such as PNG, or adopt a lower lossy compression ratio, thereby reducing storage space waste while ensuring OCR recognition accuracy.

[0113] For low-quality bills, the compression strategy is adjusted based on the image quality of the electronic bill image to reduce the compression ratio. The specific steps are as follows:

[0114] Calculate the image compression abundance index. The calculation expression is:

[0115] ,

[0116] In the formula, Δ quality is the image quality difference metric, which indicates the difference of the image quality of the current electronic receipt relative to the reference threshold, τ ref is the preset image quality reference threshold, C quality It is the image quality coefficient output by the machine learning model when the current electronic receipt image is evaluated for image quality;

[0117] This step is used to evaluate the difference between the current quality of the image and the ideal quality and perform nonlinear upscaling, especially when the difference is large.

[0118] Calculate the new compression adjustment factor. The calculation expression is:

[0119] ,

[0120] In the formula, k adjust is the compression adjustment coefficient, which determines the adjustment range of the compression ratio. γ is the adjustment parameter that controls the sensitivity of the compression adjustment. When the image quality difference Δ quality When it is larger, the compression adjustment coefficient k adjust It becomes smaller, indicating that the compression ratio will be significantly reduced to improve image quality;

[0121] Through this adjustment factor, the system can dynamically adjust the compression ratio according to the difference in image quality.

[0122] Adjust the compression ratio. The expression for adjustment is: P new =P current ·k adjust , where P new is the new compression ratio. The adjusted compression ratio will be optimized based on the current image quality.current is the current compression ratio;

[0123] This step dynamically reduces the compression ratio by calculating a new compression ratio to reduce artifacts and contrast loss caused by over-compression.

[0124] The electronic receipt image is compressed again using the readjusted new compression ratio. After the compression is completed, the compressed electronic receipt image is evaluated again, and the image quality coefficient is output based on the machine learning model. The image quality coefficient is compared and analyzed again with the preset image quality coefficient reference threshold. If the image quality coefficient is greater than or equal to the image quality coefficient reference threshold, it indicates that the deterioration of the electronic receipt image is caused by excessive compression. The problem can be solved by reducing the compression ratio. This means that the compression strategy is the root cause of the problem, and the image quality has been restored after adjustment. If the image quality coefficient is less than the image quality coefficient reference threshold, it indicates that the deterioration of the electronic receipt image is not caused by excessive compression, and there may be other problems (such as scanning errors, image damage, etc.). At this time, an alarm needs to be issued to prompt the technician to conduct further troubleshooting to find other reasons for the deterioration of the electronic receipt image and repair them. Ensure that the system can intelligently distinguish the source of image quality problems and take corresponding solutions to improve the accuracy and efficiency of electronic receipt processing.

[0125] The present invention obtains the image quality features of electronic receipts, uses a pre-trained machine learning model to evaluate the extracted image quality features, intelligently identifies electronic receipts that are at risk of over-compression, and improves the image quality of electronic receipts by adjusting the compression ratio. This intelligent prediction and adjustment mechanism ensures that the system can promptly detect and correct image blur or artifact problems caused by over-compression, thereby improving the accuracy of OCR recognition, reducing financial data errors, and ensuring the reliability of financial information.

[0126] The present invention can dynamically adjust the image compression strategy, which not only improves the quality of electronic invoice processing, but also greatly improves the operating efficiency of the system. Through the automated machine learning evaluation and classification mechanism, the system can quickly identify high-quality and low-quality electronic invoices, reducing unnecessary manual intervention and repeated operations. For high-quality invoices, the system can continue to process directly, and for low-quality invoices, the quality is automatically optimized by reducing the compression ratio, avoiding the time waste caused by manual adjustment. In addition, if the image quality problem is not caused by excessive compression, the system will automatically issue an alarm to prompt the technician to conduct more in-depth troubleshooting. This automation-based intelligent system significantly improves the efficiency of electronic invoice processing, reduces the error rate, and reduces the need to rely on manual troubleshooting, enabling enterprises to process large quantities of electronic invoices in a more efficient and accurate manner.

[0127] The present invention provides Figure 2 The associated transaction business collaboration system based on business and financial bill data shown includes an electronic bill acquisition module, an image compression module, an image quality assessment module, a quality classification module, a high-quality bill processing module, and a low-quality bill optimization module;

[0128] The electronic receipt acquisition module converts paper receipts into electronic receipt images through a high-resolution scanner;

[0129] An image compression module, after generating an electronic receipt, compresses the image according to the complexity of the receipt information;

[0130] The image quality assessment module transmits the compressed electronic receipt to the OCR system for image recognition. During the OCR recognition process, the image quality features of the electronic receipt are obtained and the extracted image quality features are evaluated using a pre-trained machine learning model.

[0131] The quality classification module classifies the compressed electronic bills into low-quality bills and high-quality bills based on the output of the machine learning model;

[0132] The high-quality bill processing module continues to compress the image of the bill classified as high-quality using a preset electronic bill image compression strategy;

[0133] Low-quality bill optimization module, for low-quality bills, adjusts the compression strategy based on the image quality of the electronic bill image, and improves the image quality problem caused by over-compression by reducing the compression ratio;

[0134] The related transaction business collaboration method based on business, finance, and bill and payment data provided in an embodiment of the present invention is realized through the above-mentioned related transaction business collaboration system based on business, finance, and bill and payment data. The specific methods and processes of the related transaction business collaboration system based on business, finance, and bill and payment data are detailed in the above-mentioned embodiment of the related transaction business collaboration method based on business, finance, and bill and payment data, which will not be repeated here.

[0135] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0136] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0137] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A method for collaborative business of related transactions based on business and financial bill data, characterized in that: The following steps are involved: Convert paper receipts into electronic receipt images through high-resolution scanners; After the electronic receipt is generated, the image is compressed according to the complexity of the receipt information; The compressed electronic receipt is transmitted to the OCR system for image recognition. During the OCR recognition process, the image quality features of the electronic receipt are obtained, and the extracted image quality features are evaluated using a pre-trained machine learning model. Based on the output of the machine learning model, the compressed electronic bills are divided into low-quality bills and high-quality bills; For the bills classified as high quality, continue to perform image compression according to the pre-set electronic bill image compression strategy; For low-quality bills, the compression strategy is adjusted based on the image quality of the electronic bill image, and the image quality problem caused by over-compression is improved by reducing the compression ratio.

2. The method for collaborative business of related transactions based on business and financial bill data according to claim 1 is characterized in that: During the OCR recognition process, the image quality features of the electronic invoice obtained include image compression artifacts and contrast changes after image compression. After analyzing the image compression artifacts and contrast changes after image compression, artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are generated respectively. The artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are used as key features for evaluating the image quality of the electronic invoice.

3. The method for collaborative business of related transactions based on business and financial bill payment data according to claim 2 is characterized in that: Image compression artifacts refer to unreal textures, noise or distorted areas in electronic receipt images due to data loss or compression algorithm during the image compression process. Artifacts will affect the visual quality and information clarity of the image. The contrast change after image compression refers to the reduction of the brightness difference of the electronic receipt image during the compression process, which leads to the weakening of the distinction between elements in the image, affecting the readability and recognition effect of the electronic receipt image.

4. The method for collaborative business of related transactions based on business and financial bill data according to claim 2 is characterized in that: The acquired artifact metric evaluation values ​​and contrast loss metric evaluation values ​​are input into a pre-trained machine learning model to generate an image quality coefficient, and the image quality of the compressed electronic receipt is intelligently predicted and evaluated through the image quality coefficient.

5. The method for collaborative business of related transactions based on business and financial bill data according to claim 4 is characterized in that: The image quality coefficient generated by predicting the compressed electronic receipt image through the machine learning model is compared and analyzed with the preset image quality coefficient reference threshold to divide the electronic receipt. The specific division steps are as follows: If the image quality coefficient is greater than or equal to the image quality coefficient reference threshold, the compressed electronic receipt is classified as a high-quality receipt; If the image quality coefficient is less than the image quality coefficient reference threshold, the compressed electronic receipt is classified as a low-quality receipt.

6. The method for collaborative business of related transactions based on business and financial bill data according to claim 4 is characterized in that: The specific steps of obtaining the compression artifact information of the compressed electronic receipt and analyzing the image compression artifact to generate the artifact metric evaluation value are as follows: The Sobel operator image processing algorithm is used to calculate the gradient of the compressed electronic receipt image to extract the edge information in the electronic receipt image. The calculation expression of the electronic receipt image gradient is: ， In the formula, G(x, y) is the gradient of the electronic bill image at pixel x, y, which reflects the clarity of the edge of the electronic bill image. I(x, y) is the pixel value of the electronic bill image at pixel x, y. and They are the gradient changes in the horizontal and vertical directions of the electronic receipt image respectively; The difference analysis is performed on adjacent pixels in the electronic receipt gradient image to capture the transition anomaly in the electronic receipt image artifact area. The calculation expression is: D(x, y) = |G(x, y)-G(x+1, y)|+|G(x, y)-G(x, y+1)|, where D(x, y) represents the gradient change difference at pixels x and y; The overall artifact metric evaluation value is generated by weighted integration of the gradient differences of all pixels. The artifact metric evaluation value calculation expression is: ， In the formula, AI artifact is the artifact metric evaluation value, which represents the overall artifact intensity of the electronic receipt image. W(x, y) is the weight function, which dynamically assigns weights according to the importance of the electronic receipt image area. Z is the normalization factor, which is used to balance the impact of different electronic receipt image sizes and complexities on the artifact metric evaluation value W. τ Indicates the total number of pixels of the electronic receipt image in the horizontal direction, H τ Indicates the total number of pixels of the electronic receipt image in the vertical direction.

7. The method for collaborative business of related transactions based on business and financial bill data according to claim 4 is characterized in that: The specific steps of obtaining contrast change information of the compressed electronic receipt image and analyzing the contrast change of the compressed electronic receipt image to generate a contrast loss metric evaluation value are as follows: Calculate the local brightness difference between the original electronic receipt image and the compressed electronic receipt image. The calculation expression is: ， Where D local (i, j) is the local brightness difference between the original electronic receipt image and the compressed electronic receipt image at pixel position (i, j), L orig (i, j) and L comp (i, j) are the brightness values ​​of the original electronic receipt image and the compressed electronic receipt image at the pixel position (i, j), respectively. α and β are parameters for adjusting the sensitivity of brightness difference, which control the weight of the local difference of the electronic receipt image in the calculation; The global contrast change factor is calculated to quantify the degree of change in global contrast after the electronic receipt image is compressed. The calculation expression of the global contrast change factor is: ， In the formula, C global is the global contrast change factor, W(i, j) is the weight function of the electronic receipt image at position (i, j), which gives each pixel a different weight to distinguish between important and unimportant areas max(L comp ) and min(L comp ) are respectively the maximum brightness value and the minimum brightness value in the compressed electronic receipt image; Calculate the contrast loss metric evaluation value, which is used to quantify the contrast loss of the compressed electronic receipt image. The calculation expression is: ， In the formula, Centrast loss is the contrast loss metric evaluation value, λ is the parameter that adjusts the effect of global brightness change on local contrast loss, and is used to suppress the amplification effect of extreme changes in brightness distribution after compression on contrast loss. It is the global brightness reference value of the compressed electronic receipt image, which is calculated based on the function form of complex brightness changes.

8. The method for collaborative business of related transactions based on business and financial bill data according to claim 5 is characterized in that: For low-quality bills, the compression strategy is adjusted based on the image quality of the electronic bill image to reduce the compression ratio. The specific steps are as follows: Calculate the image compression abundance index, the calculation expression is: ， In the formula, Δ quality is the image quality difference measure, τ ref is the preset image quality reference threshold, C quality It is the image quality coefficient output by the machine learning model when the image quality of the current electronic receipt image is evaluated; Calculate the new compression adjustment factor. The calculation expression is: ， In the formula, k adjust is the compression adjustment coefficient, which determines the adjustment amplitude of the compression ratio, and γ is the adjustment parameter that controls the sensitivity of the compression adjustment; Adjust the compression ratio. The expression for adjustment is: P new =P current ·k adjust , where P new is the new compression ratio. The adjusted compression ratio will be optimized based on the current image quality. current is the current compression ratio.

9. The method for collaborative business of related transactions based on business and financial bill data according to claim 8 is characterized in that: The electronic receipt image is compressed again using the readjusted new compression ratio. After the compression is completed, the compressed electronic receipt image is evaluated again, and the image quality coefficient is output based on the machine learning model. The image quality coefficient is compared and analyzed again with the preset image quality coefficient reference threshold. If the image quality coefficient is greater than or equal to the image quality coefficient reference threshold, it indicates that the deterioration of the electronic receipt image is caused by excessive compression. If the image quality coefficient is less than the image quality coefficient reference threshold, it indicates that the deterioration of the electronic receipt image is not caused by excessive compression. At this time, an alarm needs to be issued to prompt the technician to conduct further troubleshooting to find out other reasons for the deterioration of the electronic receipt image and repair them.

10. A collaborative system for related transactions based on business, finance and bills data, used to implement the collaborative method for related transactions based on business, finance and bills data as described in any one of claims 1 to 9, characterized in that: It includes an electronic bill acquisition module, an image compression module, an image quality assessment module, a quality classification module, a high-quality bill processing module, and a low-quality bill optimization module; The electronic receipt acquisition module converts paper receipts into electronic receipt images through a high-resolution scanner; An image compression module, after generating an electronic receipt, compresses the image according to the complexity of the receipt information; The image quality assessment module transmits the compressed electronic receipt to the OCR system for image recognition. During the OCR recognition process, the image quality features of the electronic receipt are obtained and the extracted image quality features are evaluated using a pre-trained machine learning model. The quality classification module classifies the compressed electronic bills into low-quality bills and high-quality bills based on the output of the machine learning model; The high-quality bill processing module continues to compress the image of the bill classified as high-quality using a preset electronic bill image compression strategy; The low-quality bill optimization module adjusts the compression strategy for low-quality bills based on the image quality of the electronic bill image, and improves the image quality problem caused by over-compression by reducing the compression ratio.

Images