Signature forgery detection device, method, computing device and storage medium

Abstract

The present invention discloses a signature forgery detection device, method, computing equipment and storage medium. The device includes: an interaction layer, an interface layer and a service layer, wherein the service layer includes: a receiving module, adapted to receive a signature image to be detected and its text mark dispatched from the interface layer; a first segmentation module, adapted to perform single-word segmentation processing on the signature image to be detected to obtain each text image to be detected; a first feature extraction module, adapted to extract the hash code of each text image to be detected as a text image feature; a comparison module, adapted to perform a query in a text feature library according to the text mark, compare the text image features of each text image to be detected with each text image feature of the query, and if there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, then the signature image to be detected is determined to be a forged signature image. The device can improve the precision and accuracy of detecting forged signature images.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a signature forgery detection device, method, computing equipment and storage medium. Background Art

[0002] With the rapid development of social information technology, handwriting input has become a natural and convenient form of human-computer interaction. In the digital age, handwritten signatures on smart devices (mobile phones, tablets) have been used in more and more agreement signing scenarios, such as: payment voucher signatures, bank business confirmation signatures, paperless business work order signatures, etc. Users sign their names on the handwritten signature device to confirm the business information. While safeguarding the legitimate rights and interests of both parties A and B, it improves the efficiency of business handling and brings convenience to people's production and life. However, when the client collects the user's signature, it lacks data compliance and security verification and is easily tampered with by the outside. In the existing production, it is frequently found that the signatures of different customers are completely consistent, that is, the signature image is a synthesis of single-word forged images, which brings huge security risks to the signature system. In order to improve the security of the signature system, signature anti-counterfeiting technology has become a key research direction.

[0003] Existing signature anti-counterfeiting related technologies mainly include two types: handwritten text recognition and page covering invisible signature document anti-counterfeiting method. Handwritten text recognition mainly uses two technologies, deep learning or character handwriting segmentation, to identify the handwritten text content for subsequent text verification. Deep learning handwriting recognition method: Under the premise of having a large amount of handwritten text data and labels, use deep neural network training to enable the model to learn the characteristic pattern of each character, so as to achieve the purpose of more accurate recognition of handwritten fonts. Character handwriting segmentation method: obtain character handwriting; obtain a segmentation model corresponding to the language category to which the character handwriting belongs; use the segmentation model to segment the character handwriting to obtain a segmentation block sequence; identify the segmentation block sequence to obtain the character string corresponding to the character handwriting.

[0004] The page-covered invisible signature document anti-counterfeiting method uses the signature of the certificate holder to form a signature image, and uses the same color but different spectrum color matching ink to print the signature image on an area of ​​the certificate page that is the same size and shape as the signature image. The signature image on this area is printed on top of the original text and image on the certificate page, generating a signature image that is hidden on the certificate page. The original text and image printed on the certificate page are displayed under visible light, and the signature image printed on the certificate page is displayed under infrared light, thereby effectively identifying whether the certificate is counterfeit.

[0005] However, the inventors discovered during the process of implementing the present invention that although the handwritten text recognition method can recognize text content, it cannot adaptively update the user's handwritten text in a targeted manner, and cannot effectively monitor whether there is a text template that is almost identical to the handwritten text features, and thus has no basis for judging whether the signature is forged; and the page-covered invisible signature certificate anti-counterfeiting requires printing on the signature medium such as the certificate using color-matching ink, which relies on printing technology and paper media, and is not suitable for the paperless signature scenario of the operator industry. Summary of the invention

[0006] In view of the above problems, the present invention is proposed to provide a signature forgery detection apparatus, method, computing device and storage medium that overcome the above problems or at least partially solve the above problems.

[0007] According to one aspect of the present invention, a signature forgery detection device is provided, the device comprising an interaction layer, an interface layer and a service layer, wherein the service layer comprises:

[0008] A receiving module, adapted to receive the signature image to be detected and its text mark scheduled by the interface layer;

[0009] The first segmentation module is adapted to perform single-word segmentation processing on the signature image to be detected to obtain each character image to be detected;

[0010] A first feature extraction module, adapted to extract the hash code of each to-be-detected text image as a text image feature;

[0011] The comparison module is suitable for querying in the text feature library according to the text mark, comparing the text image features of each text image to be detected with each text image feature of the query. If there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, it is determined that the signature image to be detected is a forged signature image.

[0012] Optionally, the device further comprises:

[0013] A second segmentation module is adapted to segment the acquired historical forged signature image to obtain at least one text image;

[0014] A second feature extraction module, adapted to extract a hash code of at least one text image as a text image feature;

[0015] The storage module is adapted to store the text image features of at least one text image in a text feature library; wherein different text image features corresponding to the same text are stored in a feature table corresponding to the text.

[0016] Optionally, the first feature extraction module is further adapted to:

[0017] For any text image to be detected, the difference between the row pixels of the text image to be detected is calculated using the difference hash algorithm to obtain a difference matrix, and the average value of all elements in the difference matrix is ​​calculated; each element in the difference matrix is ​​compared with the average value, and a feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image to be detected;

[0018] The second feature extraction module is further suitable for: for any text image, using the difference hash algorithm to calculate the difference between the row pixels of the text image, obtaining a difference matrix, and calculating the average value of all elements in the difference matrix; comparing each element in the difference matrix with the average value, determining the feature string based on the comparison result, converting the feature string into hexadecimal, and obtaining the hash code of the text image.

[0019] Optionally, the first segmentation module is further adapted to: perform binarization processing on the signature image to be detected, perform connected domain analysis on the binarized signature image to be detected, and perform segmentation processing on different connected domains to obtain various text images to be detected;

[0020] The second segmentation module is further adapted to: perform binarization processing on the historical forged signature image, perform connected domain analysis on the binarized historical forged signature image, and perform segmentation processing on different connected domains to obtain various text images.

[0021] Optionally, the device further comprises:

[0022] The filtering module is adapted to determine whether the number of different connected domains of the signature image to be detected is consistent with the number of text marks, and if not, filter the signature image to be detected.

[0023] Optionally, the device further comprises:

[0024] The adaptive adding module is adapted to store the text image features of other text images to be detected, except for the text image features of the text image to be detected which are similar to the queried text image features, into the text feature library.

[0025] Optionally, the device further comprises:

[0026] an adaptive deletion module, adapted to sort each text image feature according to the frequency with which each text image feature in the text feature library is detected to be similar to the text image feature of the text image to be detected; and,

[0027] When the number of text image features detected in the text feature library exceeds a preset value, the text image features detected to be similar to the text image features of the text image to be detected and whose frequency is lower than the preset value are deleted.

[0028] According to another aspect of the present invention, a method for detecting signature forgery is provided, the method being applied to a service layer, comprising:

[0029] Receive the signature image to be detected and its text mark scheduled by the interface layer;

[0030] Perform single-word segmentation processing on the signature image to be detected to obtain each text image to be detected;

[0031] Extract the hash code of each text image to be detected as the text image feature;

[0032] A query is performed in the text feature library based on the text mark, and the text image features of each text image to be detected are compared with each text image feature of the query. If there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, it is determined that the signature image to be detected is a forged signature image.

[0033] Optionally, the method further comprises:

[0034] Segment the acquired historical forged signature image to obtain at least one text image;

[0035] Extracting a hash code of at least one text image as a text image feature;

[0036] The text image features of at least one text image are stored in a text feature library; wherein different text image features corresponding to the same text are stored in a feature table corresponding to the text.

[0037] Optionally, extracting the hash code of each to-be-detected text image as a text image feature further includes:

[0038] For any text image to be detected, the difference between the row pixels of the text image to be detected is calculated using the difference hash algorithm to obtain a difference matrix, and the average value of all elements in the difference matrix is ​​calculated; each element in the difference matrix is ​​compared with the average value, and a feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image to be detected;

[0039] Extracting a hash code of at least one text image as a text image feature further includes:

[0040] For any text image, the difference hash algorithm is used to calculate the difference between the row pixels of the text image, obtain the difference matrix, and calculate the average value of all elements in the difference matrix; each element in the difference matrix is ​​compared with the average value, and the feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image.

[0041] Optionally, performing single-word segmentation processing on the signature image to be detected to obtain each text image to be detected further includes: performing binarization processing on the signature image to be detected, performing connected domain analysis on the binarized signature image to be detected, and segmenting different connected domains to obtain each text image to be detected;

[0042] Segmenting the acquired historical forged signature image to obtain at least one text image further includes: binarizing the historical forged signature image, performing connected domain analysis on the binarized historical forged signature image, and segmenting different connected domains to obtain various text images.

[0043] Optionally, the method further includes: determining whether the number of different connected domains of the signature image to be detected is consistent with the number of text marks; if not, filtering the signature image to be detected.

[0044] Optionally, the method further comprises: storing the text image features of other text images to be detected, except for the text images to be detected whose text image features are similar to the queried text image features, in a text feature library.

[0045] Optionally, the method further includes: sorting each text image feature in the text feature library according to the frequency with which each text image feature in the text feature library is detected to be similar to the text image feature of the text image to be detected; and when the number of text image features detected in the text feature library exceeds a preset value, deleting the text image features whose frequency of being detected to be similar to the text image features of the text image to be detected is lower than the preset value.

[0046] According to another aspect of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0047] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned signature forgery detection method.

[0048] According to another aspect of the present invention, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned signature forgery detection method.

[0049] According to the detection device, method, computing device and storage medium of signature forgery of the present invention, the device comprises: an interaction layer, an interface layer and a service layer, wherein the service layer comprises: a receiving module, adapted to receive the signature image to be detected and its text mark dispatched from the interface layer; a first segmentation module, adapted to perform single-word segmentation processing on the signature image to be detected to obtain each text image to be detected; a first feature extraction module, adapted to extract the hash code of each text image to be detected as a text image feature; a comparison module, adapted to perform a query in a text feature library according to the text mark, and compare the text image features of each text image to be detected with the text image features of each query, and if there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, then the signature image to be detected is determined to be a forged signature image. Through the above-mentioned method, the problem that the existing signature forgery method fails to solve the problem that there is no fixed template for electronic signatures and forged texts in actual business scenarios is solved, and the single-word image to be detected and the single-word image in the feature library are directly compared by feature differences without comparing between images, which can improve the processing speed, can improve the precision and accuracy of detecting synthetic forged signature images, and has portability and extensibility.

[0050] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0052] Figure 1 A schematic diagram of the structure of a signature forgery detection device provided by an embodiment of the present invention is shown;

[0053] Figure 2 A schematic diagram showing the structure of a signature forgery detection device provided by another embodiment of the present invention;

[0054] Figure 3a An example of a false synthetic signature is shown;

[0055] Figure 3b An example of an invalid signature is shown;

[0056] Figure 4 A schematic diagram showing a flow chart of a signature forgery detection method provided by another embodiment of the present invention is shown:

[0057] Figure 5 A schematic flow chart of a method for detecting signature forgery provided by another embodiment of the present invention is shown;

[0058] Figure 6 A schematic diagram of the structure of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0059] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0060] Signature synthesis forgery refers to the situation where a single word is forged in the signature image and then synthesized into the signature image. Figure 3a An example of a false synthetic signature is shown, Figure 3a As shown, the handwriting of the character "Li" in the two signature images of "Li Mengna" and "Li Peng" is exactly the same, so these two signatures may be synthesized using the text template of "Li".

[0061] Figure 1 FIG. 4 is a schematic diagram showing the structure of a signature forgery detection device provided by an embodiment of the present invention. Figure 1 As shown, the device includes an interaction layer 11, an interface layer 12 and a service layer 13.

[0062] The interaction layer 11 provides a microservice interface with a restful style, which is convenient for any client application to quickly call, and has portability and scalability. The interface layer 12 provides functions such as receiving requests, calling service modules, parameter verification, logging, and data retention. The service layer 23 includes the following modules: a receiving module 131, a first segmentation module 132, a first feature extraction module 133, and a comparison module 134.

[0063] The receiving module 131 is adapted to receive the signature image to be detected and its text mark scheduled by the interface layer. The text mark of the signature image to be detected refers to each text in the signature image to be detected.

[0064] Among them, the interface layer provides functions such as receiving requests, calling service modules, parameter verification, logging, and data retention.

[0065] The first segmentation module 132 is adapted to perform single-word segmentation processing on the signature image to be detected to obtain various text images to be detected. The received signature image to be detected is segmented into multiple single-word images to obtain various text images to be detected.

[0066] The first feature extraction module 133 is adapted to extract the hash code of each to-be-detected text image as a text image feature. For each to-be-detected text image, a preset hash algorithm is used to determine the hash code of the to-be-detected text image.

[0067] The comparison module 134 is suitable for querying in the text feature library according to the text mark, and comparing the text image features of each text image to be detected with each text image feature of the query. If there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, it is determined that the signature image to be detected is a forged signature image.

[0068] Among them, the text feature library stores the text image features (hash codes) corresponding to each single word, that is, the text template. For example, the text image features corresponding to the text images of the word "Li" in different fonts and different writing methods are stored. For any text image to be detected, multiple text image features of the corresponding text are queried in the text feature library according to its text mark, and then the text image features of the text image to be detected are compared with each of the queried text image features, and it is calculated whether the text image features of the text image to be detected are similar to the queried text image features. If the text image features of the text image to be detected are similar to the text image features in the text feature library, it is considered that the text image to be detected is successfully compared.

[0069] The logic for determining whether the signature image to be detected is a forged signature image is: if the text image features of at least one text image to be detected are similar to the text image features in the text feature library, that is, at least one text image to be detected is successfully compared, then it is determined that the signature image to be detected is a synthesized forged signature image.

[0070] For example, if the text mark of the signature image to be detected is "Li", all text image features corresponding to "Li" are queried in the text library, and then the text image features of the text image to be detected are compared with the text image features corresponding to "Li". If they are similar, it means that the character "Li" in the signature image to be detected has an almost identical text template in the feature library, and the "Li" in the signature image to be detected is synthesized, then it is determined that the signature image to be detected is a forged signature image.

[0071] According to the signature forgery detection device provided by the present embodiment, by receiving the signature image to be detected and its text mark scheduled from the interface layer, the signature image to be detected is segmented with characters as the granularity, and then, the text image features of each single-word image to be detected are extracted, and the text image features in the text feature library are queried according to the text mark. Finally, the text image features of each single-word image to be detected are compared with the queried text image. If there is at least one text image feature of the single-word image to be detected that is similar to the queried text image feature, the signature image to be detected is determined to be a forged signature image. Through the above-mentioned method, the problem that the existing signature forgery method fails to solve the problem of electronic signatures and forged texts without fixed templates in actual business scenarios is solved. The single-word image to be detected and the single-word image in the feature library are directly compared by feature differences without the need for comparison between images, which can improve the processing speed, can improve the precision and accuracy of detecting synthetic forged signature images, and has portability and extensibility.

[0072] Figure 2 A schematic diagram showing a schematic diagram of the structure of a signature forgery detection device provided by another embodiment of the present invention is shown in FIG. Figure 2 As shown, the device includes an interaction layer 21, an interface layer 22 and a service layer 23.

[0073] Among them, the interaction layer 21 provides a microservice interface with a restful style, which is convenient for any client application to quickly call, and has portability and scalability. The interface layer 22 provides functions such as receiving requests, calling service modules, parameter verification, logging, and data retention. The service layer 23 includes the following modules: a second segmentation module 231, a second feature extraction module 232, a storage module 233, a receiving module 234, a first segmentation module 235, a filtering module 236, a first feature extraction module 237, a comparison module 238, an adaptive addition module 239, and an adaptive deletion module 2310.

[0074] The service layer 23 is the core of the device of this embodiment, and its main functions include: forged text base library generation, signature forgery detection, text base library deletion and addition, which will be described below.

[0075] Regarding the generation of the forged text base library (that is, the generation of the text feature library), the second segmentation module 231 and the second feature extraction module 232 and the storage module 233 are mainly used for the generation of the forged text base library, specifically:

[0076] The second segmentation module 231 is adapted to perform single-word segmentation processing on the acquired historical forged signature image to obtain at least one text image.

[0077] To detect whether a signature belongs to a synthetic forgery, it is necessary to clarify the prototype of the forged single character, that is, the template text. In this embodiment, historical forged signatures are used as the template text, including artistic characters, machine-generated characters, handwritten signatures, PS characters, and so on.

[0078] Specifically, the second segmentation module 231 performs binarization processing on the obtained historical forged signature image, performs connected component analysis on the binarized historical forged signature image, and performs segmentation processing on different connected components to obtain individual character images.

[0079] For example, the obtained historical forged signature image is binarized, and the connected component analysis of the signature in the image is performed using the 8-neighborhood pixel method. A connected component is composed of foreground pixel points with the same pixel value and adjacent positions, representing a single character. By segmenting different connected components, the signature image is segmented into single-character images.

[0080] Optionally, to facilitate the representation of the image hash feature, the second segmentation module 231 is also adapted to: scale each character image to a preset size, and perform grayscale processing on the scaled character images. For example, the segmented single-character images are uniformly scaled to a size of 12*11, and the scaled images are grayscale processed.

[0081] The second feature extraction module 232 is adapted to extract the hash code of at least one character image as the character image feature.

[0082] Specifically, for any character image, the difference between the row pixels of the character image is calculated using the difference hash algorithm to obtain a difference matrix, and the average value of all elements in the difference matrix is calculated; each element in the difference matrix is compared with the average value, and a feature string is determined according to the comparison result, and the feature string is converted to hexadecimal to obtain the hash code of the character image.

[0083] Continuing with the above example, using the DHA algorithm (difference hash algorithm), by calculating the difference values between the row pixels, a 11*11 difference matrix G is obtained, the average value a of all elements in the G matrix is calculated, and dhash represents the string after the difference calculation. Traverse the pixels of the matrix G row by row from left to right. If G(i,j)>=a, then dhash+="1"; if G(i,j)<a, then dhash+="0". The obtained string composed of 121 numbers after the difference calculation is converted to hexadecimal, which is the hash code corresponding to the single-character image.

[0084] The storage module 233 is adapted to store the text image features of at least one text image in a text feature library; wherein different text image features corresponding to the same text are stored in a feature table corresponding to the text. A data table is constructed for each single word, and different text image features corresponding to the same single word are stored in the data table corresponding to the single word, thereby generating a text feature library.

[0085] The receiving module 234 receives the signature image to be detected and its text mark scheduled by the interface layer.

[0086] The first segmentation module 235 is adapted to perform single-word segmentation processing on the signature image to be detected to obtain individual character images to be detected.

[0087] The principle of the first segmentation module is similar to that of the second segmentation module, except that the object of processing is the signature image to be detected. Specifically, the signature image to be detected is binarized, and a connected domain analysis is performed on the binarized signature image to be detected, and different connected domains are segmented to obtain various text images to be detected.

[0088] The filtering module 236 is adapted to determine whether the number of different connected domains of the signature image to be detected is consistent with the number of text marks, and if not, filter the signature image to be detected.

[0089] The filtering module filters invalid signatures. Combined with the spacing characteristics between single words in the signature synthesis and forgery scenario, the number of different connected domains in the same image is counted during image segmentation. If the number of different connected domains is inconsistent with the number of text marks, it is judged that there may be connected strokes or missing strokes. It is believed that synthetic forgery will not occur in this case, and the signature to be detected is judged invalid, which better realizes the filtering of invalid signatures and improves the recognition accuracy. Figure 3b An example of an invalid signature is shown, in which the individual words are connected and are not usually synthetic forgeries.

[0090] The first feature extraction module 237 is adapted to extract the hash code of each text image to be detected as a text image feature.

[0091] The principle of the first feature extraction module is similar to that of the second feature extraction module, except that the object of processing is the various text images to be detected that are segmented from the signature image to be detected. Specifically, for any text image to be detected, the difference hash algorithm is used to calculate the difference between the row pixels of the text image to be detected, and a difference matrix is ​​obtained, and the average value of all elements in the difference matrix is ​​calculated; each element in the difference matrix is ​​compared with the average value, and a feature character string is determined according to the comparison result, and the feature character string is converted into hexadecimal to obtain the hash code of the text image to be detected.

[0092] The comparison module 238 is suitable for querying in the text feature library according to the text mark, and comparing the text image features of each text image to be detected with each text image feature of the query. If there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, it is determined that the signature image to be detected is a forged signature image.

[0093] The comparison module 238 mainly realizes the query and comparison of the signature image to be detected and the text feature library. Specifically, for any text image to be detected, multiple text image features are queried in the text feature library according to the text mark of the text image to be detected, and the similarity between the text image feature of the text image to be detected and each text image feature of the query is calculated. If the similarity is greater than the similarity threshold, it is determined that the text image to be detected is similar to the text image feature of the query, and the comparison of the text image to be detected is successful. Otherwise, it indicates that the comparison of the text image to be detected is unsuccessful.

[0094] Specifically, the Hamming distance dist between the text image feature of the text image to be detected and the text image feature of the query is calculated, and then the similarity between the two is calculated according to the formula similarity sim=1-dist / (n*n), where n represents the horizontal number of the difference matrix. The similarity threshold is obtained based on the existing negative class data and labels.

[0095] The logic of the comparison module 238 for determining whether the signature image to be detected is a forged signature image is: if the text image features of at least one text image to be detected are similar to the text image features in the text feature library, that is, at least one text image to be detected is successfully compared, then it is determined that the signature image to be detected is a synthesized forged signature image.

[0096] Optionally, the comparison module 238 is further adapted to: first compare all the text images to be detected corresponding to the signature image to be detected, and determine whether the comparison of each text image to be detected is successful. If all the comparisons are successful, it indicates that the single words in the signature image to be detected are all forged and synthesized, and the signature image to be detected is directly determined to be a signature forgery. If some of the comparisons are successful, it indicates that some of the single words in the signature image to be detected are forged and synthesized, and the signature image to be detected is still determined to be a signature forgery.

[0097] Optionally, the device also includes: a result returning module, adapted to return to the interface layer the result of whether the signature image to be detected is a forged signature image; and return to the interface layer the position information of the text image to be detected whose features are similar to the queried text image features in the signature image to be detected.

[0098] Normally, a signature image will be segmented into 2-3 text images, and the control of the overall processing time needs to be considered. In the device of this embodiment, each character corresponds to the image features of the different writing methods of the character. The image features of the character are quickly found through single-character marking, and the difference in hash features is used to compare the text to be detected with the text in the text feature library. There is no need to compare between pictures, and the difference in pictures can be quickly and effectively judged, thereby locating forged signature images.

[0099] The adaptive adding module 239 is adapted to store the text image features of other text images to be detected, except for the text images to be detected whose text image features are similar to the queried text image features, into the text feature library.

[0100] The adaptive adding module 239 mainly realizes the expansion of the text template in the text feature library, and determines the content added to the text feature library according to the result of the comparison module. Specifically, the text image features of other text images to be detected except the successfully compared text images to be detected are stored in the text feature library. If the text corresponding to the text image to be detected already exists in the text feature library, the text image features are directly added to the feature table corresponding to the text; if the text corresponding to the text image to be detected does not exist in the text feature library, a feature table of the text is created according to the single-word mark, and the text image features are added to the created feature table.

[0101] The reason why the device of this embodiment uses the text feature library of existing data to continuously enter incremental data is as follows: (1) Signature synthesis is random, and the forged text does not have a fixed template, and the source of the forged text is unknown; (2) There are different channels that use the same text to synthesize multiple signatures to commit forgery. Therefore, by continuously expanding the text feature library, the forgery hit rate can be improved.

[0102] The adaptive deletion module 2310 is suitable for sorting each text image feature in the text feature library according to the frequency with which each text image feature is detected to be similar to the text image feature of the text image to be detected; and when the number of text image features detected in the text feature library exceeds a preset value, the text image features whose frequency of being detected to be similar to the text image features of the text image to be detected is lower than the preset value are deleted.

[0103] Unlimited addition of text image features to the text feature library will lead to overflow of storage space and increase comparison time. Based on this, the device of this embodiment also provides an adaptive deletion module to appropriately delete text image features in the text feature library.

[0104] Among them, the frequency at which the text image features are detected to be similar to the text image features of the text image to be detected is essentially the frequency at which the text image is used for forged signatures. For example, for a text image feature of the character "Li" in the text feature library, when detecting three signature images to be detected, the character "Li" features in these three images to be detected are all similar to the text image features of the character "Li" in the text feature library. That is, the text image features of the character "Li" are used for forged signatures three times. According to the frequency at which each text image feature is detected to be used for synthesizing forged signatures, each text image feature is arranged in descending order of frequency.

[0105] Considering that most of the text image features added to the text feature library are not forged, text image features with a relatively low frequency of being used for synthesizing forged signatures can be deleted. Specifically, when the number of text image features stored in the text feature library exceeds the limit, text image features with a frequency lower than the preset value are deleted. In this way, text image features with a low frequency of being used for forged signatures are deleted to achieve deleting useless text and releasing new space for adding new potential text image features.

[0106] According to the signature forgery detection device provided in this embodiment, the device provides an interface through the interaction layer for any client to call, and has portability and scalability. By processing the signature image and converting it into a feature sequence for storage in the database, when detecting a signature image to be detected, on the one hand, filtering invalid signatures according to the connected regions in the image can improve the accuracy of detection. On the other hand, by directly comparing the single-character images to be detected with the single-character images in the feature library through features, without the need for image-to-image comparison, the detection time can be shortened, the forged signature image can be quickly and effectively located, and the accuracy of detection can be ensured. In addition, it can also adaptively add and delete the content in the text feature library, which can improve the hit rate of detecting synthesized forged signatures.

[0107] Figure 4 The flowchart of the signature forgery detection method provided by another embodiment of the present invention is shown, as Figure 4 shown, the method includes the following steps:

[0108] Step S410, receiving the signature image to be detected and its text mark scheduled from the interface layer.

[0109] Step S420, performing single-character segmentation processing on the signature image to be detected to obtain each text image to be detected.

[0110] Step S430, extracting the hash code of each text image to be detected as the text image feature.

[0111] Step S440, querying in the text feature library according to the text mark, comparing the text image features of each text image to be detected with each text image feature of the query, if there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, then determining that the signature image to be detected is a forged signature image.

[0112] Through the above method, the problem that the existing signature forgery methods fail to solve the problem of electronic signatures and forged texts having no fixed templates in actual business scenarios is solved. The single-word image to be detected and the single-word image in the feature library are directly compared through feature differences without the need for image-to-image comparison. This can improve the processing speed, the precision and accuracy of detecting synthetic forged signature images, and has portability and scalability.

[0113] Figure 5 A schematic diagram of a method for detecting signature forgery according to another embodiment of the present invention is shown. The method is applied to the service layer, such as Figure 5 As shown, the method comprises the following steps:

[0114] Step S501, receiving a signature image to be detected and its text mark scheduled by the interface layer.

[0115] Step S502, binarizing the signature image to be detected, performing connected domain analysis on the binarized signature image to be detected, segmenting different connected domains, and obtaining various text images to be detected.

[0116] Step S503, determining whether the number of different connected domains of the signature image to be detected is consistent with the number of text marks, if so, executing step S505; if not, executing step S504.

[0117] Step S504, confirm that the signature is invalid, and jump to step S513. If the number of different connected domains of the signature image to be detected is inconsistent with the number of text marks, the signature to be detected is confirmed to be invalid, and the result that the signature to be detected is invalid is output to the interface layer.

[0118] Step S505: extracting the hash code of each text image to be detected as a text image feature.

[0119] Step S506, searching in the text feature library according to the text mark, and comparing the text image features of each text image to be detected with each text image feature of the search.

[0120] Step S507, determine whether all the text images to be detected are successfully compared; if so, execute step S508; if not, execute step S509.

[0121] Step S508, confirm that the signature image to be detected is a forged signature image, and jump to step S513; if all the text images to be detected are successfully compared, directly confirm that the signature image to be detected is a forged signature image, and output the result that the signature image to be detected is a forged signature image to the interface layer.

[0122] Step S509, determining whether the comparison of some of the texts to be detected with the pictures is successful, if so, executing step S510; if not, executing step S511;

[0123] Step S510, confirm that the signature image to be detected is a forged signature image, mark the position of the forged single word, and jump to step S512 and step S513. If some of the text images to be detected are successfully compared, the signature image to be detected is confirmed to be a forged signature image, the position information of the forged single word in the signature image to be detected is determined, and the interface layer outputs the result, and the text image features corresponding to the normal text are added to the text feature library.

[0124] Step S511, determine that the signature image to be detected is a normal signature image, and jump to step S512 and step S513. If no text image to be detected is successfully compared, it is determined that the signature image to be detected is a normal signature image, the text image features corresponding to the normal text are added to the text feature library, and the result that the signature image to be detected is a normal signature image is output to the interface layer.

[0125] It can be seen that by processing the signature image and converting it into a feature sequence and storing it in the database, when detecting the signature image to be detected, on the one hand, invalid signatures can be filtered out according to the connected domains in the image, which can improve the accuracy of detection. On the other hand, by directly comparing the single-word image to be detected with the single-word image in the feature library through features, without the need to compare images with each other, the detection time can be shortened, the forged signature image can be quickly and effectively located, and the accuracy of detection can be guaranteed; in addition, the content in the text feature library can be adaptively added and deleted, which can improve the hit rate of detecting synthetic forged signatures.

[0126] An embodiment of the present invention provides a non-volatile computer storage medium, wherein the computer storage medium stores at least one executable instruction, and the computer executable instruction can execute the signature forgery detection method in any of the above method embodiments.

[0127] The executable instructions may be used to cause the processor to perform the following operations:

[0128] Receive the signature image to be detected and its text mark scheduled by the interface layer;

[0129] Perform single-word segmentation processing on the signature image to be detected to obtain each text image to be detected;

[0130] Extract the hash code of each text image to be detected as the text image feature;

[0131] Search the text feature library based on the text mark, compare the text image features of each text image to be detected with the text image features of the query, and if there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, then determine that the signature image to be detected is a forged signature image

[0132] In an optional manner, the executable instruction causes the processor to perform the following operations:

[0133] Segment the acquired historical forged signature image to obtain at least one text image;

[0134] Extracting a hash code of at least one text image as a text image feature;

[0135] The text image features of at least one text image are stored in a text feature library; wherein different text image features corresponding to the same text are stored in a feature table corresponding to the text.

[0136] In an optional manner, the executable instruction causes the processor to perform the following operations:

[0137] For any text image to be detected, the difference between the row pixels of the text image to be detected is calculated using the difference hash algorithm to obtain a difference matrix, and the average value of all elements in the difference matrix is ​​calculated; each element in the difference matrix is ​​compared with the average value, and a feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image to be detected;

[0138] For any text image, the difference hash algorithm is used to calculate the difference between the row pixels of the text image, obtain the difference matrix, and calculate the average value of all elements in the difference matrix; each element in the difference matrix is ​​compared with the average value, and the feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image.

[0139] In an optional manner, the executable instruction causes the processor to perform the following operations:

[0140] Binarize the signature image to be detected, perform connected domain analysis on the binary-processed signature image to be detected, segment different connected domains, and obtain various text images to be detected;

[0141] The historical forged signature images are binarized, connected domain analysis is performed on the binarized historical forged signature images, and different connected domains are segmented to obtain various text images.

[0142] In an optional manner, the executable instruction enables the processor to perform the following operations: determine whether the number of different connected domains of the signature image to be detected is consistent with the number of text marks, and if not, filter the signature image to be detected.

[0143] In an optional manner, the executable instructions cause the processor to perform the following operations: storing the text image features of other text images to be detected except for the text image to be detected whose text image features are similar to the queried text image features in a text feature library.

[0144] In an optional manner, the executable instructions cause the processor to perform the following operations: sorting each text image feature in the text feature library according to the frequency with which each text image feature is detected to be similar to the text image feature of the text image to be detected; and when the number of text image features detected in the text feature library exceeds a preset value, deleting the text image features whose frequency of being detected to be similar to the text image features of the text image to be detected is lower than the preset value.

[0145] Through the above method, the problem that the existing signature forgery methods fail to solve the problem of electronic signatures and forged texts having no fixed templates in actual business scenarios is solved. The single-word image to be detected and the single-word image in the feature library are directly compared through feature differences without the need for image-to-image comparison. This can improve the processing speed, the precision and accuracy of detecting synthetic forged signature images, and has portability and scalability.

[0146] Figure 6 The schematic diagram of the structure of the computing device embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0147] like Figure 6 As shown, the computing device may include: a processor (processor) 602 , a communication interface (Communications Interface) 604 , a memory (memory) 606 , and a communication bus 608 .

[0148] The processor 602, the communication interface 604, and the memory 606 communicate with each other via a communication bus 608. The communication interface 604 is used to communicate with other devices such as a client or other server network elements. The processor 602 is used to execute a program 610, which can specifically execute the relevant steps in the above-mentioned signature forgery detection method embodiment for a computing device.

[0149] Specifically, the program 610 may include program codes, which include computer operation instructions.

[0150] The processor 602 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs; or may be processors of different types, such as one or more CPUs and one or more ASICs.

[0151] The memory 606 is used to store the program 610. The memory 606 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0152] The program 610 may be specifically configured to enable the processor 602 to perform the following operations:

[0153] Receive the signature image to be detected and its text mark scheduled by the interface layer;

[0154] Perform single-word segmentation processing on the signature image to be detected to obtain each text image to be detected;

[0155] Extract the hash code of each text image to be detected as the text image feature;

[0156] A query is performed in the text feature library based on the text mark, and the text image features of each text image to be detected are compared with each text image feature of the query. If there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, it is determined that the signature image to be detected is a forged signature image.

[0157] In an optional manner, the program 610 enables the processor 602 to perform the following operations:

[0158] According to another aspect of the present invention, a method for detecting signature forgery is provided, the method being applied to a service layer, comprising:

[0159] Optionally, the method further comprises:

[0160] Segment the acquired historical forged signature image to obtain at least one text image;

[0161] Extracting a hash code of at least one text image as a text image feature;

[0162] The text image features of at least one text image are stored in a text feature library; wherein different text image features corresponding to the same text are stored in a feature table corresponding to the text.

[0163] Optionally, extracting the hash code of each to-be-detected text image as a text image feature further includes:

[0164] For any text image to be detected, the difference between the row pixels of the text image to be detected is calculated using the difference hash algorithm to obtain a difference matrix, and the average value of all elements in the difference matrix is ​​calculated; each element in the difference matrix is ​​compared with the average value, and a feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image to be detected;

[0165] Extracting a hash code of at least one text image as a text image feature further includes:

[0166] For any text image, the difference hash algorithm is used to calculate the difference between the row pixels of the text image, obtain the difference matrix, and calculate the average value of all elements in the difference matrix; each element in the difference matrix is ​​compared with the average value, and the feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image.

[0167] Optionally, performing single-word segmentation processing on the signature image to be detected to obtain each text image to be detected further includes: performing binarization processing on the signature image to be detected, performing connected domain analysis on the binarized signature image to be detected, and segmenting different connected domains to obtain each text image to be detected;

[0168] Segmenting the acquired historical forged signature image to obtain at least one text image further includes: binarizing the historical forged signature image, performing connected domain analysis on the binarized historical forged signature image, and segmenting different connected domains to obtain various text images.

[0169] Optionally, the method further includes: determining whether the number of different connected domains of the signature image to be detected is consistent with the number of text marks; if not, filtering the signature image to be detected.

[0170] Optionally, the method further comprises: storing the text image features of other text images to be detected, except for the text images to be detected whose text image features are similar to the queried text image features, in a text feature library.

[0171] Optionally, the method further includes: sorting each text image feature in the text feature library according to the frequency with which each text image feature in the text feature library is detected to be similar to the text image feature of the text image to be detected; and when the number of text image features detected in the text feature library exceeds a preset value, deleting the text image features whose frequency of being detected to be similar to the text image features of the text image to be detected is lower than the preset value.

[0172] Through the above method, the problem that the existing signature forgery methods fail to solve the problem of electronic signatures and forged texts having no fixed templates in actual business scenarios is solved. The single-word image to be detected and the single-word image in the feature library are directly compared through feature differences without the need for image-to-image comparison. This can improve the processing speed, the precision and accuracy of detecting synthetic forged signature images, and has portability and scalability.

[0173] The algorithm or display provided herein is not inherently related to any particular computer, virtual system or other equipment. Various general purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious to construct the structure required for this type of system. In addition, the embodiment of the present invention is not directed to any specific programming language yet. It should be understood that various programming languages ​​can be utilized to realize the content of the present invention described herein, and the description made to specific languages ​​above is for disclosing the best mode of the present invention.

[0174] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures and techniques are not shown in detail so as not to obscure the understanding of this description.

[0175] Similarly, it should be understood that in order to streamline the present invention and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof. However, this disclosed method should not be interpreted as reflecting the following intention: that the claimed invention requires more features than the features explicitly recited in each claim. More specifically, as reflected in the claims below, inventive aspects lie in less than all the features of the individual embodiments disclosed above. Therefore, the claims that follow the specific embodiment are hereby expressly incorporated into the specific embodiment, with each claim itself serving as a separate embodiment of the present invention.

[0176] Those skilled in the art will appreciate that the modules in the devices in the embodiments may be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments may be combined into one module or unit or component, and in addition they may be divided into a plurality of submodules or subunits or subcomponents. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification (including the accompanying claims, abstracts and drawings) and all processes or units of any method or device disclosed in this manner may be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification (including the accompanying claims, abstracts and drawings) may be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0177] In addition, those skilled in the art will appreciate that, although some embodiments herein include certain features included in other embodiments but not other features, the combination of features of different embodiments is meant to be within the scope of the present invention and form different embodiments. For example, in the claims below, any one of the claimed embodiments may be used in any combination.

[0178] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all functions of some or all components according to embodiments of the present invention. The present invention can also be implemented as a device or apparatus program (e.g., computer program and computer program product) for executing part or all of the methods described herein. Such a program implementing the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0179] It should be noted that the above embodiments illustrate the present invention rather than limit it, and that those skilled in the art may devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference symbol between brackets shall not be construed as a limitation on the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "one" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention may be implemented by means of hardware comprising a number of different elements and by means of a suitably programmed computer. In a unit claim enumerating a number of devices, several of these devices may be embodied by the same hardware item. The use of the words first, second, and third, etc. does not indicate any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be understood as limitations on the order of execution.

Claims

1. A signature forgery detection device, the device comprising an interaction layer, an interface layer and a service layer, wherein: The service layer includes: A receiving module, adapted to receive the signature image to be detected and its text mark scheduled by the interface layer; The first segmentation module is adapted to perform single-word segmentation processing on the signature image to be detected to obtain individual character images to be detected; A first feature extraction module, adapted to extract the hash code of each of the to-be-detected text images as a text image feature; The first feature extraction module is further adapted to: for any text image to be detected, use a difference hash algorithm to calculate the difference between row pixels of the text image to be detected, obtain a difference matrix, and calculate the average value of all elements in the difference matrix; compare each element in the difference matrix with the average value, determine a feature string according to the comparison result, convert the feature string into hexadecimal, and obtain a hash code of the text image to be detected; a comparison module, adapted to perform a query in a text feature library according to the text mark, compare the text image features of each of the text images to be detected with the text image features of the query, and determine that the signature image to be detected is a forged signature image if at least one of the text image features of the text image to be detected is similar to the text image feature of the query; The device also includes: a second feature extraction module, suitable for extracting a hash code of at least one text image as a text image feature; the second feature extraction module is further suitable for: for any text image, using a difference hash algorithm to calculate the difference between row pixels of the text image, obtaining a difference matrix, and calculating the average value of all elements in the difference matrix; comparing each element in the difference matrix with the average value, determining a feature string according to the comparison result, converting the feature string into hexadecimal, and obtaining the hash code of the text image; wherein the text image is obtained by segmenting the acquired historical forged signature image, and the text image features of the text image are stored in a text feature library.

2. The device according to claim 1, wherein: The device further comprises: A second segmentation module is adapted to segment the acquired historical forged signature image to obtain at least one text image; The device further comprises: a storage module adapted to store the text image features of the at least one text image in a text feature library; wherein different text image features corresponding to the same text are stored in a feature table corresponding to the text.

3. The device according to claim 2, wherein: The first segmentation module is further adapted to: perform binarization processing on the signature image to be detected, perform connected domain analysis on the binarized signature image to be detected, and perform segmentation processing on different connected domains to obtain various text images to be detected; The second segmentation module is further adapted to: binarize the historical forged signature image, perform connected domain analysis on the binarized historical forged signature image, and segment different connected domains to obtain various text images.

4. The device according to claim 1, wherein: The device further comprises: The filtering module is adapted to determine whether the number of different connected domains of the signature image to be detected is consistent with the number of the text marks, and if not, filter the signature image to be detected.

5. The device according to claim 1, wherein: The device further comprises: The adaptive adding module is adapted to store the text image features of other text images to be detected, except for the text image features of the text image to be detected which are similar to the queried text image features, into the text feature library.

6. The device according to claim 1, wherein: The device further comprises: an adaptive deletion module, adapted to sort the text image features in the text feature library according to the frequency with which the text image features are detected to be similar to the text image features of the text image to be detected; and When the number of text image features detected in the text feature library exceeds a preset value, the text image features detected to be similar to the text image features of the text image to be detected and whose frequency is lower than the preset value are deleted.

7. A method for detecting signature forgery, the method being applied to a service layer, comprising: Receive the signature image to be detected and its text mark scheduled by the interface layer; Performing single-word segmentation processing on the signature image to be detected to obtain individual text images to be detected; Extract the hash codes of the text images to be detected as text image features; wherein, for any text image to be detected, use the difference hash algorithm to calculate the difference between the row pixels of the text image to be detected, obtain a difference matrix, and calculate the average value of all elements in the difference matrix; compare each element in the difference matrix with the average value, determine the feature string according to the comparison result, convert the feature string into hexadecimal, and obtain the hash code of the text image to be detected; Searching in a text feature library according to the text mark, comparing the text image features of each of the text images to be detected with the text image features of the query, and if there is at least one text image feature of the text image to be detected that is similar to the text image feature of the query, determining that the signature image to be detected is a forged signature image; The text image is obtained by segmenting the acquired historical forged signature image, and the text image features of the text image are stored in a text feature library; Among them, the text image features of the text image are determined through the following steps: for any text image, the difference between the row pixels of the text image is calculated using the difference hash algorithm to obtain a difference matrix, and the average value of all elements in the difference matrix is ​​calculated; each element in the difference matrix is ​​compared with the average value, and a feature string is determined according to the comparison result, and the feature string is converted into hexadecimal to obtain the hash code of the text image, and the hash code is used as the text image feature of the text image.

8. A computing device comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the signature forgery detection method as claimed in claim 7.

9. A computer storage medium, wherein at least one executable instruction is stored in the storage medium, wherein the executable instruction enables a processor to execute operations corresponding to the signature forgery detection method according to claim 7.

Images