An artificial intelligence-based seal deviation intelligent identification method

Abstract

The application discloses an artificial intelligence-based seal deviation intelligent identification method, relates to the technical field of seal identification, and solves the technical problem that the prior art cannot completely avoid the mistakes in the work process when performing seal work, and once the mistakes occur, the seal abnormality is caused, the seal effectiveness is affected, and timely remedy cannot be performed; the signature area and the seal image in the file are identified after file correction, and then it is judged whether the seal image deviates; after the file image is identified, the file correction is performed to ensure the quality, and then whether the deviation occurs is judged according to the overlapping area between the signature area and the seal image, the application range is widened on the basis of ensuring the identification efficiency; in the process of file correction, the standard content in the file image is reasonably determined based on the standard gray scale, and then the established file framework is filled in a radiation mode; the file correction is completed with a small data processing amount, and the data processing efficiency is effectively improved.

Description

Technical Field

[0001] This invention belongs to the field of seal recognition and relates to seal position recognition and verification technology based on artificial intelligence, specifically an intelligent recognition method for seal deviation based on artificial intelligence. Background Technology

[0002] Currently, most engineering drawings or contracts are manually signed and stamped by staff after review, and then kept as evidence of their validity. However, staff may need to sign multiple engineering drawings or contracts in a short period of time, which inevitably leads to omissions or misplacement of signatures, affecting execution efficiency.

[0003] Most existing technologies employ electronic seals. For example, patent application CN114399278A discloses a method for intelligent positioning and batch signing of engineering drawings. This method exports engineering drawings to PDF documents and establishes a storage path; it automatically identifies the seal position and coordinates based on the storage path, and then performs intelligent batch stamping based on user authorization after selecting the seal. However, for small-batch signing tasks, manual processing remains the primary method. Neither electronic nor manual seals can completely eliminate errors. Errors can lead to abnormal seals, affecting their validity and making timely remediation impossible. Therefore, there is an urgent need for an AI-based intelligent seal misalignment recognition method. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an intelligent identification method for stamp deviation based on artificial intelligence, which is used to solve the technical problem that the prior art cannot completely avoid errors in the process of signing, and once an error occurs, it will cause abnormal signing, which will affect the validity of the signing and cannot be remedied in time.

[0005] To achieve the above objectives, a first aspect of the present invention provides an artificial intelligence-based intelligent method for recognizing stamp misalignment, comprising:

[0006] The intelligent recognition module batches and collects file images to identify their standard dimensions. Based on these standard dimensions, the file images are then corrected. The file correction process begins with several standard elements.

[0007] After document correction, the system identifies the signature area and seal image in the document image; it determines whether the seal image is offset relative to the signature area; if yes, an alert is issued; otherwise, the seal is deemed valid.

[0008] Preferably, the intelligent recognition module includes a central analysis unit and an image acquisition unit, wherein:

[0009] The image acquisition unit captures images of the documents to be identified and sends them to the central analysis unit in an orderly manner.

[0010] The central analysis unit performs document correction on the received document images and determines the signature offset.

[0011] Preferably, the intelligent recognition module acquires the file image and identifies the standard size of the file image, including:

[0012] Set a ruler in the shooting area and collect file images of various types of files in the shooting area;

[0013] Identify the size of the corresponding file based on the ruler in the file image and match it to a standard size; the standard size includes A1, A2, A3, A4 or a custom size.

[0014] Preferably, the intelligent recognition module performs document correction on the document image based on a standard size, including:

[0015] A file frame is created based on standard dimensions, and several standard contents are identified from the file image.

[0016] The document attributes are determined based on several standard contents, and the document is identified from the starting point of these standard contents outwards. The document frame is then filled based on the identification results. The document attributes include line spacing, font, and page margins.

[0017] Preferably, the intelligent recognition module recognizes several standard contents in the file image, including:

[0018] Acquire the acquisition information of the file image; the acquisition information includes acquisition height and acquisition brightness;

[0019] Standard grayscale values ​​are obtained by matching the acquisition height and brightness. The grayscale values ​​of the file image are compared with the standard grayscale values ​​to determine several standard contents.

[0020] Preferably, the intelligent recognition module establishes a correlation between standard grayscale and acquisition height and acquisition brightness, including:

[0021] Set a standard array; the standard array includes combinations of different acquisition heights and different acquisition brightness.

[0022] The intelligent recognition module is adjusted according to each standard array, and various types of files are captured by the adjusted intelligent recognition module. The standard grayscale of the corresponding file image is determined by combining the expert scoring method.

[0023] Establish the association between standard grayscale values ​​and their corresponding standard arrays, and store them in the intelligent recognition module.

[0024] Preferably, after the file correction is completed, the intelligent recognition module identifies the signature area of ​​the file image, including:

[0025] Extract a preset region from the file image; where the preset region is the regular signature region of the corresponding file;

[0026] Determine if a text signature exists in the preset area; if yes, define the signature area around the text signature; otherwise, issue a warning.

[0027] Preferably, the intelligent recognition module determines whether the signature image has shifted based on the signature area, including:

[0028] Extract whether a signature image exists; if yes, proceed to the next step; otherwise, issue a warning.

[0029] The overlap area between the signature image and the signature area is analyzed and calculated, and marked as CDM. When CDM ≥ CDY, it is determined that the signature image has no offset; otherwise, it is determined that the signature image has been offset. Here, CDY is the overlap threshold.

[0030] Compared with the prior art, the beneficial effects of the present invention are:

[0031] 1. This invention automatically acquires signed document images through an image acquisition unit and performs document correction based on standard dimensions. After document correction, it identifies the signature area and the seal image, and then determines whether the seal image has shifted. This invention performs document correction after identifying the document image to ensure quality, and then determines whether shift has occurred based on the overlap area between the signature area and the seal image, thus expanding the application scope while ensuring recognition efficiency.

[0032] 2. In the process of document correction, this invention extracts the corresponding standard grayscale based on the acquisition conditions of the document image, and reasonably determines the standard content in the document image based on the standard grayscale, and then fills the established document frame in a radial manner to achieve document correction; this invention completes document correction with a small amount of data processing, effectively improving data processing efficiency. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0034] Figure 1 This is a schematic diagram illustrating the working steps of the present invention. Detailed Implementation

[0035] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Please see Figure 1 The first aspect of the present invention provides an intelligent identification method for stamp deviation based on artificial intelligence, comprising: batch acquiring document images through an intelligent identification module and identifying the standard size of the document images; performing document correction on the document images based on the standard size; identifying the signature area and the stamp image in the document images after document correction; determining whether the stamp image is offset relative to the signature area; if yes, issuing a warning; if no, the stamp is qualified.

[0037] Current technologies typically employ automation to embed the seal during the signing process, allowing the image to be inserted into a pre-defined location according to a pre-defined program. However, this method requires developing a separate program for each type of document, embedding the image based on document type identification, which demands significant preparation work and cannot guarantee the stability of the automation technology, inevitably leading to omissions or errors. Furthermore, automated signing has limited applicability, with manual signing still being the most common method. Therefore, identifying signing anomalies is a crucial issue.

[0038] This invention automates the acquisition of signed document images using an image acquisition unit and performs document correction based on standard dimensions. After document correction, it identifies the signature area and the seal image, thereby determining whether the seal image has shifted. This invention ensures quality by performing document correction after image recognition and then determining whether shift has occurred based on the overlap area between the signature area and the seal image, thus broadening the application range while maintaining recognition efficiency.

[0039] The intelligent recognition module in this invention includes a central analysis unit and an image acquisition unit. The image acquisition unit captures images of the documents to be recognized and sends them to the central analysis unit in an orderly manner. The central analysis unit performs document correction and signature offset judgment on the received document images.

[0040] The image acquisition unit primarily consists of a height-adjustable camera that continuously captures images of the document within the designated area. When necessary, a background light source needs to be set in conjunction with the image acquisition unit to ensure image quality. After acquiring the document image, pre-detection can be performed to determine if a standard region exists. If it does, the process continues; otherwise, the camera height or background light source should be adjusted to re-acquire the document image. The central analysis unit mainly performs document correction, matching, and other processing, and communicates and / or is electrically connected to the image acquisition unit. It's important to understand that standard content refers to areas or content in the document image whose grayscale values ​​meet the requirements.

[0041] In a preferred embodiment, the intelligent recognition module acquires a file image and identifies the standard size of the file image, including: setting a ruler in the shooting area, acquiring file images of various types of files in the shooting area; identifying the size of the corresponding file based on the ruler in the file image, and matching the standard size.

[0042] A ruler is set within the image acquisition unit's capture area for document images. When a document is placed on the ruler, it is captured along with the ruler in the document image, thus obtaining the document's standard size. In this embodiment, standard sizes include A1, A2, A3, A4, or a custom size. This means that some contract documents will be recognized as A4 standard size, while engineering drawings may be A1 or A2. Similarly, staff can customize the size for less frequently used documents. It should be noted that this invention can also be used for electronic documents; in this case, the remaining document image acquisition process can be directly sent to the central analysis unit for further processing.

[0043] In a preferred embodiment, the intelligent recognition module performs file correction on the file image based on the standard size, including: establishing a file frame according to the standard size, and recognizing several standard contents from the file image; determining file attributes based on the several standard contents, and recognizing from the several standard contents outwards, and filling the file frame according to the recognition results.

[0044] To ensure document image quality, after determining the standard size of the document corresponding to the image, the image is corrected based on this standard size. Specifically, standard content within the image is first identified and extracted. Then, document attributes are identified and organized based on this standard content. Finally, the entire document frame is filled using this standard content as the starting point. It's important to note that document attributes include line spacing, font, and margins.

[0045] Once the standard size is known, a document framework can be established. Combining this with document attributes reveals the document's layout style. Then, image recognition technology identifies text, symbols, and other elements, which are then reassembled and filled into the document framework, completing the document correction. For example, if the standard size is A4, the document size is 210mm × 297mm. Combining this with page margins reveals the content distribution area, and combining this with line spacing and font determines the document layout.

[0046] In an optional embodiment, the intelligent recognition module identifies several standard contents in the file image, including: acquiring the acquisition information of the file image; acquiring standard grayscale values ​​based on the acquisition height and acquisition brightness matching; comparing the grayscale values ​​of the file image with the standard grayscale values; and determining several standard contents.

[0047] The most important aspect of document correction is the standard content. This embodiment uses grayscale values ​​for determination. After obtaining the standard grayscale value by combining the acquisition height and brightness, the area in the document image whose grayscale value is closest to the standard grayscale value is identified as the standard content. For example, if the standard grayscale value is 100, then values ​​close to it, such as 98 and 102, can be identified as standard content. It's important to understand that when determining the grayscale value of a document image, the smallest unit is a single font or symbol. If the grayscale value of a font meets the requirements, it is included in the standard content until the grayscale values ​​of adjacent fonts (adjacent includes left-right and top-bottom adjacent) no longer meet the requirements. It is worth noting that the amount and distribution of standard content will affect the document correction effect. Therefore, when determining standard content, it should not be excessive; only enough to identify and determine the document attributes is needed.

[0048] In an optional embodiment, the intelligent recognition module establishes a correlation between standard grayscale and acquisition height and acquisition brightness, including: setting a standard array; adjusting the intelligent recognition module according to each standard array, and capturing various types of files through the adjusted intelligent recognition module, and determining the standard grayscale of the corresponding file image by combining expert scoring; establishing a correlation between the standard grayscale and the corresponding standard array, and storing it in the intelligent recognition module.

[0049] Shooting experiments were conducted using different combinations of capture height and brightness. Expert scoring was used to determine the optimal grayscale values ​​for each file type under various standard arrays, and the correlation between these values ​​was established. Therefore, during file correction, the corresponding standard grayscale values ​​can be directly extracted based on the information obtained during shooting.

[0050] In a preferred embodiment, after the file correction is completed, the intelligent recognition module identifies the signature area of ​​the file image, including: extracting a preset area of ​​the file image; determining whether there is a text signature in the preset area; if yes, then delineating the signature area around the text signature; if no, then issuing a warning.

[0051] The preset area is the standard signature area for the corresponding document, such as the top of the first page and the bottom of the last page of a contract. It identifies whether a signature, such as the name of the person in charge or the company name, exists in the preset area. If it does not exist, a warning should be issued; if it exists, the area containing the signature should be marked as the signature area. It's important to understand that the signature area should not be too large; it should just be large enough to contain the signature.

[0052] In an optional embodiment, the intelligent recognition module determines whether the signature image has shifted based on the signature area, including: extracting whether a signature image exists; if yes, proceed to the next step; if no, issue an early warning; analyze and calculate the overlap area between the signature image and the signature area, and mark it as CDM; when CDM≥CDY, it is determined that the signature image has not shifted, otherwise it is determined that the signature image has shifted.

[0053] After determining the signature area, the stamp image, i.e., the stamp position, is identified. Then, the degree of overlap between the stamp image and the signature area is used to determine whether the stamp image has shifted. Understandably, the signature area can also be a pre-defined area by the staff.

[0054] The data in the above formula are all calculated by removing the dimensions and taking the numerical values. The formula is the closest to the real situation obtained by software simulation of a large amount of collected data. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained through simulation of a large amount of data.

[0055] Working principle of the invention:

[0056] The intelligent recognition module batches and collects file images, identifies the standard size of the file images, and then performs file correction based on the standard size.

[0057] After document correction, the system identifies the signature area and seal image in the document image; it determines whether the seal image is offset relative to the signature area; if yes, an alert is issued; otherwise, the seal is deemed valid.

[0058] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for intelligent recognition of stamp deviation based on artificial intelligence, characterized in that, include: The intelligent recognition module can batch collect file images and identify the standard size of the file images. File correction is performed on the file image based on standard dimensions; the file correction begins with several standard elements, including: A file frame is created based on standard dimensions, and several standard contents are identified from the file image. The document attributes are determined based on several standard contents, and the document is identified from the starting point of these standard contents outwards. The document frame is then filled based on the identification results. The document attributes include line spacing, font, and page margins. After document correction, the signature area and seal image in the document image are identified; it is determined whether the seal image is offset relative to the signature area; if yes, an alert is issued; otherwise, the seal is deemed valid; among these, The intelligent recognition module also includes establishing the correlation between standard grayscale and acquisition height and acquisition brightness in the file image, including: Set a standard array; the standard array includes combinations of different acquisition heights and different acquisition brightness. The intelligent recognition module is adjusted according to each standard array, and various types of files are captured by the adjusted intelligent recognition module. The standard grayscale of the corresponding file image is determined by combining the expert scoring method. Establish the association between standard grayscale values ​​and their corresponding standard arrays, and store them in the intelligent recognition module; After the file correction is completed, the intelligent recognition module identifies the signature area of ​​the file image, including: Extract a preset region from the file image; where the preset region is the regular signature region of the corresponding file; Determine if a text signature exists in the preset area; if yes, define the signature area around the text signature; otherwise, issue a warning.

2. The method for intelligent recognition of stamp deviation based on artificial intelligence according to claim 1, characterized in that, The intelligent recognition module includes a central analysis unit and an image acquisition unit, wherein: The image acquisition unit captures images of the documents to be identified and sends them to the central analysis unit in an orderly manner. The central analysis unit performs document correction on the received document images and determines the signature offset.

3. The method for intelligent recognition of stamp deviation based on artificial intelligence according to claim 2, characterized in that, The intelligent recognition module acquires the file image and identifies the standard size of the file image, including: Set a ruler in the shooting area and collect file images of various types of files in the shooting area; Identify the size of the corresponding file based on the ruler in the file image and match it to a standard size; the standard size includes A1, A2, A3, A4 or a custom size.

4. The method for intelligent recognition of stamp deviation based on artificial intelligence according to claim 1, characterized in that, The intelligent recognition module identifies several standard contents in the file image, including: Acquire the acquisition information of the file image; the acquisition information includes acquisition height and acquisition brightness; Standard grayscale values ​​are obtained by matching the acquisition height and brightness. The grayscale values ​​of the file image are compared with the standard grayscale values ​​to determine several standard contents.

5. The method for intelligent recognition of stamp deviation based on artificial intelligence according to claim 1, characterized in that, The intelligent recognition module determines whether the signature image has shifted based on the signature area, including: Extract whether a signature image exists; if yes, proceed to the next step; otherwise, issue a warning. The overlap area between the signature image and the signature area is analyzed and calculated, and marked as CDM. When CDM ≥ CDY, it is determined that the signature image has no offset; otherwise, it is determined that the signature image has been offset. Here, CDY is the overlap threshold.

Images