Paper document inspection method, device, equipment, medium and program product

Abstract

The application provides a paper file inspection method, device, equipment, medium and program product, which can be applied to the technical field of Internet of Things. The method comprises the following steps: acquiring a photo and a spectrum image of a paper file; performing feature extraction on the photo and the spectrum image to obtain a first page paper feature, a first page ink feature and a page edge feature; performing aging degree calculation according to the first page paper feature, the page edge feature and an environmental parameter of the paper file to obtain a paper aging degree score; performing ink diffusion prediction according to the first page paper feature, the page edge feature, the first page ink feature and the environmental parameter to obtain an ink diffusion degree score; the environmental parameter comprises at least one of temperature, humidity, file weight and illumination intensity; and performing weighted fusion on the paper aging degree score and the ink diffusion degree score to obtain a fusion quality score.

Description

Technical Field

[0001] This application relates to the field of the Internet of Things, specifically to a method, apparatus, equipment, medium, and program product for inspecting paper documents. Background Technology

[0002] In scenarios such as receiving pledged assets, banking institutions need to receive original warehouse receipts, pledge contracts, ownership certificates, original insurance policies, and other documents, which need to be stored long-term. However, in some cases, when these documents are needed, problems such as paper aging or blurred handwriting occur due to long-term storage, causing adverse effects on the bank and losses to relevant parties. Therefore, a paper document inspection method is needed to promptly identify these problems and provide early warnings. Summary of the Invention

[0003] In view of the above problems, embodiments of this application provide a paper document inspection method, apparatus, equipment, medium, and program product.

[0004] According to a first aspect of this application, a paper document inspection method is provided, comprising: acquiring photographs and spectral images of the paper document, wherein the spectral images characterize the reflectivity of the paper document to light in multiple wavelength bands; extracting features from the photographs and spectral images to obtain first-page paper features, first-page ink features, and page edge features, wherein the first-page paper features indicate the paper type and paper aging degree of the top page of the paper document, the first-page ink features indicate the text clarity of the top page, and the page edge features indicate the paper thickness and aging degree of the middle pages of the paper document; calculating the aging degree based on the first-page paper features, page edge features, and environmental parameters of the paper document to obtain a paper aging degree score; predicting ink diffusion based on the first-page paper features, page edge features, first-page ink features, and environmental parameters to obtain an ink diffusion degree score, wherein the environmental parameters include at least one of temperature, humidity, document weight, and light intensity; and weightedly fusing the paper aging degree score and the ink diffusion degree score to obtain a fusion quality score.

[0005] According to embodiments of this application, feature extraction from photographs and spectral images to obtain first-page paper features includes: feature extraction from photographs to obtain paper features; feature extraction from spectral images to obtain light reflectance; paper features include paper size, first-page paper color, surface smoothness, and paper texture features; paper size is calculated based on the photograph's acquisition location, the distance to the paper document, and the number of paper pixels in the photograph; color, surface smoothness, paper texture features, and light reflectance are fused and analyzed to obtain paper type and first-page aging degree; and paper size, paper type, and first-page aging degree are used as first-page paper features.

[0006] According to an embodiment of this application, the aging degree is calculated based on the first page paper characteristics, page edge characteristics, and environmental parameters of the paper document to obtain a paper aging degree score. This includes: vertically propagating information about the aging degree of the first page based on environmental parameters and paper type to obtain a first vertical feature, which characterizes the degree of paper aging at the bottom of the paper document caused by pressure due to weight under temperature and humidity conditions; fusing the first vertical feature and the first horizontal feature to obtain a fused feature; wherein, the first horizontal feature is obtained by horizontally propagating information about the bottom page edge features in the page edge characteristics based on environmental parameters and paper size, and the first horizontal feature indicates the degree of aging at the bottom center of the paper document; and obtaining a paper aging degree score based on the score corresponding to the fused feature mapping.

[0007] According to embodiments of this application, feature extraction of photographs and spectral images to obtain ink blot features includes: performing optical character recognition on the photograph to obtain the recognized characters and the confidence scores corresponding to the recognized characters, and calculating the mean and standard deviation of the confidence scores; performing sharpness analysis on the character boundaries corresponding to the standard glyphs of the recognized characters in the photograph to obtain the ink blot boundary sharpness, which characterizes the clarity of the character boundaries in the file; extracting the light reflectance of the region corresponding to the recognized characters in the spectral image, calculating the average value of the difference between the light reflectance and a preset reference value to obtain the grayscale uniformity; and using the mean and standard deviation of the confidence scores, the ink blot boundary sharpness, and the grayscale uniformity as ink blot features.

[0008] According to an embodiment of this application, before performing optical character recognition on a photograph to obtain the recognized characters, handwritten characters or seals in the photograph are identified and determined; if handwritten characters or seals exist in the photograph, the target area where the handwritten characters or seals are located in the photograph is extracted, and the photograph is replaced with the target area to perform optical character recognition on the target area to obtain the recognized characters.

[0009] According to embodiments of this application, ink diffusion prediction based on first page paper characteristics, page edge characteristics, first page ink characteristics, and environmental parameters includes: predicting vertical diffusion of ink boundary sharpness and grayscale uniformity based on environmental parameters, paper type, paper thickness, mean and standard deviation of confidence level, to obtain vertical ink features, which characterize the degree of ink smudging between upper and lower pages of a paper document; predicting horizontal diffusion of ink boundary sharpness and grayscale uniformity based on environmental parameters, paper type, paper size, mean and standard deviation of confidence level, to obtain horizontal ink features, which characterize the degree of ink smudging in the horizontal direction of each page of paper, with paper size obtained by feature extraction from photographs; and performing feature fusion and scoring operations based on vertical and horizontal ink features to obtain an ink diffusion score.

[0010] According to embodiments of this application, during pre-identified low system load periods, feature extraction is performed based on photographs and spectral images to obtain first page paper features, ink features, and page edge features.

[0011] According to a second aspect of this application, a paper document inspection device is provided, comprising: a data acquisition module for acquiring photographs and spectral images of paper documents, wherein the spectral images characterize the reflectivity of the paper documents to light in multiple wavelength bands; a feature extraction module for extracting features from the photographs and spectral images to obtain first-page paper features, first-page ink features, and page edge features, wherein the first-page paper features indicate the paper type and paper aging degree of the top page of the paper document, the first-page ink features indicate the text clarity of the top page, and the page edge features indicate the paper thickness of the middle pages and the aging degree of the page edges; an inference module for calculating the aging degree based on the first-page paper features, page edge features, and environmental parameters of the paper document to obtain a paper aging degree score, and predicting ink diffusion based on the first-page paper features, page edge features, first-page ink features, and environmental parameters to obtain an ink diffusion degree score, wherein the environmental parameters include at least one of temperature, humidity, document weight, and light intensity; and an evaluation module for weighted fusion of the paper aging degree score and the ink diffusion degree score to obtain a fusion quality score.

[0012] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.

[0013] According to a fourth aspect of this application, a computer-readable storage medium is also provided, on which a computer program or instructions are stored, wherein the computer program or instructions, when executed by a processor, implement the steps of the above-described method.

[0014] According to a fifth aspect of this application, a computer program product is also provided, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method. Attached Figure Description

[0015] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0016] Figure 1 The illustrations depict application scenarios of paper document inspection methods, apparatuses, devices, media, and program products according to embodiments of this application.

[0017] Figure 2 A flowchart illustrating a paper document inspection method according to an embodiment of this application is shown schematically.

[0018] Figure 3 The illustration shows a schematic diagram of the process for extracting features of the first page of paper according to an embodiment of this application;

[0019] Figure 4 The illustration shows a schematic diagram of the process for extracting page edge features according to an embodiment of this application;

[0020] Figure 5 The illustration shows a flowchart of extracting ink features from the homepage according to an embodiment of this application;

[0021] Figure 6 The illustration shows a flowchart of extracting ink features from the first page of a signature / stamp portion according to an embodiment of this application;

[0022] Figure 7 This schematic diagram illustrates the structural block diagram of a paper document inspection device according to an embodiment of the present application;

[0023] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a paper document inspection method according to an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0026] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0027] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0028] As used in this paper, the term "model" refers to a model that learns the relationship between inputs and outputs from training data, enabling it to generate corresponding outputs for a given input after training. Model generation can be based on machine learning techniques. Deep learning is a machine learning algorithm that processes inputs and provides corresponding outputs using multiple layers of processing units. A neural network model is an example of a deep learning-based model. In this paper, "model" may also be referred to as a "machine learning model," "learning model," "machine learning network," or "learning network," and these terms are used interchangeably.

[0029] It's important to note that the term "neural network" can refer to a machine learning network based on deep learning. A neural network processes input and provides corresponding output, typically consisting of an input layer, an output layer, and one or more hidden layers between them. Neural networks used in deep learning applications often include many hidden layers, increasing the network's depth. The layers of a neural network are connected sequentially, so that the output of the previous layer serves as the input to the next layer. The input layer receives the input to the neural network, while the output layer's output becomes the final output. Each layer of a neural network includes one or more nodes (also called processing nodes or neurons), each processing the input from the layer above.

[0030] It should be understood that machine learning generally includes three phases: training, testing, and application (also known as inference). In the training phase, a given model is trained using a large amount of training data, iteratively updating parameter values ​​until the model can consistently generate inferences that meet the expected goals from the training data. Through training, the model can be considered to have learned the relationship between inputs and outputs (also known as the input-output mapping) from the training data. The parameter values ​​of the trained model are determined. In the testing phase, test inputs are applied to the trained model to test whether it can provide the correct output, thus determining the model's performance. In the application phase, the model can be used to process actual inputs based on the trained parameter values ​​to determine the corresponding output.

[0031] In one or more embodiments described herein, the term "large model" can refer to a deep learning model with a large number of model parameters, which can include hundreds of millions, tens of billions, hundreds of billions, trillions, or even tens of trillions of model parameters. Large models can also be called foundational models or basic models. They are pre-trained using large-scale unlabeled corpora to produce pre-trained models with hundreds of millions of parameters. Such models can adapt to a wide range of downstream tasks and have good generalization ability, such as large language models and multimodal pre-trained models. It should be understood that in practical applications, large models only require a small number of samples to fine-tune the pre-trained model before being applied to different tasks. Large models can be widely used in natural language processing, computer vision, and other fields. Specifically, they can be applied to computer vision tasks such as visual question answering, image captioning, and image generation, as well as natural language processing tasks such as text-based sentiment classification, text summarization, and machine translation. Major application scenarios for large models can include digital assistants, intelligent robots, search, online education, office software, e-commerce, and intelligent design.

[0032] Figure 1 The illustration schematically depicts an application scenario of a paper document inspection method, apparatus, device, medium, and program product according to embodiments of this application. It should be noted that... Figure 1 The examples shown are merely examples of scenarios in which the embodiments of this application can be applied, to help those skilled in the art understand the technical content of this application, but do not mean that the embodiments of this application cannot be used in other devices, systems, environments or scenarios.

[0033] like Figure 1 As shown, application scenario 100 according to this embodiment may include a file archiving cabinet or file safe application scenario. Network 104 is used as a medium to provide a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0034] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be devices equipped with cameras, multispectral sensors, temperature sensors, humidity sensors, and light intensity sensors.

[0035] Server 105 can be a server providing various services, such as a backend management server (for example only) supporting transactions supported by the first terminal device 101, the second terminal device 102, and the third terminal device 103. The backend management server can analyze and process received user requests and other data, and then feed the processing results back to the terminal devices. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services such as cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks, and basic cloud computing services such as big data.

[0036] It should be noted that the paper document inspection method provided in this application embodiment can generally be executed by server 105 and / or terminal devices 101-103. Accordingly, the paper document inspection device provided in this application embodiment can generally be installed in server 105 and / or terminal devices 101-103.

[0037] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0038] Figure 2 A flowchart illustrating a paper document inspection method according to an embodiment of this application is shown schematically. Figure 2 As shown, the paper document inspection method 200 according to the embodiments of this application may include steps S210 to S240.

[0039] In step S210, a photograph and spectral image of the paper document are acquired. The spectral image characterizes the reflectivity of the paper document to light in multiple wavelengths. In step S220, feature extraction is performed on the photograph and spectral image to obtain the first page paper features, the first page ink features, and the page edge features. The first page paper features indicate the paper type and paper aging degree of the top page of the paper document, the first page ink features indicate the text clarity of the top page, and the page edge features indicate the paper thickness and aging degree of the middle pages of the paper document. In step S230, the aging degree is calculated based on the first page paper features, page edge features, and environmental parameters of the paper document to obtain a paper aging degree score. Ink diffusion is predicted based on the first page paper features, page edge features, first page ink features, and environmental parameters to obtain an ink diffusion degree score. The environmental parameters include at least one of temperature, humidity, document weight, and light intensity. In step S240, the paper aging degree score and the ink diffusion degree score are weighted and fused to obtain a fusion quality score.

[0040] For example, institutions such as bank warehouses (archives / voucher repositories) must maintain a large number of paper documents. In legal, risk control, regulatory, and pledge scenarios, computer systems cannot yet completely replace paper documents. These paper documents include legally mandated paper-based certificates of rights, such as paper warehouse receipts, pledge contracts, collateral handover lists, original ownership certificates, and original insurance policies. They also include various documents requiring signature confirmation, such as risk notices and demand letters. Some of these documents need to be stored long-term and retrieved only when needed. However, during long-term storage, these paper documents may be affected by the storage environment (temperature, humidity, light, dust, corrosive gases, etc.), leading to paper deterioration (e.g., becoming brittle or sticking together), blurred handwriting, and ink seeping between documents. This results in paper documents being difficult to handle properly or distorted information. Therefore, it is necessary to inspect paper documents regularly to intervene promptly when document deterioration is detected.

[0041] In step S210, obtaining photographs of the paper documents can be achieved by using a high-definition camera to take front and side photos of the stacked documents. The front photo includes the first page of the stacked documents and the sides of the stack, i.e., taking a front photo of the paper documents parallel to the ground and a side photo of the paper documents perpendicular to the ground. The side photos can include side photos from multiple angles. The front photo provides information about the paper of the first page and information related to the handwriting / ink marks in the document; the side photos show the edge condition of the stacked document pages, such as edge curling, yellowing, etc.

[0042] In step S210, the photos and spectral images can be acquired according to a preset time interval, for example, the time interval can be one month or one week.

[0043] In step S210, acquiring the spectral image of the paper document can be achieved by using a multispectral imager or a hyperspectral imager to acquire spectral images of the front and side views of the document, where each pixel of the spectral image is a spectral curve. The spectral image is essentially a three-dimensional data block (x, y, λ), where (x, y) represents the pixel coordinates of the document, and λ represents the wavelength dimension. For example, the wavelength dimension can include multiple bands, and the spectral image can indicate the reflectance of multiple bands. Exemplarily, the bands include multiple bands in the visible light band (450~700nm) and near-infrared shortwave (700~900nm), such as the 450nm, 550nm, 600nm, 650nm, 700nm, and 750nm bands.

[0044] In step S210, for documents stored in paper document boxes, a probe camera can be used to capture photos, and a multi-band spectral detector can be used to collect spectral data.

[0045] In step S210, photos and spectral data are typically collected during the day and sent to a computer for data processing via the Internet of Things (IoT) for further data processing.

[0046] In step S220, since the first page photo records the paper color, paper texture, paper size (i.e. the length and width of the paper) and the handwriting on it, which may include the signature or seal of relevant personnel; different paper fibers, different handwriting colors, the composition of the ink used for writing, the composition of the ink pad for stamping, the initial clarity of the stamp, etc. will all affect the reflectivity of the paper under different wavelengths of light. Therefore, by combining the photo and spectral data, comprehensive feature information of the first page can be obtained.

[0047] In step S220, in the stacked documents, except for the first page, the aging of the page edges can only be observed from this side, such as yellowing or paper curling due to the influence of water molecules in the air. Based on the side photograph and the spectral image of the side, relevant information about the changes in the page edges can be obtained.

[0048] In step S230, two dimensions of paper document deterioration are considered: the vertical dimension perpendicular to the ground and the horizontal dimension parallel to the ground. In the vertical dimension, firstly, due to gravity, pressure affects paper aging. For example, stacking can cause stress concentration on the bottom of the paper, and under long-term pressure, the paper may experience fiber creep and reduced flexibility. Secondly, pressure also affects ink changes. Under pressure, ink particles may physically move, and the paper exhibits capillary action, potentially causing ink to penetrate or move between pages. Both of these factors affect the quality of the paper document. In the horizontal dimension, or the transverse dimension, the sides / edges of stacked items are the deterioration front. First, air or moisture can penetrate through gaps, and oxidation and other phenomena can spread from the edges inwards; therefore, the deterioration of the paper quality spreads from the outside inwards. Second, in addition to paper issues, ink particles may spread along the paper fibers in the horizontal direction, causing ink bleeding and making signatures and seals blurry. For documents in bank vaults, blurred handwriting or paper aging can lead to easy breakage when handling, resulting in document quality problems.

[0049] In step S230, the environmental parameters may include at least one of temperature, humidity, light intensity, and pressure caused by the weight of the paper document.

[0050] In step S230, environmental parameters can be collected by setting sensors such as temperature sensors, humidity sensors, and pressure sensors. The Internet of Things (IoT) technology is used to network the various sensors, as well as probe cameras and multi-band spectral detectors, and the data collected in the IoT is used to inspect paper documents.

[0051] In step S230, considering the two dimensions of paper deterioration and the characteristics of paper deterioration and ink diffusion or migration in the two dimensions, paper deterioration and ink diffusion are evaluated separately to obtain paper aging score and ink diffusion score respectively.

[0052] In step S240, the paper aging score and ink diffusion score are weighted and fused to obtain a fusion quality score. For example, the fusion quality score can be obtained by weighted summation based on preset weights. This fusion quality score represents the quality of the paper. The more severe the paper aging or the more severe the ink diffusion, the lower the fusion quality score. An early warning is issued when the fusion quality score is lower than a preset threshold to remind relevant personnel to intervene in a timely manner.

[0053] In step S240, the quality score indicates the degree of paper deterioration. This quality refers to the ease with which the document is damaged when it is taken out. Therefore, in such cases, the document can be updated in advance.

[0054] In step S240, the photos, spectral images, and corresponding fusion quality scores can be stored on the blockchain to prevent tampering.

[0055] According to embodiments of this application, to address the issue that paper documents stored by institutions such as banks may suffer from paper aging or blurred handwriting, affecting the usability and accuracy of the documents, the surface features of the paper documents are identified by collecting photographs and spectral images of the paper documents. Based on these surface features, the internal characteristics of the paper are inferred, thereby predicting the internal quality score of the paper based on its appearance. This allows for early warning and intervention for the deterioration of paper documents, avoiding the discovery of document quality degradation only when the documents are used, and reducing the adverse effects of document quality degradation on relevant parties.

[0056] According to embodiments of this application, feature extraction from photographs and spectral images to obtain first-page paper features includes: feature extraction from photographs to obtain paper features; feature extraction from spectral images to obtain light reflectance; paper features include paper size, first-page paper color, surface smoothness, and paper texture features; paper size is calculated based on the photograph's acquisition location, the distance to the paper document, and the number of paper pixels in the photograph; color, surface smoothness, paper texture features, and light reflectance are fused and analyzed to obtain paper type and first-page aging degree; and paper size, paper type, and first-page aging degree are used as first-page paper features.

[0057] Figure 3 The illustration shows a schematic diagram of a process for extracting features of the first page of paper according to an embodiment of this application. For example... Figure 3As shown, feature extraction is performed on the front view of the photo to obtain paper size, first page paper color, surface smoothness, and paper texture features. Feature extraction is performed on the front spectral image to obtain the first page reflectance. Then, a first fusion analysis is performed on the color, surface smoothness, paper texture features, and light reflectance to obtain the first page aging degree. A second fusion analysis is performed on the color, surface smoothness, paper texture features, and light reflectance to obtain the paper type.

[0058] For example, the paper size is calculated based on the location where the photo was taken, the distance to the paper document, and the number of paper pixels in the photo. For instance, the paper size can be obtained by using Canny edge detection combined with contour fitting.

[0059] For example, feature extraction from a front-facing photograph can utilize computer vision techniques to extract target features (such as paper color, surface smoothness, and paper texture features). For instance, computer vision models such as convolutional neural networks can be used. For example, the fusion analysis could be performed using a pre-trained paper type classifier and a neural network model of first-page aging.

[0060] For example, when performing feature extraction on a frontal spectral image, the light reflectance of each band can be treated as a separate image and a pre-trained spectral feature extraction network can be used to extract the frontal reflectance of each band; alternatively, multi-band data can be treated as a spectral data cube for feature extraction to obtain frontal reflectance reflecting the features of multiple bands.

[0061] Figure 4 The illustration shows a schematic diagram of the process for extracting page edge features according to an embodiment of this application.

[0062] like Figure 4 As shown, feature extraction is performed on the side photo to obtain paper thickness, edge straightness, and edge color. Feature extraction is performed on the side spectral image to obtain edge reflectance. Based on edge reflectance, paper thickness, edge straightness, and edge color, a third fusion analysis is performed to determine the degree of edge aging. Edge straightness indicates the uniformity of the edge and its degree of parallelism to the horizontal plane.

[0063] According to embodiments of this application, by extracting the features of visible paper from the visible portion of a paper document, the aging degree of the invisible portion of the paper document can be further inferred based on the features of the visible portion, so as to assess the paper condition of the stored paper document and provide timely warning when the paper ages to the point of being easily damaged.

[0064] According to an embodiment of this application, the aging degree is calculated based on the first page paper characteristics, page edge characteristics, and environmental parameters of the paper document to obtain a paper aging degree score. This includes: vertically propagating information about the aging degree of the first page based on environmental parameters and paper type to obtain a first vertical feature, which characterizes the degree of paper aging at the bottom of the paper document caused by pressure due to weight under temperature and humidity conditions; fusing the first vertical feature and the first horizontal feature to obtain a fused feature; wherein, the first horizontal feature is obtained by horizontally propagating information about the bottom page edge features in the page edge characteristics based on environmental parameters and paper size, and the first horizontal feature indicates the degree of aging at the bottom center of the paper document; and obtaining a paper aging degree score based on the score corresponding to the fused feature mapping.

[0065] For example, the top sheets of paper have better ventilation and are easier to pick up, but the bottom sheets may become brittle under the influence of gravity and are not easily damaged. Therefore, based on the paper characteristics of the first page of a stack of papers, and considering the effect of environmental parameters, including the gravity of the document itself, on the bottom sheets of the document, longitudinal aging transmission can be performed to derive the first longitudinal characteristic under gravity.

[0066] For example, the aforementioned gravity transmission can be simulated using a multi-layer neural network. For instance, if a paper document is divided into multiple layers from top to bottom according to a preset thickness or number of pages, such as five layers, then a five-layer neural network can be used to simulate the aging state of each of these five layers of paper. This model can be obtained through iterative training using sample data (including labeled ratings).

[0067] For example, based on the horizontal conduction effect of paper fibers, such as fiber capillary action, the first transverse feature of the paper center can be inferred from the characteristics of the page edge to determine the paper aging caused by the horizontal conduction effect. For instance, the first transverse feature can be predicted based on the page edge features using a multilayer perceptron based on the horizontal penetration effect.

[0068] For example, the first vertical feature and the first horizontal feature may include multiple features, such as features representing multiple locations or positions on the page. Feature fusion is performed through a fully connected layer, and the fused feature mapping score can be obtained by using an activation function, such as the softmax activation function, to map the paper aging score.

[0069] According to embodiments of this application, by considering factors such as the pressure caused by the weight of the paper and the fiber changes of the bottom paper, as well as the horizontal fiber conduction factors such as moisture and light intensity that degrade the paper, an aging score representing the degree of paper aging in the center of the bottom page of a paper document can be derived based on the paper characteristics of the first page. This allows for the prediction of whether the paper document is prone to breakage when taken out due to paper aging, thus providing an early warning.

[0070] According to embodiments of this application, feature extraction of photographs and spectral images to obtain ink blot features includes: performing optical character recognition on the photograph to obtain the recognized characters and the confidence scores corresponding to the recognized characters, and calculating the mean and standard deviation of the confidence scores; performing sharpness analysis on the character boundaries corresponding to the standard glyphs of the recognized characters in the photograph to obtain the ink blot boundary sharpness, which characterizes the clarity of the character boundaries in the file; extracting the light reflectance of the region corresponding to the recognized characters in the spectral image, calculating the average value of the difference between the light reflectance and a preset reference value to obtain the grayscale uniformity; and using the mean and standard deviation of the confidence scores, the ink blot boundary sharpness, and the grayscale uniformity as ink blot features.

[0071] Figure 5 The illustration shows a flowchart of extracting ink features from the homepage according to an embodiment of this application. Figure 5 As shown, the frontal photograph is segmented into multiple text / character regions based on the text content. Optical Character Recognition (OCR) is performed on each character region to obtain the recognized characters (including the confidence level of each recognized character). Then, the mean and standard deviation of the confidence levels are calculated, and the recognized characters are compared with their corresponding text or character regions to obtain the character boundary sharpness. For example, the character boundary sharpness can be the average character boundary sharpness of the text content. Based on the character regions obtained from the segmentation of the frontal photograph, the frontal image is further segmented into spectral regions corresponding to each recognized character region. Then, the standard deviation of the reflectance in each character region is calculated to determine the grayscale uniformity of the recognized character region.

[0072] For example, in multi-band data, carbon / laser toner has the lowest reflectance at 500nm, which facilitates accurate positioning of character ink pixels based on the contrast between ink and paper, avoiding OCR mask omissions / mislabeling; the 550nm band is the band most sensitive to the human eye, at which point the visual representation of paper whiteness and ink density is highly consistent with the spectral data, and the spectral reflectance data can be seamlessly integrated with ordinary visible light OCR results; the 600nm band is sensitive to ink smudging: ink diffusion in the smudging area will cause the reflectance of this band to rise slowly; the 650nm band can reflect the thickness or adhesion of printed ink layers or handwritten ink, and if the ink layer peels off, the reflectance of this band will rise sharply.

[0073] According to embodiments of this application, by extracting the ink features of the visible ink portion of a paper document, the degree of ink diffusion in the invisible portion of the paper document can be further inferred based on the ink features of the visible portion, so as to assess the clarity of the ink in the stored paper document and provide timely warnings when the ink becomes unclear.

[0074] According to an embodiment of this application, before performing optical character recognition on a photograph to obtain the recognized characters, handwritten characters or seals in the photograph are identified and determined; if handwritten characters or seals exist in the photograph, the target area where the handwritten characters or seals are located in the photograph is extracted, and the photograph is replaced with the target area to perform optical character recognition on the target area to obtain the recognized characters.

[0075] For example, paper documents may include printed parts and handwritten signature and stamp parts. Generally speaking, compared with printed fonts, handwritten signatures or stamp parts are more prone to problems such as ink diffusion along the paper fibers, resulting in illegible handwriting. In this case, the ink diffusion of handwritten characters or stamps in the photo can be predicted.

[0076] For example, carbon-based (carbon black) pens have extremely low reflectivity (close to absorption) in the 720–1000 nm range, while paper and red stamps have high reflectivity in this range, showing a significant difference. Red inkpads (cinnabar / ink) have strong absorption in the 500–580 nm range, creating a large contrast with paper and black ink. Therefore, a spectrometer can be specifically configured to collect spectral images in the 940 nm and 550 nm bands, and handwriting features can be extracted based on these images.

[0077] Figure 6 The illustration shows a flowchart of extracting ink features from the first page of a signature / stamp portion according to an embodiment of this application.

[0078] like Figure 6 As shown, the signature / stamp portion in the frontal photograph is identified. When the signature / stamp portion is present, it is segmented into multiple text / character regions. OCR is performed on each character region to obtain the identified character (including the confidence level of each identified character). The mean and standard deviation of the confidence levels are then calculated. The identified characters are compared with their corresponding text or character regions to determine the character boundary sharpness. Based on the signature / stamp portion identified in the frontal photograph, the signature / stamp portion in the spectral image is extracted. Then, the character regions obtained from the segmentation of the signature / stamp portion in the photograph are used to segment the signature / stamp portion in the spectral image, dividing the signature / stamp portion in the spectral image into spectral regions corresponding to each identified character region. The standard deviation of the reflectance in each character region is then calculated to determine the grayscale uniformity of the identified character region.

[0079] According to the embodiments of this application, by limiting the handwriting recognition portion to handwritten characters or seals, subsequent ink diffusion is carried out on characters that are more prone to handwriting diffusion problems. Compared with processing all the text content on the homepage, the data processing scale is reduced, computer computing resources are saved, and better recognition results can be achieved.

[0080] According to embodiments of this application, ink diffusion prediction based on first page paper characteristics, page edge characteristics, first page ink characteristics, and environmental parameters includes: predicting vertical diffusion of ink boundary sharpness and grayscale uniformity based on environmental parameters, paper type, paper thickness, mean and standard deviation of confidence level, to obtain vertical ink features, which characterize the degree of ink smudging between upper and lower pages of a paper document; predicting horizontal diffusion of ink boundary sharpness and grayscale uniformity based on environmental parameters, paper type, paper size, mean and standard deviation of confidence level, to obtain horizontal ink features, which characterize the degree of ink smudging in the horizontal direction of each page of paper, with paper size obtained by feature extraction from photographs; and performing feature fusion and scoring operations based on vertical and horizontal ink features to obtain an ink diffusion score.

[0081] For example, a pre-trained longitudinal prediction model performs vertical diffusion prediction on the ink blot edge sharpness and grayscale uniformity based on environmental parameters, paper type, paper thickness, mean and standard deviation of confidence level to obtain longitudinal ink blot features. The longitudinal ink blot features may include a first ink blot edge sharpness and a first grayscale uniformity. The first ink blot edge sharpness represents the inferred value of the ink blot edge sharpness and grayscale uniformity at the bottom of the paper document obtained by longitudinal propagation prediction based on the ink blot edge sharpness and grayscale uniformity of the first page.

[0082] For example, a pre-trained lateral prediction model can perform horizontal diffusion prediction of ink blot edge sharpness and grayscale uniformity based on environmental parameters, paper type, paper size, mean and standard deviation of confidence levels to obtain lateral ink blot features. These lateral ink blot features can include a second ink blot edge sharpness and a second grayscale uniformity. The second ink blot edge sharpness represents the inferred value of the ink blot edge sharpness and grayscale uniformity at the center of the paper document, obtained through vertical propagation prediction based on the ink blot edge sharpness and grayscale uniformity of the first page. Here, the center of the paper document refers to the portion of the paper document containing the text, i.e., the document portion excluding page margins, and is not limited to the center position of the document page.

[0083] For example, the pre-trained longitudinal prediction model / network and the lateral prediction model / network described above can be obtained by training a neural network model using sample datasets and labeled data.

[0084] For example, feature fusion can be performed using fully connected layers, and the scoring operation can be implemented using the Softmax activation function.

[0085] According to the embodiments of this application, by considering the effect of ink on the upper and lower pages of a paper document and the diffusion effect on the horizontal page based on the ink edge sharpness and grayscale uniformity of the first page, an ink diffusion score is obtained, representing the ink fusion of the center part of the bottom page of the paper document. This score is determined based on the ink characteristics of the first page combined with paper thickness, environmental factors, etc., to predict whether there is a problem of unclear text in the paper document due to ink diffusion, so as to provide early warning.

[0086] According to embodiments of this application, during pre-identified low system load periods, feature extraction is performed based on photographs and spectral images to obtain first page paper features, ink features, and page edge features.

[0087] For example, during the daytime, photos and spectral images of paper documents, along with environmental parameters and other data, are collected, transmitted via the Internet of Things, and stored. During the nighttime period when the system is under low load, feature extraction is performed based on the photos and spectral images to obtain the first page paper features, ink features, and page edge features, which are then used for subsequent analysis to obtain a fusion quality score.

[0088] For example, based on statistical analysis of the system load of banks and other institutions over the past month, low-load periods can be identified as pre-identified low-load periods. Here, "system" refers to the servers used by banks and other institutions to perform data processing, and server load can be measured by the activity level of the CPU (Central Processing Unit) or CPU utilization rate, etc.

[0089] According to the embodiments of this application, by reasonably allocating the time for data acquisition and data processing, the computing resources of the system can be effectively utilized, and the inspection can be avoided from causing a load impact on the business processing of institutions such as banks.

[0090] Based on the above-described paper document inspection method, embodiments of this application also provide a paper document inspection device. The following will be combined with... Figure 7 The device is described in detail.

[0091] Figure 7 A schematic block diagram of a paper document inspection device according to an embodiment of this application is shown.

[0092] like Figure 7 As shown, the paper document inspection device 700 of this embodiment includes a data acquisition module 710, a feature extraction module 720, an inference module 730, and an evaluation module 740.

[0093] The data acquisition module 710 is used to acquire photographs and spectral images of paper documents. The spectral images characterize the reflectivity of the paper documents to light in multiple wavelength bands. In one embodiment, the data acquisition module 710 can be used to perform step S210 described above, which will not be repeated here.

[0094] The feature extraction module 720 is used to extract features from the photograph and spectral image to obtain first-page paper features, first-page ink features, and page edge features. The first-page paper features indicate the paper type and paper aging degree of the top page of the paper document; the first-page ink features indicate the text clarity of the top page; and the page edge features indicate the paper thickness of the middle pages and the aging degree of the page edges. In one embodiment, the feature extraction module 720 can be used to perform step S220 described above, which will not be repeated here.

[0095] The inference module 730 is used to calculate the degree of aging based on the characteristics of the first page paper, the characteristics of the page edges, and the environmental parameters of the paper document, to obtain a paper aging score. It also predicts ink diffusion based on the characteristics of the first page paper, the characteristics of the page edges, the characteristics of the ink on the first page, and the environmental parameters, to obtain an ink diffusion score. The environmental parameters include at least one of temperature, humidity, document weight, and light intensity. In one embodiment, the inference module 730 can be used to perform step S230 as described above, which will not be repeated here.

[0096] The evaluation module 740 is used to perform a weighted fusion of the paper aging score and the ink diffusion score to obtain a fusion quality score. In one embodiment, the evaluation module 740 can be used to perform step S240 described above, which will not be repeated here.

[0097] According to an embodiment of this application, the feature extraction module 720 is further configured to: extract features from the photograph to obtain paper features; extract features from the spectral image to obtain light reflectance; the paper features include paper size, color of the first page, surface smoothness, and paper texture features; the paper size is calculated based on the photograph's acquisition location, the distance to the paper document, and the number of paper pixels in the photograph; perform fusion analysis on color, surface smoothness, paper texture features, and light reflectance to obtain paper type and first page aging degree; and use paper size, paper type, and first page aging degree as the first page paper features.

[0098] According to an embodiment of this application, the inference module 730 is further configured to: propagate information about the first page aging degree vertically based on environmental parameters and paper type to obtain a first vertical feature, the first vertical feature representing the degree of paper aging at the bottom of the paper document caused by pressure due to weight under temperature and humidity conditions; fuse the first vertical feature and the first horizontal feature to obtain a fused feature; wherein, the first horizontal feature is obtained by propagating information about the bottom edge feature in the page edge feature horizontally based on environmental parameters and paper size, the first horizontal feature indicating the degree of aging at the bottom center of the paper document; and obtain a paper aging degree score based on the score corresponding to the fused feature mapping.

[0099] According to an embodiment of this application, the feature extraction module 720 is further configured to: perform optical character recognition on the photograph to obtain the recognized character and the confidence level corresponding to the recognized character, and calculate the mean and standard deviation of the confidence level; perform sharpness analysis on the character boundary corresponding to the standard glyph of the recognized character in the photograph to obtain the ink boundary sharpness, wherein the ink boundary sharpness characterizes the clarity of the character boundary in the file; extract the light reflectance of the region corresponding to the recognized character in the spectral image, calculate the average value of the difference between the light reflectance and the preset reference value, and obtain the gray uniformity value; and use the mean, standard deviation, ink boundary sharpness, and gray uniformity value of the confidence level as ink features.

[0100] According to an embodiment of this application, the feature extraction module 720 is further configured to: identify and determine handwritten characters or seals in a photograph before performing optical character recognition based on the photograph; if handwritten characters or seals exist in the photograph, extract the target area where the handwritten characters or seals are located in the photograph, and replace the photograph with the target area to perform optical character recognition on the target area.

[0101] According to an embodiment of this application, the inference module 730 is further configured to: predict the vertical diffusion of ink edge sharpness and grayscale uniformity based on environmental parameters, paper type, paper thickness, mean and standard deviation of confidence level, to obtain vertical ink feature, which characterizes the degree of ink smudging between upper and lower pages of a paper document; predict the horizontal diffusion of ink edge sharpness and grayscale uniformity based on environmental parameters, paper type, paper size, mean and standard deviation of confidence level, to obtain horizontal ink feature, which characterizes the degree of ink smudging in the horizontal direction of each page of paper, and the paper size is obtained by feature extraction from a photograph; and perform feature fusion and scoring operations based on the vertical and horizontal ink feature to obtain an ink diffusion score.

[0102] According to an embodiment of this application, the feature extraction module 720 is further configured to: extract features based on photographs and spectral images during a pre-identified low system load period to obtain first page paper features, ink features, and page edge features.

[0103] According to embodiments of this application, any multiple modules among the data acquisition module 710, feature extraction module 720, inference module 730, and evaluation module 740 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the data acquisition module 710, feature extraction module 720, inference module 730, and evaluation module 740 can be at least partially implemented as hardware circuits, such as field-programmable gate arrays, programmable logic arrays, systems-on-a-chip, systems-on-a-substrate, systems-on-package, application-specific integrated circuits, or other reasonable means of integrating or packaging circuits, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the data acquisition module 710, feature extraction module 720, inference module 730, and evaluation module 740 can be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0104] Figure 8 A block diagram schematically illustrates an electronic device suitable for implementing a paper document inspection method according to an embodiment of this application.

[0105] like Figure 8 As shown, an electronic device 800 according to an embodiment of this application includes a processor 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory 802 or a program loaded from a storage portion 808 into a random access memory 803. The processor 801 may include, for example, a general-purpose microprocessor, an instruction set processor and / or an associated chipset and / or a dedicated microprocessor. The processor 801 may also include onboard memory for caching purposes. The processor 801 may include a single processing unit or multiple processing units for executing different steps of the method flow according to an embodiment of this application.

[0106] Random access memory 803 stores various programs and data required for the operation of electronic device 800. Processor 801, read-only memory 802, and random access memory 803 are interconnected via bus 804. Processor 801 executes various steps of the method flow according to embodiments of this application by executing programs stored in read-only memory 802 and / or random access memory 803. It should be noted that the programs may also be stored in one or more memories other than read-only memory 802 and random access memory 803. Processor 801 may also execute various steps of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0107] According to embodiments of this application, the electronic device 800 may further include an input / output interface 805, which is also connected to a bus 804. The electronic device 800 may also include one or more of the following components connected to the input / output interface 805: an input section 806 including a keyboard, mouse, etc.; an output section 807 including a cathode ray tube, liquid crystal display, etc., and a speaker, etc.; a storage section 808 including a hard disk, etc.; and a communication section 809 including a network interface card, such as a local area network card, modem, etc. The communication section 809 performs communication processing via a network such as the Internet. A drive 810 is also connected to the input / output interface 805 as needed. A removable medium 811, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 810 as needed so that computer programs read from it can be installed into the storage section 808 as needed.

[0108] Embodiments of this application also provide a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0109] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory, read-only memory, erasable programmable read-only memory, portable compact disk read-only memory, optical storage devices, magnetic storage devices, or any suitable combination thereof. In embodiments of this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include the read-only memory 802 described above, and / or random access memory 803, and / or one or more memories other than read-only memory 802 and random access memory 803.

[0110] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.

[0111] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 809, and / or installed from a removable medium 811. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0112] In embodiments of this application, the computer program can be downloaded and installed from a network via communication section 809, and / or installed from removable medium 811. When the computer program is executed by processor 801, it performs the functions defined in the system of embodiments of this application. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0113] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0114] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0115] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A method for inspecting paper documents, characterized in that, The method includes: Acquire photographs and spectral images of paper documents, wherein the spectral images characterize the reflectivity of the paper documents to light in multiple wavelength bands; Feature extraction is performed on the photograph and the spectral image to obtain the first page paper features, the first page ink features, and the page edge features. The first page paper features indicate the paper type and paper aging degree of the top page of the paper document, the first page ink features indicate the text clarity of the top page, and the page edge features indicate the paper thickness and aging degree of the middle pages of the paper document. The aging degree is calculated based on the first page paper characteristics, the page edge characteristics, and the environmental parameters of the paper document to obtain a paper aging degree score. The ink diffusion is predicted based on the first page paper characteristics, the page edge characteristics, the first page ink characteristics, and the environmental parameters to obtain an ink diffusion degree score. The environmental parameters include at least one of temperature, humidity, document weight, and light intensity. The paper aging score and ink diffusion score are weighted and fused to obtain a fusion quality score.

2. The method according to claim 1, characterized in that, The feature extraction from the photograph and the spectral image to obtain the features of the first page paper includes: Feature extraction is performed on the photograph to obtain paper features, and feature extraction is performed on the spectral image to obtain light reflectance. The paper features include paper size, color of the first page, surface smoothness, and paper texture features. The paper size is calculated based on the photograph's acquisition location, the distance to the paper document, and the paper pixels in the photograph. By performing a fusion analysis on the color, surface smoothness, paper texture features, and light reflectance, the paper type and first page aging degree can be obtained. The paper size, paper type, and first page aging degree are used as the characteristics of the first page paper.

3. The method according to claim 2, characterized in that, The process of calculating the degree of aging based on the characteristics of the first page paper, the characteristics of the page edges, and the environmental parameters of the paper document to obtain a paper aging score includes: Based on the environmental parameters and the paper type, the aging degree of the first page is propagated vertically to obtain a first longitudinal feature. The first longitudinal feature characterizes the degree of paper aging at the bottom of the paper document caused by the pressure formed by weight under the temperature and humidity conditions. The first vertical feature and the first horizontal feature are fused to obtain a fused feature; wherein, the first horizontal feature is obtained by horizontally propagating information from the bottom edge feature in the page edge feature according to the environmental parameters and the paper size, and the first horizontal feature indicates the degree of aging of the bottom paper center of the paper document. Based on the score corresponding to the fused feature map, the paper aging score is obtained.

4. The method according to claim 1, characterized in that, The feature extraction from the photograph and the spectral image to obtain the ink stain features includes: Optical character recognition is performed on the photograph to obtain the recognized characters and the confidence scores corresponding to the recognized characters, and the mean and standard deviation of the confidence scores are calculated. Sharpness analysis is performed on the character boundaries in the photo corresponding to the standard glyphs of the identified characters to obtain the ink boundary sharpness, which characterizes the clarity of the character boundaries in the file. Extract the light reflectance of the region corresponding to the identified character in the spectral image, calculate the average value of the difference between the light reflectance and the preset reference value, and obtain the grayscale uniformity value; The mean and standard deviation of the confidence level, the sharpness of the ink mark boundary, and the uniformity of gray level are used as ink mark features.

5. The method according to claim 4, characterized in that, Before performing optical character recognition on the photograph to obtain the recognized characters, the handwritten characters or seals in the photograph are identified and determined. If handwritten characters or seals are present in the photograph, the target area containing the handwritten characters or seals in the photograph is extracted, and the photograph is replaced with the target area to perform optical character recognition on the target area to obtain the recognized characters.

6. The method according to claim 4 or 5, characterized in that, The method of predicting ink diffusion based on the first page paper characteristics, the page edge characteristics, the first page ink characteristics, and the environmental parameters includes: Based on environmental parameters, paper type, paper thickness, mean and standard deviation of confidence level, vertical diffusion prediction is performed on the sharpness of ink boundary and gray uniformity to obtain vertical ink feature, which characterizes the degree of ink smudging between upper and lower pages of the paper document. Based on environmental parameters, paper type, paper size, mean and standard deviation of the confidence level, the horizontal diffusion prediction of the ink mark boundary sharpness and the gray uniformity value is performed to obtain the horizontal ink mark feature. The horizontal ink mark feature characterizes the degree of ink smudging in the horizontal direction of each page of paper. The paper size is obtained by feature extraction from the photograph. Based on the vertical ink feature and the horizontal ink feature, feature fusion and scoring operations are performed to obtain the ink diffusion score.

7. The method according to claim 1, characterized in that, During pre-identified periods of low system load, feature extraction is performed based on the photograph and the spectral image to obtain the first page paper features, ink features, and page edge features.

8. A paper document inspection device, characterized in that, The device includes: Data acquisition module: used to acquire photographs and spectral images of paper documents, wherein the spectral images characterize the reflectivity of the paper documents to light in multiple wavelength bands; Feature extraction module: used to extract features from the photo and the spectral image to obtain first page paper features, first page ink features and page edge features. The first page paper features indicate the paper type and paper aging degree of the top page of the paper document, the first page ink features indicate the text clarity of the top page, and the page edge features indicate the paper thickness and aging degree of the middle page of the paper document. Inference module: used to calculate the degree of aging based on the first page paper characteristics, the page edge characteristics, and the environmental parameters of the paper document to obtain a paper aging score; and to predict ink diffusion based on the first page paper characteristics, page edge characteristics, first page ink characteristics, and the environmental parameters to obtain an ink diffusion score. The environmental parameters include at least one of temperature, humidity, document weight, and light intensity. Evaluation module: used to perform weighted fusion of the paper aging score and ink diffusion score to obtain a fusion quality score.

9. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 7.

Images