TWI767837B - Nuclear print detection method and system - Google Patents

Abstract

The present invention provides a verification method and system for verification of signature, which is to perform signature verification on a document file A with a signature seal on a document and a signature signature B with a signature signature previously stored on a white paper. First, a scale-invariant feature conversion algorithm is used to perform the signature verification. If the number of corresponding point groups does not reach the basic value, the similarity will be excluded first. If the number of corresponding point groups does not reach the basic value, then affine transformation will be used to judge the size consistency of the two. If they do not match, they will be excluded. Otherwise, the correlation coefficient will be used. The formula inquires about the correlation between the two printed texts. When the correlation does not meet the predetermined threshold value, it is considered irrelevant. Otherwise, the terminal verification procedure is continued, and the corresponding point groups are used to generate a color printed text information, which is combined with various Compare with the standard value of the reservation. In this way, the inspection accuracy can be greatly improved by using the sequential judgment of the present invention.

Description

Nuclear print detection method and system

The present invention is related to the field of checking whether the seal on any document is consistent with the seal previously used, such as a seal certificate, especially a kind of information that uses multiple sequential analysis mechanisms to greatly improve the check. Accurate nuclear print detection method and system.

In daily life, there is no lack of experience of using a seal to express the intention of the parties. In addition, Article 3, Paragraph 2 of the Civil Code of our country also regulates that "if a person signs with a seal, the seal has the same effect as the signature." In other words, it is common to use a seal instead of a signature, and a legal person has no possibility of signing, so when a legal person is a sign of intent, it still needs to be printed by a natural person. Therefore, the use of seals instead of signatures has become the norm for all kinds of clerical behaviors in life to be accompanied by expressions of intent.

However, since the seal is a movable property, it is not unpredictable that the movable property will be used after the possession of the movable property. Therefore, if the seal is stolen or not held by the real right holder, it may lead to mistakes that the real right holder does not want to present. or hypocrisy and other untrue expressions of meaning, which in turn lead to subsequent legal issues and social disputes. The job is that, in practice, for important attributes that need to be expressed as a document, the undertaker often needs to verify the consistency of the seal before and after, or even requires a seal certificate to be submitted to the household registration authority to apply for retention as a basis for comparison. However, at present, the comparison between before and after the two seals is mostly carried out manually, and the manual comparison method is nothing more than side-by-side and then recognized by the naked eye at the same time, or some personnel supplemented by superimposing and then transmissive lighting for the naked eye. Comparison, but its effect is limited by the shading of the physical paper, and the effect is not good. Since the process of making seals is not complicated, there is a risk of misjudgment if the details are deliberately counterfeited. Therefore, there is no shortage of technologies for accurate comparison using computer-aided systems.

In order to improve the convenience of judgment after the aforementioned manual overlapping, for example, "CN 1313965 C; Seal checking device and seal checking method" uses computer system analysis to replace the visual judgment mechanism after manual overlapping. Therefore, in practice, after the two printed images to be compared are superimposed, it is judged whether they are consistent according to the degree of difference of the overlap. However, the prerequisite for the accuracy of this approach is how to effectively align and then superimpose. However, how to effectively superimpose the previous technology and not much ink, so this technology is at best just a manual superimposition of the printed text. After the combination, the judgment of the differences between them is supplemented by computer judgment.

Another example is "CN 109657666 A; A method and device for identifying circular seals", for circular seals with circular characters, further taking the center point of the circle as the origin, and placing the circular characters in the rectangular coordinate system of the seal text map. For example, with the rotation correction, the polar coordinate transformation process is performed, and then the stamp text image under the corresponding polar coordinates can be saved for subsequent comparison, thereby greatly reducing the occupation of storage resources. This technical method is mainly aimed at the circular seal as the object, so the concept of the center point position and the polar coordinate conversion after the selection and correction is derived. Therefore, it is not applicable to general seals.

Another example is "CN 110378351 A; Seal identification method and device", which compares the image of the seal to be identified with the corresponding template seal, and divides it into multiple areas, and then calculates the total number of seals in each area. The pixel ratio of the printed text. Further, after calculating the inner product of the geometric mean distance by using the ratio of these pixels, and comparing it with a threshold value, it is judged whether the two are the same printed text. This method uses the pixel ratio as the representation of the printed text, and then calculates the correlation between the pixel ratio and the inner product of the geometric average distance. The accuracy of this single judgment mechanism depends on the number of cutting areas. However, if the cutting area is too small, the resulting computing system is overburdened and easy to read printing errors. If the cutting area is too large, the accuracy will be greatly reduced. Therefore, the technology only has a single judgment mechanism, and it cannot effectively improve the recognition accuracy for all types of printed texts.

Another example is "CN 110472623 A; Image detection method, equipment and system", which first locates the object to be detected, and further separates the printed text area from the background area and then extracts it. The purpose is to improve the accuracy of the extracted printed text area, which can be used as the basis for judging whether the printed text is consistent in the future, and the technology of extracting the printed text area is supplemented by color difference basis or image segmentation algorithm, etc. The area is effectively separated from the printed area. For example, for the edge position of the print area, the edge line of the print area is adjusted by using the difference from the standard value. Therefore, the prior art focuses on how to accurately and effectively extract the to-be-identified printed text area to be judged, and the judgment after extraction is not the focus of the prior art.

In view of the fact that the seal cannot be perfectly displayed in a standard horizontal or vertical position when it is stamped, the prior art also has the technology of positioning the two first and then superimposing and comparing, thereby replacing the manual superposition judgment, such as "CN 111353485 A ;Stamp recognition method, device, equipment and medium", the center of the stamp pattern area to be recognized and the center point of the target stamp image are coincident, and then according to the number of overlapping pixels, the matching degree between them is inquired as a technical means to judge whether they are consistent or not. . Since the center point depends on the angle of the printed text, it often happens that it is not correct after overlapping, so this technology is supplemented by rotating the printed text once for adjusting the positioning and correction means; or projecting in the horizontal and vertical directions to find the center point position; or use the color composition of the print area to determine the possible range, and then seek the subsequent positioning accuracy; or adjust the grayscale image of the print brightness to facilitate subsequent judgments. In other words, the prior art is mainly aimed at how to find the center point and the technical means of determining whether the printed texts are consistent by using the overlapping number of pixels after overlapping. Therefore, if there is a large deviation between the center point and the print angle, the accuracy of this single comparison technique of pixel overlap ratio will be greatly reduced.

In view of this, the inventor of the present invention has assembled many years of rich experience in related industries, and considered the deficiencies of the above-mentioned various prior art, and then conceived a method and system for the detection of nuclear seals. Any discrepancies are eliminated in stages, thereby greatly improving the inspection accuracy.

The purpose of the present invention is to improve the conventional method of judging whether the seal before and after the seal is consistent. Initially, the method of finding the center point is replaced by the corresponding point matching method, and then the size consistency, grayscale correlation, and finally The degree of difference after matching is compared and judged in stages to improve the efficiency and accuracy of inspection.

In order to achieve the above-mentioned purpose, the present invention proposes a method for checking the seal, which is for a document file A with a seal on a document, and a seal file B left on a white paper with a seal for seal checking, Wherein the document file A and the signature file B are both computer image information, including the following steps: using a scale-invariant feature conversion algorithm to calculate and extract at least one feature matching between the document file A and the signature file B. Corresponding point group; wherein, the corresponding point group includes (PA, PB), PA is a point on the document file A, and PB is a point on the signature file B; determine whether the number of the corresponding point group is less than the preset A basic value, when the number of corresponding point groups is less than the preset basic value, it is determined that the seals of the document file A and the seal file B are different; when the number of corresponding point groups is greater than or equal to the basic value, continue to the next step Utilize this corresponding point group to calculate the affine transformation, to judge whether the seal size of this document file A and this signature document B is consistent, when the size is inconsistent, namely judge that this document document A and the seal of this signature document B are not identical; When the sizes are the same, go to the next step; use the correlation coefficient formula to calculate a correlation r with the grayscale values within the seal range of the document file A and the seal file B, if the correlation r is less than a threshold value, it is determined that the seals of the document file A and the seal file B are not the same; if the correlation r is greater than or equal to the threshold value, proceed to the next step; and the corresponding point group (PA) in the document file A and the seal file B , PB) according to the characteristics to generate a color print information with four color attributes, and define the pixel color of the PA point as CA, the pixel color of the PB point as CB, and (1) a first color: CA and CB are similar to white; (2) a second color: CA is similar to white, CB is not similar to white; (3) a third color: CA is not similar to white, CB is similar to white; (4) a first color Four colors: CA and CB are not similar to white; wherein, when the area of any of the second color blocks or the area of any of the third color blocks exceeds a preset area value, it is determined that the file A is different from The seal of the seal file B is not the same; otherwise, it is determined that the seal of the file A and the seal file B are the same; or the perimeter of any second color block or the perimeter of any third color block When it exceeds a preset perimeter value, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are the same; or any PB point is used as the In the center, when the number of dots of the second color or the third color within the enclosed range of the outer frame is greater than a preset value, it is determined that the seals of the document A and the seal file B are different; otherwise, That is, it is determined that the seal of the document file A and the seal file B are the same.

，當(A _00*A ₁₁)-(A _01*A ₁₀)不等於1時，即判定該文件檔A與該印鑑檔B之尺寸不同；當(A _00*A ₁₁)-(A _01*A ₁₀)等於1時，即判定該文件檔A與該印鑑檔B之尺寸相同。 In one embodiment, the step of "using the corresponding point group to calculate an affine transformation to determine whether the size of the seal of the document A and the seal file B is consistent" is to define the point coordinates of the document A as ( Xa, Ya), the point coordinates of the signature file B are (Xb, Yb), and the linear relationship satisfies

, when (A _00* A ₁₁ )-(A _01* A ₁₀ ) is not equal to 1, it is determined that the size of the document file A and the signature file B are different; when (A _00* A ₁₁ )-(A _01* When A ₁₀ ) is equal to 1, it is determined that the document file A and the signature file B have the same size.

。 In the previous embodiment, the step of "using the correlation coefficient formula to calculate the correlation r with the grayscale values within the seal range of the document file A and the seal file B" is to define a corresponding point group (PA, PB ) pixel grayscales are Ga(PA) and Gb(PB), and the average grayscale value MA of the seal range of the document file A, the average grayscale value MB of the seal range of the seal file B, the correlation

.

Continuing from the previous embodiments, the basic value on which the corresponding point group is compared is the number 3. And wherein the first color is white, the second color is red, the third color is blue, and the fourth color is green.

In order to achieve the above-mentioned purpose of the present invention, the present invention further proposes a stamp detection system, which is aimed at using an image capture device to scan a document with a stamp and a white paper with a pre-existing stamp to generate a document file. A and a signature file B for performing the signature checking of the document file A and the signature file B, including: a storage for storing the imported information of the document file A and the signature file B; a matching A device is telecommunicationly connected to the storage, and the matcher performs scale-invariant feature conversion for calculation and extracts the information of at least one corresponding point group whose features match each other between the document file A and the signature file B; wherein, the corresponding The point set includes (PA, PB), PA is a point on the document file A, PB is a point on the signature file B; The comparator receives the information of the corresponding point group, and determines whether the number of the corresponding point group is less than a preset basic value, and when the corresponding point group number is less than the preset basic value, it determines the document file A and the seal The seals of file B are not the same; on the contrary, when the number of the corresponding point group is greater than or equal to the basic value, a number of approval information is generated; a size comparator, telecommunication connects the quantity comparator and the storage, the size comparison After receiving the quantity approval information, the device extracts the information of the document file A and the seal file B from the storage, and performs affine transformation to determine whether the size of the seals of the document file A and the seal file B are consistent, and when the sizes are inconsistent , that is, it is determined that the seals of the document file A and the seal file B are not the same; otherwise, a size approval information is generated; After receiving the size approval information, the device extracts the information of the document file A and the seal file B from the storage, and calculates a Relevance r, if the relevancy r is less than a threshold value, it is determined that the seals of the document file A and the seal file B are different; otherwise, if the relevancy r is greater than or equal to the threshold value, a relevancy approval information is generated ; a color marker, telecommunication connects the correlation comparator and the storage, after the color marker receives the correlation approval information, extracts the document file A and the signature file B information from the storage, the Document file A and the corresponding point group (PA, PB) in the signature file B generate color stamp information with one of four color attributes according to the characteristics. The color of the PA point is defined as CA, and the color of the PB point is defined as CB , (1) a first color: CA and CB are similar to white; (2) a second color: CA and white are similar, CB and white are not similar; (3) a third color: CA and white are not similar, CB is similar to white; (4) a fourth color: CA and CB are not similar to white; and a terminal checker, which is connected to the color marker by telecommunication, and receives the color print information for terminal check, wherein, When the area of any of the second color blocks or the area of any of the third color blocks exceeds a predetermined area value, it is determined that the seals of the document file A and the seal file B are different; It is determined that the seals of the document file A and the seal identification file B are the same; or when the perimeter of any of the second color blocks or the perimeter of any of the third color blocks exceeds a preset perimeter value, it is determined that the The seals of the document file A and the seal file B are different; on the contrary, it is determined that the seal of the document file A and the seal file B are the same; When the number of dots of the second color or the third color is greater than a preset value, it is determined that the seals of the document A and the seal B are different; otherwise, it is determined that the document A and the seal The seal of file B is the same.

，當(A _00*A ₁₁)-(A _01*A ₁₀)不等於1時，即判定該文件檔A與該印鑑檔B之尺寸不同；當(A _00*A ₁₁)-(A _01*A ₁₀)等於1時，即判定該文件檔A與該印鑑檔B之尺寸相同，而生成該尺寸核可資訊。 In one embodiment, the size comparator performs affine transformation to define the point coordinates of the document file A as (Xa, Ya), and the point coordinates of the signature file B as (Xb, Yb), and the linear relationship satisfies

, when (A _00* A ₁₁ )-(A _01* A ₁₀ ) is not equal to 1, it is determined that the size of the document file A and the signature file B are different; when (A _00* A ₁₁ )-(A _01* When A ₁₀ ) is equal to 1, it is determined that the document file A and the signature file B have the same size, and the size approval information is generated.

。 According to the previous embodiment, the correlation comparator executes the correlation coefficient formula to define the pixel grayscales of the corresponding point groups (PA, PB) as Ga(PA) and Gb(PB), and the seal of the file A The range gray mean value MA, the seal range gray mean value MB of the seal file B, and the correlation is calculated

.

Continuing from the previous embodiments, the basic value on which the corresponding point group is compared is the number 3. And wherein the first color is white, the second color is red, the third color is blue, and the fourth color is green.

To sum up, the seal detection method and system of the present invention is initially based on a scale-invariant feature conversion algorithm to calculate and extract at least one corresponding point group whose features match each other between the document file A and the seal file B, and then calculate and extract at least one corresponding point group. Judging the number of corresponding point groups for preliminary comparison, then using affine conversion to size consistency for comparison, and then using the correlation coefficient formula and using the image gray value as the correlation analysis and comparison, and finally using the corresponding point group. Based on matching, color print information with four color types is generated for difference analysis. In general, this kind of staged comparison analysis and judgment can greatly improve the inspection efficiency and comparison accuracy.

In order to enable those skilled in the art to clearly understand the content of the present invention, please refer to the following descriptions and drawings.

Please refer to FIGS. 1A to 1E , which are flow charts of steps of the print detection method according to the preferred embodiment of the present invention. As can be seen from the figure, the step S1 of the verification method of the present invention is to scan a document with a seal and a white paper with a pre-existing seal by using an image capture device to generate a document file A and a blank paper respectively. A seal file B for performing the seal check of the document file A and the seal file B. The document is written for any entity of social significance, such as various contracts, applications, certificates, withdrawal slips, checks, etc., and the seal retained before the first refers to the written object for correct comparison, such as practice The above seal certificate is a suitable example. Therefore, in this step, the document file A and the signature file B of the computer image information are converted into the document file A and the signature file B by the image capture device such as a scanner or a color video camera, a camera and the like.

Continuing, step S2: using a scale-invariant feature conversion algorithm to calculate and extract at least one corresponding point group whose features match each other between the document file A and the signature file B; wherein, the corresponding point group includes (PA, PB) , PA is a point on the document file A, and PB is a point on the signature file B. If the document file A and the seal file B are stamped by the same seal, theoretically, PA and PB will correspond to the same position on the physical seal. Among them, Scale-invariant feature transform (SIFT) is a machine vision algorithm used to detect and describe local features in images. It finds extreme points in digital images, such as the edges of objects , corners, intersections, etc., and extract their position, scale, and rotation invariants. Therefore, the method does not search for the concept of the so-called center point, but searches for the matching point between the two images based on the aforementioned invariant, that is, the corresponding point group in the present invention.

Continuing, step S3: judging whether the corresponding point group quantity is less than a preset basic value, when the corresponding point group quantity is less than the preset basic value, it is determined that the seals of the document file A and the seal file B are different, When the number of corresponding point groups is greater than or equal to the basic value, continue to the next step. According to the complexity and size of the seal, the number of basic values can be set to 3, that is, when the number of corresponding point groups is less than 3, it is directly determined that the two seals are different; otherwise, the number of corresponding point groups is greater than When it is equal to 3, it will not be excluded for the time being, and the next comparison step will be performed successively.

，當(A _00*A ₁₁)-(A _01*A ₁₀)不等於1時，即判定該文件檔A與該印鑑檔B之尺寸不同；當(A _00*A ₁₁)-(A _01*A ₁₀)等於1時，即判定該文件檔A與該印鑑檔B之尺寸相同。附帶說明，雖該行列式值為-1時尺寸亦相同，但印章用印應不存在反射關係下之翻轉之情況，故僅於行列式值為1時判定尺寸相符。 Continue, step S4: utilize this corresponding point group to calculate affine transformation, to judge whether the seal size of this document file A and this seal document B is consistent, when the size is inconsistent, namely judge that this document file A and this seal document B have the same size. The stamps are not the same; when the dimensions are the same, continue to the next step. Among them, affine transformation (Affine transformation), refers to in geometry, a linear transformation of a vector space followed by a translation, transformed into another vector space. In other words, by linear translation, rotation, scaling, reflection, shearing, a combination of the aforementioned transformations in any order is performed. Translation transformation and rotation transformation are as the name implies, and the combination of the two is called Euclidean transformation or rigid body transformation; scaling transformation can be further divided into uniform scaling (uniform scaling) and non-uniform scaling (non-uniform scaling). The scaling factor of each coordinate axis is the same (isotropic), the latter is different; if the scaling factor is negative, the reflection transformation will be superimposed; the shear mapping will translate all points proportionally along a specified direction . Accordingly, the point coordinates of the document file A are defined as (Xa, Ya), and the point coordinates of the signature file B are defined as (Xb, Yb), and the linear relationship satisfies

, when (A _00* A ₁₁ )-(A _01* A ₁₀ ) is not equal to 1, it is determined that the size of the document file A and the signature file B are different; when (A _00* A ₁₁ )-(A _01* When A ₁₀ ) is equal to 1, it is determined that the document file A and the signature file B have the same size. Incidentally, although the size is the same when the determinant value is -1, the seal used for the seal should not be inverted under the reflection relationship, so the size is determined to match only when the determinant value is 1.

，據此作為二者之關聯性判斷。 Continue, step S5: use the correlation coefficient formula to calculate a correlation r with the grayscale values within the seal range of the document file A and the seal file B, if the correlation r is less than a threshold value, then determine the document File A and the seal of the seal file B are not the same. Since the correlation is only a statistical value, it may not have a judgment effect for small local differences, such as Wang and Yu, Da and Tai, or cracks that may exist in strokes. Therefore, if the correlation r is greater than or equal to the threshold value , continue to the next step. Among them, the correlation coefficient is a statistical index used to reflect the closeness of the correlation between variables. It is calculated according to the product difference method. It is also based on the deviation between the two variables and their respective averages. The two deviations are multiplied to reflect the two degree of correlation between variables. Therefore, define the pixel grayscales of the corresponding point groups (PA, PB) as Ga(PA) and Gb(PB), and the average grayscale value MA of the seal range of the document file A, and the average grayscale value MB of the seal range of the seal file B , the correlation

, as the correlation judgment between the two.

Continuing, step S6: The corresponding point groups (PA, PB) in the document file A and the signature file B are generated according to the characteristics with one of four color attributes of color print information, and the pixel color of the PA point is defined as CA , the pixel color of the PB point is defined as CB, and (1) a first color: CA and CB are both similar to white; (2) a second color: CA is similar to white, CB is not similar to white; (3) A third color: CA is not similar to white, CB is similar to white; (4) A fourth color: CA and CB are not similar to white; wherein, the area of any of the second color blocks or any of the first When the area of the three-color block exceeds a preset area value, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are the same; When the perimeter of the second color block or the perimeter of any third color block exceeds a predetermined perimeter value, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are different. The document file A has the same seal as the seal file B; or with any PB point as the center, the number of points of the second color or the third color within the enclosed range of the outer frame is greater than a preset number When the value is set, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are the same. The four blocks can be set to the following different colors, the first color is white, the second color is red, the third color is blue, and the fourth color is green. When the identification personnel directly judge with the naked eye that only the green seal appears to be on a white background, the two seals are the same. Even if the color printed information has red or blue, it can be judged whether the two are the same through the size and perimeter of the red or blue. Or, using several of the PB points as the center basis, the number of red or blue pixels under the enclosed area surrounding the outer frame is used as a way of judging whether the two seals are consistent. The above various terminal judgment methods can exist alone or be executed in plural combinations to enhance the accuracy verification.

Please refer to FIG. 2 , which is a functional block diagram of the print detection apparatus according to the preferred embodiment of the present invention. From the figure, the device includes a host 1 and a display 2 . And the stamp detection system 10 of the present invention is installed in the host 1, and functionally, the image capture device is used to scan the document with the stamp and the white paper left in the pre-stamp to generate a document file A and A signature file B for performing signature checking of the document A and the signature file B, comprising: a storage 101, a matching unit 102, a quantity comparator 103, a size comparator 104, a Relevance comparer 105 , a color marker 106 , and an end checker 107 . The storage 101 is used for storing the imported information of the document file A and the signature file B. The document is written for any entity of social significance, such as various contracts, applications, certificates, withdrawal slips, checks, etc., and the seal retained before the first refers to the written object for correct comparison, such as practice The above seal certificate is a suitable example. Therefore, the storage 101 is used to store the document file A and the signature file B which are converted into computer image information by image capturing devices such as scanners, color video cameras or cameras and the like.

The matcher 102 is connected to the storage 101 by telecommunication, and the matcher 102 performs scale-invariant feature transformation for calculating and extracting the information of at least one corresponding point group whose features match each other between the document file A and the signature file B ; wherein, the corresponding point group includes (PA, PB), PA is a point on the document file A, and PB is a point on the signature file B. Among them, Scale-invariant feature transform (SIFT) is a machine vision algorithm used to detect and describe local features in images. It finds extreme points in digital images, such as the edges of objects , corners, intersections, etc., and extract their position, scale, and rotation invariants. Therefore, the method does not search for the concept of the so-called center point, but searches for the matching point between the two images based on the aforementioned invariant, that is, the corresponding point group in the present invention.

The quantity matching unit 103 is connected to the matching unit 102 and the storage 101 by telecommunication, and the quantity matching unit 103 receives the information of the corresponding point group, and determines whether the number of the corresponding point group is less than a predetermined basic value, When the number of corresponding point groups is less than the preset basic value, it is determined that the seals of the document file A and the seal file B are different; on the contrary, when the number of corresponding point groups is greater than or equal to the basic value, a quantity approval information is generated . According to the complexity and size of the seal, the number of basic values can be set to 3, that is, when the number of corresponding point groups is less than 3, it is directly determined that the two seals are different; otherwise, the number of corresponding point groups is greater than When it is equal to 3, it is not excluded for the time being.

，當(A _00*A ₁₁)-(A _01*A ₁₀)不等於1時，即判定該文件檔A與該印鑑檔B之尺寸不同；當(A _00*A ₁₁)-(A _01*A ₁₀)等於1時，即判定該文件檔A與該印鑑檔B之尺寸相同，而生成該尺寸核可資訊。 The size comparator 104 is connected to the quantity comparator 103 and the storage 101 by telecommunication. After receiving the quantity approval information, the size comparator 104 extracts the document file A and the signature file B from the storage 101 information, and perform affine transformation to determine whether the size of the seals of the document file A and the seal file B are consistent. When the sizes are inconsistent, it is determined that the seals of the document file A and the seal file B are different; Approved information. Affine transformation means that in geometry, a linear transformation is performed on a vector space followed by a translation to transform it into another vector space. In other words, by linear translation, rotation, scaling, reflection, shearing, a combination of the aforementioned transformations in any order is performed. Translation transformation and rotation transformation are as the name implies, and the combination of the two is called Euclidean transformation or rigid body transformation; scaling transformation can be further divided into uniform scaling (uniform scaling) and non-uniform scaling (non-uniform scaling). The scaling factor of each coordinate axis is the same (isotropic), the latter is different; if the scaling factor is negative, the reflection transformation will be superimposed; the shear mapping will translate all points proportionally along a specified direction . . Accordingly, the point coordinates of the document file A are defined as (Xa, Ya), the point coordinates of the signature file B are (Xb, Yb), and the linear relationship satisfies

, when (A _00* A ₁₁ )-(A _01* A ₁₀ ) is not equal to 1, it is determined that the size of the document file A and the signature file B are different; when (A _00* A ₁₁ )-(A _01* When A ₁₀ ) is equal to 1, it is determined that the size of the document file A and the signature file B are the same, and the size approval information is generated.

，據此作為二者之關聯性判斷。 The correlation comparator 105 is telecommunicationly connected to the size comparator 104 and the storage 101. After receiving the size approval information, the correlation comparator 105 extracts the document A and the signature from the storage 101. file B information, and calculate a correlation r based on the grayscale values within the seal range of the document file A and the seal file B. If the correlation r is less than a threshold value, it is determined that the document file A and the seal The seals of file B are not the same; on the contrary, if the correlation r is greater than or equal to the threshold value, a correlation approval information is generated. Among them, the correlation coefficient is a statistical index used to reflect the closeness of the correlation between variables. It is calculated according to the product difference method. It is also based on the deviation between the two variables and their respective averages. The two deviations are multiplied to reflect the two degree of correlation between variables. Therefore, define the pixel grayscales of the corresponding point groups (PA, PB) as Ga(PA) and Gb(PB), and the average grayscale value MA of the seal range of the document file A, and the average grayscale value MB of the seal range of the seal file B , the correlation

, as the correlation judgment between the two.

The color marker 106 is telecommunicationly connected to the correlation comparator 105 and the storage 101. After receiving the correlation approval information, the color marker 106 extracts the document file A and the signature file B from the storage 101 Information, the corresponding point groups (PA, PB) in the document file A and the signature file B are generated according to the characteristics of the color seal information with four color attributes, wherein the pixel color of the PA point is defined as the CA and PB points. The pixel color is defined as CB, (1) a first color: CA and CB are similar to white; (2) a second color: CA is similar to white, CB is not similar to white; (3) a third color: CA is not similar to white, CB is similar to white; (4) a fourth color: CA and CB are not similar to white. The four blocks can be set to the following different colors, the first color is white, the second color is red, the third color is blue, and the fourth color is green. When the identification personnel directly judge with the naked eye that only the green seal appears to be on a white background, the two seals are the same.

The terminal checker 107 is connected to the color marker 106 by telecommunication, and receives the color print information for terminal check, wherein the area of any second color block or the area of any third color block When it exceeds a preset area value, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are the same; or any of the second color blocks When the perimeter or the perimeter of any of the third color blocks exceeds a preset perimeter value, it is determined that the seals of the document A and the seal B are different; otherwise, it is determined that the document A and the seal are different. The seals of the seal file B are the same; or with any point of the seal file B as the center, when the number of points of the second color or the third color within the enclosed range of the outer frame is greater than a predetermined value, That is, it is determined that the seals of the document file A and the seal file B are not the same; otherwise, it is determined that the seals of the document file A and the seal file B are the same. Even if the color printed information has red or blue, it can be judged whether the two are the same through the size and perimeter of the red or blue. Or, using several of the PB points as the center basis, the number of red or blue pixels under the enclosed area surrounding the outer frame is used as a way of judging whether the two seals are consistent. The above various terminal judgment methods can exist alone or be executed in plural combinations to enhance the accuracy verification.

To sum up, the seal detection method and system of the present invention is initially based on a scale-invariant feature conversion algorithm to calculate and extract at least one corresponding point group whose features match each other between the document file A and the seal file B, and then calculate and extract at least one corresponding point group. Judging the number of corresponding point groups for preliminary comparison, then using affine conversion to size consistency for comparison, and then using the correlation coefficient formula and using the image gray value as the correlation analysis and comparison, and finally using the corresponding point group. Based on matching, color print information with four color types is generated for difference analysis. In general, this kind of staged comparison analysis and judgment can greatly improve the inspection efficiency and comparison accuracy.

The above descriptions are only preferred embodiments within the scope of the patent application of the present invention, and should not be used to limit the scope of rights of the present invention according to the content of the embodiments; therefore, the meaning changes made without departing from the equal scope of the present invention or modification, should still be covered within the scope of the patent application of the present invention.

S1~S6: Steps 1: host 10: Nuclear printing detection system 101: Storage 102: Matcher 103: Quantity Comparator 104: Size Comparator 105: Associative aligner 106: Color marker 107: Terminal Checker 2: Display

FIG. 1A is a flow chart (1) of part of the steps of the print detection method according to the preferred embodiment of the present invention. FIG. 1B is a flow chart (2) of part of the steps of the verification method according to the preferred embodiment of the present invention. FIG. 1C is a flow chart (3) of part of the steps of the verification method according to the preferred embodiment of the present invention. FIG. 1D is a flow chart (4) of part of the steps of the verification method according to the preferred embodiment of the present invention. Fig. 1E is a flow chart (5) of part of the steps of the verification method of a preferred embodiment of the present invention. FIG. 2 is a functional block diagram of the stamp detection device according to the preferred embodiment of the present invention.

1: host

10: Nuclear printing detection system

101: Storage

102: Matcher

103: Quantity Comparator

104: Size Comparator

105: Associative aligner

106: Color marker

107: Terminal Checker

2: Display

Claims (10)

A kind of verification seal detection method, is to carry out seal verification for a document file A with a seal on a document, and a seal file B, which is left on a white paper with a seal, wherein the document A and the seal file B are all computer image information, including the following steps: Using a scale-invariant feature transformation algorithm, calculate and extract at least one corresponding point group whose features match each other between the document file A and the signature file B; wherein, the corresponding point group includes (PA, PB), and PA is the file A point on file A, PB is a point on file B of the seal; It is judged whether the number of corresponding point groups is less than a preset basic value, and when the number of corresponding point groups is less than the preset basic value, it is determined that the seals of the document file A and the seal file B are different; the number of corresponding point groups When it is greater than or equal to the basic value, continue to the next step; Use the corresponding point group to calculate the affine transformation to judge whether the size of the seals of the document file A and the seal file B are the same. When the size is the same, continue to the next step; Using the correlation coefficient formula, a correlation r is calculated based on the gray values within the seal range of the document file A and the seal file B. If the correlation r is less than a threshold value, it is determined that the document file A and the seal are The seals of file B are not the same; if the correlation r is greater than or equal to the threshold value, proceed to the next step; and The corresponding point groups (PA, PB) in the document file A and the signature file B are generated according to the characteristics to generate color printing information with one of four color attributes, and the pixel color of the PA point is defined as CA, the pixel color of the PB point is defined as The pixel color is defined as CB, and (1) First color: CA and CB are similar to white; (2) A second color: CA is similar to white, CB is not similar to white; (3) A third color: CA is not similar to white, CB is similar to white; (4) A fourth color: CA and CB are not similar to white; Wherein, when the area of any of the second color blocks or the area of any of the third color blocks exceeds a preset area value, it is determined that the seals of the document file A and the seal file B are different; otherwise, that is It is determined that the seal of the document file A and the seal file B are the same; or when the perimeter of any of the second color blocks or the perimeter of any of the third color blocks exceeds a predetermined perimeter value, it is determined that the The seals of the document file A and the seal file B are different; on the contrary, it is determined that the seals of the document file A and the seal file B are the same; When the number of dots of the second color or the third color is greater than a preset value, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seal of the document file A and the seal file B are different The stamp is the same. The seal detection method as described in claim 1, wherein the step of "using the corresponding point group to calculate an affine transformation to determine whether the size of the seals of the document file A and the seal file B are consistent" is to define the document file The point coordinates of A are (Xa, Ya), the point coordinates of the signature file B are (Xb, Yb), and the linear relationship satisfies

, when (A _00* A ₁₁ )-(A _01* A ₁₀ ) is not equal to 1, it is determined that the size of the document file A and the signature file B are different; when (A _00* A ₁₁ )-(A _01* When A ₁₀ ) is equal to 1, it is determined that the document file A and the signature file B have the same size. The verification seal detection method according to claim 2, wherein the step of "using the correlation coefficient formula to calculate the correlation r with the grayscale values within the seal range of the document file A and the signature file B", is: Define the pixel grayscale of the corresponding point group (PA, PB) as Ga(PA) and Gb(PB), and the grayscale average value MA of the seal range of the document file A, the grayscale average value MB of the seal range of the seal file B, the Relevance

. The verification detection method according to any one of claims 1 to 3, wherein the basic value on which the corresponding point group is compared is a quantity of 3. The checkmark detection method according to claim 4, wherein the first color is white, the second color is red, the third color is blue, and the fourth color is green. A verification seal detection system is used to generate a document file A and a seal file B by scanning a document with a seal and a blank paper left in a prior seal respectively by using an image capture device, and for executing the The seal check of document file A and the seal file B, including: a storage for storing the imported information of the document file A and the signature file B; a matcher, connected to the storage by telecommunication, the matcher performs scale-invariant feature conversion for calculation and extracts information of at least one corresponding point group whose features match each other between the document file A and the signature file B; wherein, The corresponding point group includes (PA, PB), PA is a point on the document file A, and PB is a point on the signature file B; a quantity comparator, telecommunicationly connecting the matcher and the storage, the quantity comparator receives the information of the corresponding point group, and judges whether the quantity of the corresponding point group is less than a preset basic value, when the corresponding point group When the number is less than the preset basic value, it is determined that the seals of the document file A and the seal file B are different; on the contrary, when the number of the corresponding point group is greater than or equal to the basic value, a number of approval information is generated; A size comparator, connected by telecommunication to the quantity comparator and the storage, after receiving the quantity approval information, the size comparator extracts the information of the document file A and the signature file B from the storage, and executes the simulation Judging whether the size of the seals of the document file A and the seal file B is consistent, and when the sizes are inconsistent, it is determined that the seals of the document file A and the seal file B are not the same; otherwise, a size approval information is generated; a correlation comparator, connecting the size comparator and the storage by telecommunication, after receiving the size approval information, the correlation comparator extracts the information of the document file A and the signature file B from the storage, and A correlation r is calculated based on the grayscale values within the seal range of the document file A and the seal file B. If the correlation r is less than a threshold value, it is determined that the seals of the document file A and the seal file B are not identical. The same; on the contrary, if the correlation r is greater than or equal to the threshold value, a correlation approval information is generated; a color marker, connected by telecommunication between the correlation comparator and the storage, after receiving the correlation approval information, the color marker extracts the information of the document file A and the signature file B from the storage, and the document File A and the corresponding point group (PA, PB) in the signature file B generate color print information with one of four color attributes according to the characteristics. The pixel color of the PA point is defined as CA, and the pixel color of the PB point is defined as CB, (1) a first color: CA and CB are both similar to white; (2) a second color: CA is similar to white, CB is not similar to white; (3) a third color: CA is not similar to white , CB and white are similar; (4) a fourth color: CA and CB are not similar to white; and A terminal checker, connected to the color marker by telecommunication, and receives the color print information for terminal check, wherein the area of any second color block or the area of any third color block exceeds one When the area value is preset, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are the same; or the circumference of any of the second color blocks When the length or the perimeter of any of the third color blocks exceeds a preset perimeter value, it is determined that the seals of the document file A and the seal file B are different; otherwise, it is determined that the document file A and the seal file are different. The seals of B are the same; or with any point of the seal file B as the center, the number of points of the second color or the third color within the enclosed range of the outer frame is greater than a predetermined value, it is determined that The seals of the document file A and the seal file B are different; otherwise, it is determined that the seals of the document file A and the seal file B are the same. The seal detection system of claim 6, wherein the size comparator performs affine transformation to define the point coordinates of the document file A as (Xa, Ya), and the point coordinates of the seal file B as (Xa, Ya) is (Xb, Yb), and the linear relationship satisfies

, when (A _00* A ₁₁ )-(A _01* A ₁₀ ) is not equal to 1, it is determined that the size of the document file A and the signature file B are different; when (A _00* A ₁₁ )-(A _01* When A ₁₀ ) is equal to 1, it is determined that the document file A and the signature file B have the same size, and the size approval information is generated. The print detection system as claimed in claim 7, wherein the correlation comparator executes a correlation coefficient formula to define the pixel grayscales of the corresponding point groups (PA, PB) as Ga(PA) and Gb(PB) , and the grayscale mean value MA of the seal range of the document file A, and the grayscale mean value MB of the seal range of the seal document file B, to calculate the correlation

. The stamp detection system according to any one of claims 6 to 8, wherein the base value on which the corresponding point group is compared is a quantity of 3. The stamp detection system of claim 9, wherein the first color is white, the second color is red, the third color is blue, and the fourth color is green.

Images