CN115063804A

CN115063804A - Seal identification and anti-counterfeiting detection method and system

Info

Publication number: CN115063804A
Application number: CN202210747658.3A
Authority: CN
Inventors: 赵楠; 吕东东
Original assignee: Alipay Hangzhou Information Technology Co Ltd
Current assignee: Alipay Hangzhou Information Technology Co Ltd
Priority date: 2022-06-29
Filing date: 2022-06-29
Publication date: 2022-09-16

Abstract

The present disclosure provides a seal identification and anti-counterfeiting detection method and system. The method comprises the steps of extracting a seal image in a picture; correcting the text part in the extracted seal image; identifying the corrected text part to obtain a seal character; extracting corresponding seal anti-counterfeiting characteristics according to the seal image and the seal characters; and detecting the authenticity of the seal based on the extracted seal anti-counterfeiting feature.

Description

Seal identification and anti-counterfeiting detection method and system

Technical Field

The present disclosure relates to the field of artificial intelligence, and more particularly, to seal identification and anti-counterfeiting detection.

Background

At present, a great deal of manual investment is needed in the process of entering and checking a large number of collaborative documents, so that algorithm and system intervention is needed to realize automatic entering of documents, and the method becomes a key guarantee that a large number of collaborative tasks are not mistaken and are executed efficiently.

In the automatic document input process, the seal identification and the anti-counterfeiting detection are required to be carried out in parallel to ensure that the name of the mechanism to which the seal belongs is consistent with that of the initiating mechanism. In the process of identifying the seal, the traditional character identification effect is poor because, for example, the text of the name of the seal mechanism is distributed in the curved area, but the traditional character identification does not consider that the text of the name of the mechanism on the seal is stretched and distributed on the curved area, so the direction of the characters on the seal is not necessarily in the anchor orientation, the direction of the characters needs to be judged before identifying the text, in addition, after the characters on the seal rotate, the distribution of the text lines is changed into an arch shape or even a polygon shape from a square frame, and therefore the requirement of character identification of the seal can not be met by a square window for general target detection.

In addition, when the seal is subjected to anti-counterfeiting detection, manual identification or comparison is required, which causes poor anti-counterfeiting detection effect and requires a large amount of manpower.

Therefore, it is desirable to provide a seal identification and anti-counterfeiting detection method, so that the situation that characters are bent and stretched in an actual scene can be considered, a curved text region is corrected firstly, the characters are identified, the accuracy of seal character identification is improved, anti-counterfeiting detection of a seal is automatically performed by extracting anti-counterfeiting characteristics of the seal, the accuracy of seal anti-counterfeiting detection is greatly improved, and the efficiency of document entry can be further improved.

Disclosure of Invention

This disclosure is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This disclosure is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

In view of the above problems, according to a first aspect of the present disclosure, there is provided a method for identifying and detecting an anti-counterfeit seal, the method including: extracting a seal image in the picture; correcting the text part in the extracted seal image; identifying the corrected text part to obtain a seal character; extracting corresponding seal anti-counterfeiting characteristics according to the seal image and the seal characters; and detecting the authenticity of the seal based on the extracted seal anti-counterfeiting characteristic.

According to the technical scheme of the embodiment of the invention, the bending and stretching conditions of the text in the real scene are considered, the name of the mechanism to which the seal belongs is identified by utilizing target detection and scene character identification, the accuracy rate of seal identification is obviously improved, and the authenticity of the seal is automatically judged through anti-counterfeiting detection, so that the risk of an enterprise is released in the scene of a risk service.

According to one embodiment of the present disclosure, the stamp anti-counterfeiting feature includes one or more of a stamp font, a stamp color, a stamp specification, a stamp position, an anti-counterfeiting line, an anti-counterfeiting code, and a stamp character curvature.

According to a further embodiment of the present disclosure, when the extracted seal anti-counterfeiting feature is an anti-counterfeiting line, detecting authenticity of the seal based on the extracted seal anti-counterfeiting feature further comprises: acquiring a stamp corresponding to the real stamp from a server side based on the mechanism name in the stamp text; comparing the anti-counterfeiting patterns extracted from the stamp image with the anti-counterfeiting patterns in the stamp corresponding to the real stamp to determine the image similarity; and determining that the seal is true when the image similarity is greater than a preset threshold value.

According to a further embodiment of the present disclosure, when the extracted stamp anti-counterfeiting feature is an anti-counterfeiting code, detecting authenticity of the stamp based on the extracted stamp anti-counterfeiting feature further comprises: inquiring a corresponding organization name from a server side based on the extracted anti-counterfeiting code; and determining that the seal is true when the inquired mechanism name is consistent with the mechanism name in the seal characters.

According to a further embodiment of the present disclosure, extracting the stamp image in the picture further comprises: identifying whether the image contains the seal or not by utilizing a target detection network and determining the position information of the seal; and cutting out a corresponding stamp image in the picture based on the position information of the stamp.

According to a further embodiment of the present disclosure, rectifying the text portion in the extracted stamp image further comprises: identifying the detected stamp image to cut out a text image; sending the text image into a positioning network to position a group of reference points; calculating text image transformation parameters by using a grid generator according to the group of reference points and generating a sampling grid; the sampling grid and the text image are fed to a sampler to obtain a rectified text portion.

According to a further embodiment of the present disclosure, identifying the corrected text portion to obtain the stamp text further comprises: extracting sequence features from the rectified text portion using an encoder; and circularly generating an output sequence according to the sequence characteristics by using a decoder based on an attention mechanism to obtain corresponding seal characters.

According to a second aspect of the present disclosure, there is provided a stamp identification and anti-counterfeiting detection system, the system comprising: the seal extraction module is used for extracting a seal image in the picture; the character recognition module corrects the text part in the extracted stamp image; identifying the corrected text part to obtain a seal character; the anti-counterfeiting detection module extracts corresponding anti-counterfeiting characteristics of the seal according to the seal image and the seal characters; and detecting the authenticity of the seal based on the extracted seal anti-counterfeiting feature.

According to the technical scheme of the embodiment of the invention, by utilizing the stamp identification and anti-counterfeiting detection system, the curved text structure can be restored into a character square frame structure from top to bottom in the stamp identification process, the stamp characters are obtained through sequence identification, the accuracy of the identification result is greatly improved, anti-counterfeiting detection is automatically carried out by extracting anti-counterfeiting features (such as anti-counterfeiting codes, anti-counterfeiting lines and the like) from the stamp in the anti-counterfeiting detection process, manual operation is not needed, and the accuracy of the detection result is obviously improved.

According to a further embodiment of the present disclosure, when the extracted stamp security feature is an anti-counterfeiting code, detecting authenticity of the stamp based on the extracted stamp security feature further comprises: inquiring a corresponding organization name from a server side based on the extracted anti-counterfeiting code; and determining that the seal is true when the inquired mechanism name is consistent with the mechanism name in the seal characters.

According to a further embodiment of the present disclosure, rectifying the text portion in the extracted stamp image further comprises: identifying the detected stamp image to cut out a text image; sending the text image into a positioning network to position a group of reference points; calculating text image transformation parameters by using a grid generator according to the group of reference points and generating a sampling grid; the sampling grid and the text image are fed into a sampler to obtain a rectified text portion.

According to a third aspect of the present disclosure, there is provided a document automatic entry method, including: extracting page characters of a document to be input and a seal image in a picture; carrying out universal named entity recognition and customized named entity recognition on the extracted page characters; correcting the text part in the extracted seal image; identifying the corrected text part to obtain a seal character; extracting corresponding seal anti-counterfeiting characteristics according to the seal image and the seal characters; detecting authenticity of the seal based on the extracted seal anti-counterfeiting feature; and filling the character recognition result and the seal recognition result.

According to a fourth aspect of the present disclosure, there is provided a computer readable storage medium having stored thereon instructions that, when executed, cause a machine to perform the method of any of the preceding first aspects.

These and other features and advantages will become apparent upon reading the following detailed description and upon reference to the accompanying drawings. It is to be understood that both the foregoing general description and the following detailed description are explanatory only and are not restrictive of aspects as claimed.

Drawings

So that the manner in which the above recited features of the present disclosure can be understood in detail, a more particular description of the disclosure, briefly summarized above, may be had by reference to embodiments, some of which are illustrated in the appended drawings. It is to be noted, however, that the appended drawings illustrate only certain typical aspects of this disclosure and are therefore not to be considered limiting of its scope, for the description may admit to other equally effective aspects.

FIG. 1 is a schematic architecture diagram of a stamp identification and anti-counterfeiting detection system according to one embodiment of the present disclosure.

FIG. 2 is a schematic flow diagram of a stamp detection and identification method according to one embodiment of the present disclosure.

Fig. 3 is a schematic diagram of a structure for text rectification based on a spatial transformation network according to an embodiment of the present disclosure.

FIG. 4 is a schematic diagram of a structure for text recognition based on a sequence recognition network according to one embodiment of the present disclosure.

FIG. 5 is a schematic flow chart diagram of a method of stamp identification and anti-counterfeiting detection according to one embodiment of the present disclosure.

Fig. 6 is a schematic flow diagram of a document entry automation method according to one embodiment of the present disclosure.

FIG. 7 is a schematic architecture diagram of a stamp identification and anti-counterfeiting detection system according to one embodiment of the present disclosure.

Detailed Description

The present disclosure is described in detail below with reference to the attached drawing figures, and features of the present disclosure will become further apparent in the following detailed description.

FIG. 1 is a schematic architecture diagram of a stamp identification and anti-counterfeiting detection system 100 according to one embodiment of the present disclosure. System 100 may include at least a stamp extraction module 101, a text recognition module 102, and an anti-counterfeiting detection module 103.

The stamp extraction module 101 can extract a stamp image containing a stamp in the picture to be identified. In one embodiment, the stamp extraction module 101 may utilize the object detection network to identify whether a stamp is included in the picture and determine position information of the stamp, and cut out a corresponding stamp image in the picture based on the position information of the stamp.

The character recognition module 102 may correct a text portion in the extracted stamp image and recognize the corrected text portion to obtain a stamp character. In one embodiment, the word recognition module 102 may utilize a Space Transformation Network (STN) to rectify the text portion in the extracted stamp image. Specifically, the STN network may include a positioning network, a mesh generator, and a sampler, and the word recognition module 102 may recognize the extracted stamp image to crop out a text image, feed the text image into the positioning network to position a set of fiducial points, calculate text image transformation parameters with the mesh generator based on the set of fiducial points and generate a sampling mesh, and feed the sampling mesh and the text image into the sampler to obtain a rectified text portion. In one embodiment, the word recognition module 102 may utilize a sequence recognition network (STN) to recognize the rectified text portion to obtain the stamp word. In particular, the STN network may include an encoder and a decoder, and the word recognition module 102 may extract sequence features from the rectified text portion using the encoder and cyclically generate an output sequence from the sequence features using an attention-based decoder to obtain corresponding stamp words. The specific stamp detection and identification process will be described in more detail in FIGS. 2-4 below.

The anti-counterfeiting detection module 103 can extract corresponding anti-counterfeiting characteristics of the stamp according to the stamp image and the stamp characters, and detect the authenticity of the stamp based on the extracted anti-counterfeiting characteristics of the stamp. In one embodiment, the stamp security features may include one or more of stamp font, stamp color, stamp specification, stamp position, security print, security code, and stamp text curvature. In one embodiment, in the case that the extracted stamp anti-counterfeiting feature includes an anti-counterfeiting line, the anti-counterfeiting detection module 103 may obtain an impression corresponding to a real stamp from a server side based on the stamp mechanism name in the stamp text, compare the anti-counterfeiting line extracted from the stamp image with the anti-counterfeiting line in the impression corresponding to the real stamp to determine an image similarity, and determine that the stamp is true when the determined image similarity is greater than a preset threshold. In another embodiment, in a case that the extracted stamp anti-counterfeiting feature includes an anti-counterfeiting code, the anti-counterfeiting detection module 103 may query a corresponding organization name from a server side based on the extracted anti-counterfeiting code, and determine that the stamp is true when the queried organization name is consistent with the organization name in the stamp text. In another embodiment, the anti-counterfeit detection module 103 may determine whether the seal meets the specification based on the seal font, the seal color, the seal specification, or the seal position, so as to identify the authenticity of the seal.

Those skilled in the art will appreciate that the system of the present disclosure and its various modules may be implemented in either hardware or software, and that the modules may be combined or combined in any suitable manner.

FIG. 2 is a schematic flow chart diagram of a stamp detection and identification method 200 according to one embodiment of the present disclosure. As shown in fig. 2, in the process of detecting and identifying the stamp, firstly, the image to be identified is sent to the target detection network to cut out the stamp image containing the stamp. The target detection network may include, for example, R-CNN, SSD, YOLO, etc., where YOLO is an end-to-end target detection model, and the main idea is to convert the problem of target detection into a regression problem, and given an input image, the position of a bounding box of a target and its classification category may be directly regressed at multiple positions of the image, so that real-time target detection may be implemented at a relatively high speed.

The cut stamp image may then be sent to a text detection network to identify the extracted stamp image and cut a text image.

The cropped text image may then be fed into a Space Transformation Network (STN) to rectify the text portion, resulting in a rectified text portion. Turning to fig. 3, fig. 3 shows a schematic diagram of a structure 300 for text correction based on an STN network according to an embodiment of the present disclosure. The structure of the STN network comprises a positioning network, a grid generator and a sampler, and the network structure can learn displacement invariance and rotation invariance, namely, the network structure corrects an object subjected to affine transformation in a scene. As shown in FIG. 3, the cropped text image may be first croppedISending the data into a positioning network to position a set of reference pointsC. The positioning network may capture the overall text shape of the input image and locate the fiducial points accordingly, wherein no fiducial coordinate points are labeled for any sample in the training of the positioning network, as the training of the positioning network is fully supervised by other parts of the STN. The grid generator may then generate a grid of points based on the set of reference pointsCComputing text image transformation parameters and generating a sampling gridP. The sampling grid may then be meshedPAnd text imagesIFeeding into a samplerVTo obtain a rectified text portion. The sampler may perform bilinear interpolation on pixels of the input image near each pixel of the sampling grid, and the result of the interpolation is each pixel value of the corrected image. Thus, the sampler performs the above operation on all the pixels to obtain a corrected imageI’That is to say that,I’=V(P, I)。

returning to FIG. 2, the rectified text portion may be fed into a Sequence Recognition Network (SRN) to obtain stamp text. Turning to fig. 4, fig. 4 shows a schematic diagram of a structure 400 for text recognition based on a sequence recognition network according to an embodiment of the present disclosure. Since the target output is an inherent sequence of characters, the recognition problem can be modeled as a sequence recognition problem. The SRN network may include an encoder and a decoder, wherein the encoder may be composed of a 7-layer CNN that may obtain a feature map of an input image and convert the feature map into a sequence whose length is the width of the feature map, such that feature vectors arranged in a left-to-right order may be obtained, and a two-layer bi-directional LSTM that may analyze the independence of one sequence in two directions and output another sequence of the same length, the decoder being based on a sequence model of a unidirectional gated cyclic unit (GRU). Specifically, sequence features are extracted from the rectified text portion using an encoder in the SRN network, and an output sequence is cyclically generated from the sequence features using an attention-based decoder to obtain corresponding stamp words.

Therefore, in the process of seal detection and identification, the situation that characters are bent and stretched in an actual scene is considered, the STN and SRN technologies are combined, the STN is used for correcting the curved text structure to restore the curved text structure into a character bottom-up square frame structure, the SRN based on the attention mechanism is used, and the final seal characters are obtained through two independent circulation networks with direction sequences, so that the accuracy and the efficiency of seal identification are obviously improved.

FIG. 5 is a schematic flow chart diagram of a method 500 of stamp identification and anti-counterfeiting detection according to one embodiment of the present disclosure. The method 500 begins at step 501, where the stamp extraction module 101 may extract a stamp image from a picture. The extraction may be performed using an object detection network (e.g., YOLO). In one embodiment, the stamp extraction module 101 may utilize the object detection network to identify whether a stamp is included in the picture and determine position information (e.g., coordinate position) of the stamp, and cut out a corresponding stamp image in the picture based on the position information of the stamp.

In step 502, the character recognition module 102 may correct the text portion in the extracted stamp image. The above-mentioned rectification may be performed by using the space transformation network STN. In one embodiment, the word recognition module 102 may recognize the extracted stamp image to crop out a text image, feed the text image into a positioning network to locate a set of fiducial points, calculate text image transformation parameters with a grid generator based on the set of fiducial points and generate a sampling network, and then feed the sampling network and the text image into a sampler to obtain a rectified text portion.

In step 503, the text recognition module 102 may recognize the corrected text portion to obtain the stamp text. The above-mentioned character recognition may be performed by using a sequence recognition network SRN. In one embodiment, the word recognition module 102 may extract sequence features from the rectified text portion using an encoder in the SRN network and then cyclically generate an output sequence from the sequence features using an attention-based decoder to obtain corresponding stamp words.

In step 504, the anti-counterfeiting detection module 103 may extract corresponding stamp anti-counterfeiting features according to the extracted stamp image and the recognized stamp text. For example, the stamp security features may include, but are not limited to, security features for stamp security, security codes, stamp fonts, stamp colors, stamp specifications, stamp text curvature, and the like.

In step 505, the anti-counterfeiting detection module 103 may detect authenticity of the stamp based on the extracted stamp anti-counterfeiting feature. Consider the following several embodiments. In the first embodiment, the extracted anti-counterfeiting features of the seal include anti-counterfeiting lines, where the anti-counterfeiting lines refer to a plurality of thin lines distributed on the surface of the common seal, and the line positions of the lines are randomly generated and are very fine, and have uniqueness and non-replicability.

In the second embodiment, the extracted anti-counterfeit feature of the stamp includes an anti-counterfeit code, where the anti-counterfeit code refers to an anti-counterfeit code located at the lower edge of the genuine stamp, the font of the code is clear and smooth, and the fake official stamp has no code or has a forged code, in this case, the anti-counterfeit detection module 103 may query a corresponding mechanism name from the server side based on the extracted anti-counterfeit code, and determine that the stamp is genuine when the queried mechanism name is consistent with the mechanism name in the stamp text.

In a third embodiment, the extracted seal security features include seal font, seal color, seal specification, or seal position. Because the seal is usually engraved based on certain specifications, the authenticity of the seal can be identified by judging whether the seal font, the seal color, the seal specification or the seal position and the like meet the specifications.

In the fourth embodiment, the extracted stamp anti-counterfeiting feature includes stamp character curvature (i.e., curvature of the stamp curved text portion), in this case, the anti-counterfeiting detection module 103 may obtain the stamp corresponding to the real stamp from the server side based on the mechanism name in the recognized stamp character, and compare the stamp character curvature with the stamp character curvature in the stamp corresponding to the real stamp to identify the authenticity of the stamp.

Therefore, in consideration of the bending and stretching conditions of the real scene text, the stamp characters (including the name of the mechanism to which the stamp belongs) are identified by utilizing the target detection and the scene character identification, and the stamp anti-counterfeiting detection is automatically carried out by extracting the anti-counterfeiting characteristics of the stamp, so that the accuracy of the stamp identification and anti-counterfeiting detection results can be obviously improved.

Fig. 6 is a schematic flow diagram of a document entry automation method 600 according to one embodiment of the present disclosure. As shown in fig. 6, a document picture to be entered is first acquired. In one embodiment, in the process of entering the document, an electronic version document PDF can be generated or the document is photographed and uploaded to an online system, and the system uniformly converts PDF files into JPG format pictures.

All the text to be entered on the document picture can then be pre-processed by Optical Character Recognition (OCR) and all the page characters extracted, which may include, for example, de-skewing, de-speckling, and then binarizing the image (i.e., converting from color or grayscale to black and white), then recognizing the layout between the characters and normalizing the character results. After all page characters are extracted, a universal named entity identification (NER) can be performed on the whole text to identify a universal key named entity (e.g., company name, mailbox, identification number, etc.), and then a custom named entity identification can be performed on the basis to identify a business key named entity predefined by a business (e.g., a business field such as an originating organization name, calling personnel information, a query start time, a query deadline, etc.). The stamp recognition and anti-counterfeiting detection method described in fig. 5 may be performed concurrently with the OCR processing, and will not be described herein again.

And then, after OCR processing, seal recognition and anti-counterfeiting detection are carried out, automatically filling the obtained recognition result, and adding system interaction confirmed by a user to finish the automatic process of uploading the document to field entry by the user.

Therefore, by utilizing various computer vision algorithms such as OCR, target detection, character recognition and the like, the automatic document entry is realized, and the document entry efficiency is greatly improved.

FIG. 7 is a schematic architecture diagram of a stamp identification and anti-counterfeiting detection system 700 according to one embodiment of the present disclosure. As shown in fig. 7, system 700 may include a memory 701 and at least one processor 702. Memory 701 may include RAM, ROM, or a combination thereof. The memory 701 may store computer-executable instructions that, when executed by the at least one processor 702, cause the at least one processor 702 to perform various functions described herein, including: extracting a seal image in the picture; correcting the text part in the extracted seal image; identifying the corrected text part to obtain a seal character; extracting corresponding seal anti-counterfeiting characteristics according to the seal image and the seal characters; and detecting the authenticity of the seal based on the extracted seal anti-counterfeiting feature. In some cases, memory 701 may include, among other things, a BIOS that may control basic hardware or software operations, such as interaction with peripheral components or devices. The processor 702 may include intelligent hardware devices (e.g., general-purpose processors, DSPs, CPUs, microcontrollers, ASICs, FPGAs, programmable logic devices, discrete gate or transistor logic components, discrete hardware components, or any combinations thereof).

The various illustrative blocks and modules described in connection with the disclosure herein may be implemented or performed with a general purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices (e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, or any other such configuration).

The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Other examples and implementations are within the scope of the disclosure and the following claims. For example, due to the nature of software, the functions described herein may be implemented using software executed by a processor, hardware, firmware, hard-wired, or any combination thereof. Features that implement functions may also be physically located at various locations, including being distributed such that portions of functions are implemented at different physical locations.

What has been described above includes examples of aspects of the claimed subject matter. It is, of course, not possible to describe every conceivable combination of components or methodologies for purposes of describing the claimed subject matter, but one of ordinary skill in the art may recognize that many further combinations and permutations of the claimed subject matter are possible. Accordingly, the disclosed subject matter is intended to embrace all such alterations, modifications and variations that fall within the spirit and scope of the appended claims.

Claims

1. A stamp identification and anti-counterfeiting detection method comprises the following steps:

extracting a seal image in the picture;

correcting the text part in the extracted seal image;

identifying the corrected text part to obtain a seal character;

extracting corresponding seal anti-counterfeiting characteristics according to the seal image and the seal characters; and

and detecting the authenticity of the seal based on the extracted seal anti-counterfeiting feature.

2. The method according to claim 1, wherein the stamp security features include one or more of stamp font, stamp color, stamp format, stamp location, security print, security code, and stamp text curvature.

3. The method of claim 2, wherein detecting authenticity of the stamp based on the extracted stamp security feature when the extracted stamp security feature is a security mark further comprises:

acquiring a stamp corresponding to the real stamp from a server side based on the mechanism name in the stamp text;

comparing the anti-counterfeiting patterns extracted from the stamp image with the anti-counterfeiting patterns in the stamp corresponding to the real stamp to determine the image similarity; and

and determining that the seal is true when the image similarity is greater than a preset threshold value.

4. The method according to claim 2, wherein detecting authenticity of the stamp based on the extracted stamp security feature when the extracted stamp security feature is an anti-counterfeiting code further comprises:

inquiring a corresponding organization name from a server side based on the extracted anti-counterfeiting code; and

and determining the seal to be true when the inquired mechanism name is consistent with the mechanism name in the seal characters.

5. The method of claim 1, wherein extracting the stamp image from the picture further comprises:

identifying whether the image contains the seal or not by utilizing a target detection network and determining the position information of the seal; and

and cutting out a corresponding stamp image in the picture based on the position information of the stamp.

6. The method of claim 1, rectifying the text portion in the extracted stamp image further comprising:

identifying the detected stamp image to cut out a text image;

sending the text image into a positioning network to position a group of reference points;

calculating text image transformation parameters by using a grid generator according to the group of reference points and generating a sampling grid; and

the sampling grid and the text image are fed into a sampler to obtain a rectified text portion.

7. The method of claim 1, wherein recognizing the corrected text portion for stamp text further comprises:

extracting sequence features from the rectified text portion using an encoder; and

and circularly generating an output sequence according to the sequence characteristics by using a decoder based on an attention mechanism to obtain corresponding seal characters.

8. A stamp identification and anti-counterfeiting detection system, the system comprising:

the seal extraction module is used for extracting a seal image in the picture;

a character recognition module for recognizing the character of the character,

correcting the text part in the extracted seal image;

identifying the corrected text part to obtain a seal character; and

an anti-counterfeiting detection module is arranged on the base,

9. The system according to claim 8, wherein the stamp security features include one or more of stamp font, stamp color, stamp format, stamp location, security print, security code, and stamp text curvature.

10. The system of claim 9, wherein detecting authenticity of the stamp based on the extracted stamp security feature when the extracted stamp security feature is a security mark further comprises:

11. The system of claim 9, wherein when the extracted seal security feature is an anti-counterfeiting code, detecting authenticity of the seal based on the extracted seal security feature further comprises:

12. The system of claim 8, wherein extracting the stamp image from the picture further comprises:

13. The system of claim 8, rectifying the text portion in the extracted stamp image further comprising:

identifying the extracted stamp image to cut out a text image;

14. The system of claim 8, wherein recognizing the corrected text portion for stamp text further comprises:

15. A document automatic entry method comprises the following steps:

extracting page characters of a document to be input and a seal image in a picture;

carrying out universal named entity recognition and customized named entity recognition on the extracted page characters;

correcting the text part in the extracted seal image;

identifying the corrected text part to obtain a seal character;

extracting corresponding seal anti-counterfeiting characteristics according to the seal image and the seal characters;

detecting authenticity of the seal based on the extracted seal anti-counterfeiting feature; and

and filling a character recognition result and a seal recognition result.

16. A computer-readable storage medium having stored thereon instructions that, when executed, cause a machine to perform the method of any of claims 1-7.