TWI739387B - Seal identification system and method thereof - Google Patents

Abstract

A seal identification system is provided. By scanning and obtaining a seal image; performing straightening, removal of edge margins, normalization, contrast enhancement, binarization, image erosion and image expansion processes on the seal image to obtain a denoised image; dividing the denoised image and a retained seal image into multiple blocks separately; performing similarity comparison processing on each block of the denoised image and a corresponding block of the retained seal image in order, and then sequentially determining whether each block of the denoised image is similar to the corresponding block of the retained seal image; and outputting a notification message when determining any block of the denoised image is not similar to the corresponding block of the retained seal image, the recognition performance can be improved.

Description

Seal identification system and method

The invention relates to an identification system and a method thereof, in particular to a seal identification system and a method thereof.

Seals are closely related and inseparable to people’s daily lives, whether it is to withdraw cash and settle accounts at financial institutions, or to register or transfer real estate (such as land or houses) at the land administration authority. The seal must be used, so the practicality of the seal is self-evident.

However, in order to ensure that the seal is used correctly, often when people go to the financial institution to withdraw money and settle the account or to the land office to register or transfer the real estate, the staff of the financial institution or the land office needs to compare with the naked eye. Whether the seal is correct. However, manual seal comparisons that are too similar often require other staff to help confirm them, resulting in inefficiency and easy comparison errors. In addition, if the comparison error caused by the identification of forged seals with the naked eye causes the people to steal or transfer the property, it will not only cause customer complaints, but also affect the reputation of financial institutions or land administration agencies. In view of this, it is necessary to propose improved technical means to solve this problem.

The invention discloses a seal identification system and method.

First, the present invention discloses a seal recognition system, which includes: an image scanning module, an image processing module, a binarization module, an image noise reduction module, an image comparison module and a notification module, and the image processing module is connected The scanning module, the binarization module is connected to the image processing module, the image noise reduction module is connected to the binarization module, the image comparison module is connected to the image noise reduction module, and the notification module is connected to the image comparison module. Among them, the image scanning module is used to scan and obtain the seal image; the image processing module is used to correct the seal image, remove white borders, normalize and strengthen the contrast to obtain the seal image; the binarization module is used to correct The stamp image is binarized to obtain a binarized image; the image noise reduction module is used to perform image erosion and image expansion processing on the binarized image to remove noise and obtain a de-noise image; image comparison The module is used to cut the noise-removed image and the stored seal image separately to cut into multiple blocks, and respectively sequentially divide each block of the noise-removed image and each corresponding block of the stored seal image Perform similarity comparison processing, and then sequentially determine whether each block of the denoising image is similar to each corresponding block of the stored seal image; and the notification module is used when the image comparison module determines the denoising When there is no similarity between any block of the message image and the corresponding block of the stored seal image, a notification message is output.

In addition, the present invention discloses a seal identification method, which includes the following steps: scanning and obtaining a seal image; correcting, removing white borders, normalizing and contrasting the seal image to obtain a seal image; performing binarization processing on the seal image To obtain a binary image; perform image erosion and image dilation processing on the binary image to remove noise, and then obtain a noise-removed image; cut the noise-removed image and the stored seal image separately to cut into Multiple blocks; sequentially compare each block of the denoising image with each corresponding block of the saved seal image, and then determine each block of the denoising image and the retention in sequence Whether each corresponding block of the seal image is similar; and when it is determined that any block of the denoising image is not similar to the corresponding block of the remaining seal image, a notification message is output.

The system and method disclosed in the present invention are as above. The difference from the prior art is that the present invention obtains the seal image by scanning; the seal image is sequentially corrected, white border removed, normalized, contrast enhanced, binarized, and image erosion is performed in sequence. And image expansion processing, and then get the denoising image; the denoising image and the saved seal image are cut into multiple blocks separately; each block of the denoising image and the saved seal are sequentially cut Perform similarity comparison processing for each corresponding block of the image, and then determine in order whether each block of the denoising image is similar to each corresponding block of the remaining seal image; and when determining the denoising image When there is no similarity between any block and the corresponding block of the saved seal image, a notification message is output.

Through the above technical means, the present invention can assist manual identification and comparison of suspected seals, and the image processing technology and image segmentation are used to increase the comparison speed, and it is not easy to make mistakes due to human negligence, thereby improving the identification performance.

Before describing the seal identification system and method disclosed in the present invention, firstly, the terms defined by the present invention will be explained. The various modules included in the seal identification system of the present invention can be implemented mainly by hardware, and at the same time Can work with software or firmware. Among them, the software or firmware used in the implementation can be stored on machine-readable storage media, such as: read-only memory (ROM), random access memory (RAM), disk storage media, optical storage media, Flash memory devices, etc., and can be executed by one or more general-purpose or special-purpose programmable microprocessors.

The following describes the implementation of the present invention in detail with the drawings and embodiments, so as to fully understand and implement the implementation process of how the present invention uses technical means to solve technical problems and achieve technical effects.

Please refer to "Figure 1" and "Figure 2", "Figure 1" is a block diagram of an embodiment of the seal recognition system of the present invention, and "Figure 2" is the "Figure 1" seal recognition system performing seal recognition A method flowchart of an embodiment of the method. In this embodiment, the seal identification system 100 can be applied to financial institutions to provide assistance to staff to identify and compare the seals when they withdraw cash and settle accounts at financial institutions. However, this embodiment is not intended to limit the present invention. , Can be used in different situations where seal identification is also required according to actual needs.

In this embodiment, the seal recognition system 100 may include: an image scanning module 110, an image processing module 120, a binarization module 130, an image noise reduction module 140, an image comparison module 150, and a notification module 160 , The image processing module 120 can be connected to the scanning module 110, the binarization module 130 can be connected to the image processing module 120, the image noise reduction module 140 can be connected to the binarization module 130, and the image comparison module 150 can be connected The image noise reduction module 140 and the notification module 160 can be connected to the image comparison module 150. Among them, between the image processing module 120 and the scanning module 110, between the binarization module 130 and the image processing module 120, between the image noise reduction module 140 and the binarization module 130, and the image comparison module The group 150 and the image noise reduction module 140 and the notification module 160 and the image comparison module 150 can transmit images, messages, and data in a wireless or wired manner, and can also communicate with each other through a network, for example: Mobile communication network, Internet, local area network, wide area network and/or wireless network for the transmission of images, messages and/or data.

In this embodiment, the seal identification method may include the following steps: scanning and obtaining a seal image (step 210); performing normalization, removing white borders, normalization and contrast enhancement processing on the seal image to obtain the seal image (step 220); Perform binarization processing on the seal image to obtain a binarized image (step 230); perform image erosion and image dilation processing on the binarized image to remove noise and obtain a de-noise image (step 240); The denoising image and the retained seal image are respectively cut into multiple blocks (step 250); each block of the denoising image is similar to each corresponding block of the retained seal image in sequence In order to determine whether each block of the denoising image is similar to each corresponding block of the stored seal image (step 260); and when it is determined whether any block of the denoising image is similar to each other When the corresponding blocks of the stored seal images are not similar, a notification message is output (step 270).

In step 210, the image scanning module 110 can scan the people to provide their withdrawal slips or settlement applications to financial institutions. The withdrawal slips or settlement applications are stamped with the people’s seal stamp, and then use the characteristics of image processing technology. The capture function captures the characteristics of the seal stamp, establishes the block where the seal seal is located, and extracts the block to obtain the seal image.

N個像素（轉正、去除白邊且正規化的印鑑影像之長度為M個像素大小，寬度為N個像素大小；即轉正、去除白邊且正規化的印鑑影像為M

N個像素所組成的二維陣列），M與N皆為大於1的正整數，可依據實際需求進行調整。 In step 220, after the image scanning module 110 has captured the seal image, since the seal stamp may not be stamped, the image processing module 120 needs to correct the seal image; in addition, the image processing module 120 also needs to correct the seal image. Remove the white borders of the corrected seal image, adjust the length and width of the corrected seal image and remove the white borders to a fixed size (that is, the normalization process described in the present invention), and convert the corrected, white border and normalized seal image The contrast is enhanced, so that the seal image captured by the image scanning module 110 becomes a seal image that has a simple seal seal and can be used for subsequent identification. Among them, the length and width of the stamp image that is normalized, white edges removed, and normalized can be M

N pixels (the length of the stamp image that is normalized, white edges removed, and normalized is M pixels, and the width is N pixels; that is, the stamp image that is normalized, white edges removed, and normalized is M

A two-dimensional array composed of N pixels), M and N are both positive integers greater than 1, which can be adjusted according to actual needs.

Among them, since the seal stamp on the withdrawal slip or settlement application is usually a square stamp, in this embodiment, the normalization process performed by the image processing module 120 in step 220 may include the following steps: ( a) Determine whether the seal frame in the seal image is square; (b) If it is determined that the seal frame in the seal image is square, define the plane coordinate system with the origin of the bottom vertex of the seal frame in the seal image _{; (C) Obtain the projection angle θ 1} of the edge of the first quadrant in the plane coordinate system of the seal frame on the horizontal axis of the plane coordinate system; (d) Determine whether the projection angle θ ₁ is equal to zero; and (e) If it is judged When the projection angle θ _{1 is} not equal to zero, rotate the seal image clockwise by a preset angle according to the rotation command, and repeat steps (c) and (d) until it is determined that the projection angle θ _{1 is} equal to zero. Among them, the preset angle can be but not limited to 1 degree, which can be adjusted according to the actual situation; it should be noted that when the preset angle is too small, the number of repeated steps (c) and (d) will increase, and furthermore Affect the efficiency of the normalization process; when the preset angle is too large, it will cause the projection angle θ ₁ to be difficult to equal zero.

In more detail, the image processing module 120 will first determine whether the seal stamp of the seal image captured by the image scanning module 110 is a square seal (that is, whether the seal frame is a square), and if it determines the seal frame in the seal image When it is a square, because the uncovered square chapter has the bottom vertex on the drawing surface, the image processing module 120 can define a plane coordinate system with the bottom vertex as the origin, and the plane coordinate system has mutually perpendicular Horizontal axis and vertical axis; at this time, a sideline extending from the origin in the square chapter can be located in the first quadrant of the plane coordinate system, and the projection angle of the sideline on the horizontal axis of the plane coordinate system can be θ ₁ ; , The image processing module 120 will determine whether the projection angle θ ₁ is equal to zero (when the projection angle θ _{1 is} equal to zero, it means that the edge is located on the horizontal axis of the plane coordinate system, that is, it does not need to be normalized), when the image processing module 120 determines the projection angle When θ _{1 is} not equal to zero, the seal image can be rotated clockwise by a preset angle according to the rotation command, and then steps (c) and (d) are repeated until it is determined that the projection angle θ _{1 is} equal to zero (that is, the seal image is corrected).

In addition, since the seal stamp on the withdrawal slip or settlement application may not be a square stamp, in another embodiment, the normalization process performed by the image processing module 120 in step 220 may also include the following steps : Obtain the uncorrected text in the seal image through the method of image recognition; obtain the seal stamp of the seal card reserved by the financial institution for the people who want to perform the seal recognition comparison; and based on the obtained seal stamp The uncorrected text in the seal and the seal image determines the rotation angle required for the seal image to be corrected, and then the seal image is corrected according to the rotation angle. Among them, the seal recognition system 100 is used to assist manual identification and comparison of suspected seals (that is, the seal image is highly similar to the stored seal image and cannot be accurately judged by the naked eye, so the seal recognition system 100 is used to assist in the identification), therefore, In the step of obtaining the text in the seal image through the method of image recognition, the image processing technology can be assisted based on the seal on the seal stamp of the seal card and its arrangement to determine the font of the text in the seal image, and then the seal image can be obtained Uncorrected text in.

In this embodiment, the white fringing process performed by the image processing module 120 in step 220 may include the following steps: determining whether any row or column of pixels in the corrected seal image is completely white; and if it is determined in the corrected seal image Any row or column of pixels is completely white, and that row or column that is completely white is removed.

N個像素所組成的二維陣列）。 In this embodiment, since the seal stamp in the seal image that is normalized, white edges removed, and normalized may be darker in some areas and lighter in some areas due to the uniform force applied during stamping or other factors, , The image processing module 120 needs to dilute the color of the darker area and deepen the color of the lighter area. Therefore, the contrast enhancement processing performed by the image processing module 120 in step 220 may include the following steps: using histogram adjustment ( Histogram Modification) method, by changing the grayscale value and distribution of the pixels of the stamp image that is normalized, removed the white border and normalized, and then achieves the effect of image enhancement. Furthermore, the grayscale distribution map (which is a discrete function) of the stamp image that is normalized, white fringed removed, and normalized can be based on the distribution map equalization method (Histogram equalization) by using the discrete probability density function value of each grayscale value Perform the conversion to obtain a uniformly distributed grayscale distribution map. It should be noted that the length and width of the seal image is the same as the length and width of the stamp image that is normalized, white border removed, and normalized (that is, the length of the seal image is M pixels in size, and the width is N pixels in size; in other words, the seal image is M

A two-dimensional array composed of N pixels).

In step 230, the binarization module 130 may perform binarization processing on the seal image to obtain a binarized image. In more detail, the binarization module 130 can convert the stamp image obtained by the image processing module 120 into a grayscale image, so that the grayscale value of each pixel in the grayscale image falls between 0 and 255. A level value of 0 indicates that the pixel is black, and a grayscale value of 255 indicates that the pixel is white; the binarization module 130 may further obtain a threshold value according to the image histogram corresponding to the grayscale image, and then Use any pixel in the above grayscale image to convert to 1 when its grayscale value is greater than the threshold, and convert any pixel to 0 when its grayscale value is less than or equal to the threshold (that is, use the threshold to convert the above Grayscale image), and then generate a two-dimensional array composed of 1 and 0 (that is, to generate a binary image), where 1 in the two-dimensional array represents white and 0 represents black.

In step 240, since the seal stamp may be blurred (ie, noise is generated) when the seal is used for too long, the image noise reduction module 140 is required to perform image erosion and image expansion on the binary image. Processing to remove the noise, and then get a de-noise image. Among them, the image erosion and image dilation processing system can interact with multiple dilation operations and erosion operations, so that the results of the operations can fill in small holes, smooth boundaries, or break some lines in the binary image. Connect, and then get the noise-removed image.

In step 250 and step 260, the image comparison module 150 starts to compare the similarity between the noise-removed image and the saved seal image (that is, performs seal recognition and comparison), where the saved seal image is for the public in the financial institution When applying for opening an account, it is printed on the seal card, and the financial institution uses the seal recognition system 100 to execute the image obtained from the above steps 210 to 240.

In more detail, when the image comparison module 150 starts to perform the seal recognition and comparison, it needs to cut the denoising image and the saved seal image respectively, for example: cut the denoising image into three blocks, and save the seal. The image is also cut into three blocks, so that each block of the denoising image has the same area as each corresponding block of the remaining seal image (step 250). In this embodiment, the number of blocks cut by the denoising image and the saved seal image can be, but is not limited to, the number of words in the printed text in the saved seal image (ie, the number of people's names).

，P為去雜訊影像的每一區塊與留存印章影像的每一對應區塊之間的相似度值，R與Q分別為去雜訊影像的任一區塊與留存印章影像的對應區塊中每一像素值的總和。最後，影像比對模組150可將計算出來的每一相似度值與一第一預設值進行大小比較，若計算出來的任一相似度值小於該第一預設值時，則判斷用以計算該相似度值的去雜訊影像的該區塊與留存印章影像的該對應區塊之間不相似；若計算出來的任一相似度值大於或等於該第一預設值時，則判斷用以計算該相似度值的去雜訊影像的該區塊與留存印章影像的該對應區塊之間相似；在本實施例中，第一預設值可為但不限於90％，可依據實際需求進行調整。 Then, the image comparison module 150 can sequentially calculate the similarity value between each block of the noise image and each corresponding block of the stored seal image based on the similarity formula (that is, obtained through the distribution of pixels). Similarity value), and based on the calculated similarity value, determine whether each block of the denoising image is similar to each corresponding block of the remaining seal image. The similarity formula is:

, P is the similarity value between each block of the denoising image and each corresponding block of the stored seal image, R and Q are respectively any block of the denoising image and the corresponding area of the stored seal image The sum of the value of each pixel in the block. Finally, the image comparison module 150 can compare each calculated similarity value with a first preset value. If any calculated similarity value is less than the first preset value, the judgment is The block of the denoising image for calculating the similarity value is not similar to the corresponding block of the stored seal image; if any calculated similarity value is greater than or equal to the first preset value, then It is determined that the block of the denoising image used to calculate the similarity value is similar to the corresponding block of the stored seal image; in this embodiment, the first preset value may be, but not limited to, 90%. Adjust according to actual needs.

）），接著，若影像比對模組150判斷P ₁小於90％時，則判斷區塊J ₁與區塊K ₁之間不相似，反之，判斷區塊J ₁與區塊K ₁之間相似；若影像比對模組150判斷區塊J ₁與區塊K ₁之間相似，則接著計算區塊J ₂與區塊K ₂之間的相似度值（需先計算區塊J ₂中每一像素值的總和R ₂以及區塊K ₂中每一像素值的總和Q ₂，再進行相似度P ₂之計算（

）），並判斷P ₂是否小於90％，若影像比對模組150判斷P ₂小於90％時，則判斷區塊J ₂與區塊K ₂之間不相似，反之，判斷區塊J ₂與區塊K ₂之間相似；若影像比對模組150判斷區塊J ₂與區塊K ₂之間相似，則接著計算區塊J ₃與區塊K ₃之間的相似度值（需先計算區塊J ₃中每一像素值的總和R ₃以及區塊K ₃中每一像素值的總和Q ₃，再進行相似度P ₃之計算（

）），並判斷P ₃是否小於90％，若影像比對模組150判斷P ₃小於90％時，則判斷區塊J ₃與區塊K ₃之間不相似，反之，判斷區塊J ₃與區塊K ₃之間相似。 For example, when the de-noise image is cut into three blocks (block J ₁ , block J ₂ and block J ₃ ) and the remaining seal image is also cut into three blocks (block K ₁ , block J 3) K ₂ and block K ₃ ), block J ₁ can correspond to block K ₁ and the area of block J ₁ and block K _{1 are} the same (that is, the number of pixels in _{block J 1} and block K _{1 is the same),} Block J ₂ can correspond to block K ₂ and the areas of block J ₂ and block K _{2 are} the same (that is, _{the number of pixels in block J 2} and block K _{2 is} the same), block J ₃ can correspond to block K ₃ And the area of the block J ₃ and the block K _{3 are} the same (that is, the number of pixels in the _{block J 3} and the block K _{3 is the same).} The image comparison module 150 can first calculate _{the similarity value between the block J 1} and the block K ₁ (it is necessary to first calculate _{the sum R 1} of each pixel value in the _{block J 1} and each pixel in the block K ₁ The sum of the values Q ₁ , since the denoising image and the stored seal image are a two-dimensional array of 1 and 0 (1 represents white and 0 represents black), when any pixel is white, its pixel value is 1. When any pixel is black, its pixel value is 0), and then calculate the similarity P ₁ (

)), Then, if the image comparison module 150 determines that P ₁ is less than 90%, it is determined that the block ₁ and the block J K ₁ between the dissimilar, and vice versa, and the block ₁ determines the block J K between ₁ Similar; if the image comparison module 150 determines that the block J ₁ and the block K ₁ are similar, then calculate _{the similarity value between the block J 2} and the block K ₂ (need to calculate the block J _{2 first} The sum R ₂ of each pixel value and the sum Q ₂ of each pixel value in the block K ₂ are then calculated for the similarity P _{2 (}

)) and judge whether P ₂ is less than 90%. If the image comparison module 150 judges that P _{2 is} less than 90%, it judges that the block J ₂ and the block K ₂ are not similar, otherwise, judges the block J ₂ It is similar to the block K ₂ ; if the image comparison module 150 determines that the block J ₂ and the block K ₂ are similar, it then calculates the similarity value between the _{block J 3} and the block K _{3 (need} first calculate the sum of the R value of each pixel block J _{3 3} and the sum of each pixel value of the block K ₃ Q _3, P ₃ then calculates the similarity of (

)) and judge whether P ₃ is less than 90%. If the image comparison module 150 judges that P _{3 is} less than 90%, it judges that the block J ₃ and the block K ₃ are not similar, otherwise, judges the block J ₃ similar to the block K _3.

In step 270, since the image comparison module 150 in step 260 can first calculate P ₁ , and determine whether the block J ₁ and the block K ₁ are similar based on P ₁ ; if the image comparison module 150 determines the area If the block J ₁ and the block K ₁ are similar, then P ₂ is calculated, and based on P ₂ it is judged whether the block J ₂ and the block K ₂ are similar; if the image comparison module 150 judges the block J ₂ Is similar to the block K ₂ , then P ₃ is calculated, and based on P _3, it is determined whether the block J ₃ and the block K ₃ are similar. Therefore, once the image comparison module 150 determines the block J between ₁ and K ₁ dissimilar block, the block ₂ and block J K J dissimilar blocks or dissimilarity between ₃ and K3 between ₂ blocks, the image comparison module 150 may identify the end seal Comparison; at this time, the notification module 160 can output a notification message (the content of which can be, but is not limited to, "seal does not match") to notify the staff operating the seal identification system 100. Since the image comparison module 150 determines that any block of the denoising image is not similar to the corresponding block of the stored seal image, it will notify the staff operating the seal recognition system 100, and there is no need to fully identify the denoising image. Whether all the blocks are similar to all the corresponding blocks of the saved seal image, therefore, the identification efficiency can be improved.

It should be noted that when the image comparison module 150 determines that the block J ₁ and the block K ₁ are similar, the block J ₂ and the block K ₂ are similar, and the block J ₃ and the block K3 are similar. Similarly, the notification module 160 can output the pass information (the content of which can be, but is not limited to, "the seal is correct") to notify the staff operating the seal identification system 100.

In the above step 260, the image comparison module 150 sequentially calculates the similarity value between each block of the noise image and each corresponding block of the stored seal image based on the similarity formula, and based on the calculation The obtained similarity value determines whether each block of the denoising image is similar to each corresponding block of the stored seal image, but this embodiment is not intended to limit the present invention. For example, in another embodiment, the image comparison module 150 can capture at least one feature point of each block of the denoising image and each corresponding block of the stored seal image, and compare them in sequence The distribution of at least one feature point between each block of the denoising image and each corresponding block of the remaining seal image, and then determining each block of the denoising image and each corresponding block of the remaining seal image in sequence Are they similar. Among them, the characteristic points can be, but are not limited to, the end points of each character, the transition point of the stroke and the intersection of the stroke.

For example, when the de-noise image is cut into three blocks (block J ₁ , block J ₂ and block J ₃ ) and the remaining seal image is also cut into three blocks (block K ₁ , block J 3) when K ₂ K _{block. 3), a} block may correspond to block J and K ₁ J ₁ block area of the same block of K _1, a block may correspond to block J ₂ and K ₂ and J ₂ block region Block K _{2 has} the same area, block J ₃ can correspond to block K _3, and block J ₃ and block K _{3 have} the same area. Image comparison module 150 may retrieve the first block and the block J _a feature point ₁ K, J and matching block ₁ and block number of feature points ₁ K, 150 determines if the image matching area module When the _{number of feature points of the block J 1} and the block K ₁ is the same, it is judged that the block J ₁ and the block K ₁ are similar; otherwise, it is judged that the block J ₁ and the block K ₁ are not similar; if the image ratio is ₁ and the similarity between the K block module 150 determines _a block J, and then retrieve the feature point block ₂ and block ₂ J K, wherein matching block ₂ and block J and K ₂ If the image comparison module 150 determines that the _{number of feature points of the block J 2} and the block K ₂ are the same, it will determine that the block J ₂ and the block K ₂ are similar. Otherwise, it will determine that the block J _{2 is} similar to the The blocks K ₂ are not similar; if the image comparison module 150 judges that the blocks J ₂ and K ₂ are similar, then it then extracts the feature points in the _{block J 3} and the block K _{3 and compares them.} Regarding the number of feature points of the block J ₃ and the block K ₃ , if the image comparison module 150 determines that the number of feature points of the block J ₃ and the block K ₃ are the same, it will determine the block J ₃ and the block K ₃ Otherwise, it is judged that the block J ₃ and the block K ₃ are not similar.

In addition, in one embodiment, the image comparison module 150 can perform a thinning process on each block of the denoising image and each corresponding block of the stored seal image to capture the denoising image. Multiple feature vectors of each block and each corresponding block of the stored seal image; calculate each feature of each block of the noise image and each corresponding block of the stored seal image based on the inner product formula The vector included angle between the vectors, and the average vector included angle between each block of the denoising image and each corresponding block of the stored seal image is obtained in sequence; and the difference between each block and each block based on the denoising image The average vector included angle between each corresponding block of the stored seal image sequentially determines whether each block of the denoising image is similar to each corresponding block of the stored seal image.

For example, when the de-noise image is cut into three blocks (block J ₁ , block J ₂ and block J ₃ ) and the remaining seal image is also cut into three blocks (block K ₁ , block J 3) when the block K and K _{2. 3), a} block may correspond to block J and K ₁ J ₁ block area of the same block of K _1, a block may correspond to block J ₂ and K ₂ and J ₂ blocks The area of the block K _{2 is} the same, the block J ₃ can correspond to the block K _3, and the area of the block J ₃ and the block K _{3 are the same.} The image comparison module 150 can _{thin each text in the block J 1} and the block K _{1 first} , and convert the coordinates of the endpoints and intersections of the thinned lines into multiple vectors, so that each line segment forms a feature vector; Then, the image comparison module 150 can use the _{inner product formula of the corresponding feature vectors of the block J 1} and the block K _{1 to} obtain the vector included angle of the feature vector, and average all the vector included angles to obtain the average vector included angle ( For example: block J ₁ and block K ₁ each have five feature vectors, so five vector included angles can be calculated, and the average vector included angle can be obtained by averaging the five vector included angles), when the image comparison module 150. When it is judged that the average vector included angle is less than the second preset value (which can be but not limited to 0 degrees to 5 degrees), it is judged that the block J ₁ and the block K ₁ are similar. Otherwise, it is judged that the block J ₁ and the zone Block K ₁ is not similar; if the image comparison module 150 determines that the block J ₁ and the block K ₁ are similar, then _{each text in the block J 2} and the block K ₂ is thinned, and then thinned. The endpoints and intersection coordinates of the lines are converted into multiple vectors, so that each line segment forms a feature vector; then, the image comparison module 150 can inner product the corresponding feature vectors of the _{block J 2} and the block K ₂ The formula obtains the vector included angle of the feature vector, and averages all the vector included angles to obtain the average vector included angle. When the image comparison module 150 determines that the average vector included angle is less than the second preset value (which can be, but is not limited to, 0 degrees to 5 degrees), it is judged that the block J ₂ and the block K ₂ are similar, otherwise, it is judged that the block J ₂ and the block K ₂ are not similar; if the image comparison module 150 judges that the block J _{2 is} similar to the Block K ₂ is similar, then _{each text in block J 3} and block K ₃ is thinned, and the endpoints and intersection coordinates of the thinned lines are converted into multiple vectors, so that each line segment forms a feature Vector; then, the image comparison module 150 can use the _{inner product formula of the feature vectors corresponding to the block J 3} and the block K _{3 to} obtain the vector included angle of the feature vector, and average the included angles of all vectors to obtain the average vector The included angle: when the image comparison module 150 determines that the average vector included angle is less than the second preset value (which can be but is not limited to 0 degrees to 5 degrees), it is determined that the block J ₃ and the block K ₃ are similar, and vice versa , It is determined that the block J ₃ and the block K ₃ are not similar.

To sum up, it can be seen that the difference between the present invention and the prior art is that the seal image is obtained through scanning; the seal image is sequentially corrected, white edges removed, normalized, contrast enhanced, binarized, image erosion and image expansion are performed. Processing to obtain the denoising image; the denoising image and the saved seal image are respectively subjected to a cutting process to cut into multiple blocks; each block of the denoising image and each remaining seal image are sequentially Perform similarity comparison processing on a corresponding block, and then determine in order whether each block of the denoising image is similar to each corresponding block of the remaining seal image; and when determining any area of the denoising image When the block and the corresponding block of the saved seal image are not similar, the notification message is output. This technical method can solve the problems of the previous technology, and then can assist the manual identification and comparison of the suspected seal, and by the image Processing technology and image segmentation processing speed up the comparison, and it is not easy to make mistakes due to human negligence, thereby improving the recognition performance.

Although the present invention is disclosed in the foregoing embodiments as above, it is not intended to limit the present invention. Anyone familiar with similar art can make some changes and modifications without departing from the spirit and scope of the present invention. Therefore, the present invention The scope of patent protection shall be subject to the definition of the scope of patent application attached to this specification.

100: Seal Identification System 110: Image scanning module 120: image processing module 130: Binarization module 140: Image noise reduction module 150: Image comparison module 160: Notification Module Step 210: Scan and obtain the seal image Step 220: Perform correction, removal of white edges, normalization and contrast enhancement on the seal image to obtain the seal image Step 230: Binarize the stamp image to obtain a binary image Step 240: Perform image erosion and image expansion processing on the binarized image to remove noise, thereby obtaining a de-noise image Step 250: Cut the noise-removed image and the stored seal image separately to cut into multiple blocks Step 260: Perform similarity comparison processing on each block of the denoising image and each corresponding block of the stored seal image in sequence, and then determine each block of the denoising image and the stored seal image in sequence Whether each corresponding block of is similar Step 270: When it is determined that any block of the denoising image is not similar to the corresponding block of the stored seal image, output a notification message

Figure 1 is a block diagram of an embodiment of the seal verification system of the present invention. Fig. 2 is a flowchart of an embodiment of the seal recognition method executed by the seal recognition system of Fig. 1.

100: Seal Identification System

110: Image scanning module

120: image processing module

130: Binarization module

140: Image noise reduction module

150: Image comparison module

160: Notification Module

Claims (8)

A seal recognition system, comprising: an image scanning module for scanning and obtaining a seal image; an image processing module connected to the scanning module for correcting, removing white edges, and normalizing the seal image And contrast enhancement processing to obtain a stamp image; a binarization module connected to the image processing module to perform a binarization process on the stamp image to obtain a binary image; an image noise reduction The module is connected to the binarization module to perform image erosion and image expansion processing on the binarized image to remove noise and obtain a denoising image; an image comparison module is connected to the image The noise reduction module is used to perform a cutting process on the denoising image and a stored stamp image respectively to cut into multiple blocks, and calculate each of the denoising images in sequence based on a similarity formula. A similarity value between a block and each corresponding block of the stored seal image, and based on the similarity value, each block of the denoising image and each block of the stored seal image are sequentially determined Whether the corresponding blocks are similar, where the similarity formula is:

, P is the similarity value between each block of the denoising image and each corresponding block of the saved seal image, R and Q are respectively any block of the denoising image and the saved The sum of each pixel value in the corresponding block of the seal image; and a notification module connected to the image comparison module for determining whether any block of the denoising image is the same as the image comparison module. When the corresponding blocks of the stored seal image are not similar, a notification message is output. The seal recognition system according to claim 1, wherein the image comparison module captures at least one feature point of each block of the denoising image and each corresponding block of the stored seal image, and according to The distribution of the at least one characteristic point between each block of the denoising image and each corresponding block of the saved seal image is sequentially compared, and then each block of the denoising image and the retained stamp image are sequentially determined Whether each corresponding block of the seal image is similar. The seal recognition system according to claim 1, wherein the image comparison module performs thinning processing on each block of the denoising image and each corresponding block of the stored seal image to capture Take multiple feature vectors of each block of the denoising image and each corresponding block of the retained seal image; calculate each block of the denoising image and the retained seal image based on an inner product formula A vector angle between each of the feature vectors in each corresponding block of each corresponding block, and obtain the difference between each block of the denoising image and each corresponding block of the remaining seal image in sequence An average vector included angle; and based on the average vector included angle between each block of the denoising image and each corresponding block of the stored seal image, it is determined in sequence that each block of the denoising image is Whether each corresponding block of the saved seal image is similar. The seal recognition system according to claim 1, wherein the conversion process includes the following steps: (a) Determine whether the seal frame in the seal image is a square; (b) If it is determined whether the seal frame in the seal image is square When it is square, define a plane coordinate system with a bottom vertex of the seal frame in the seal image as the origin; (c) Obtain the side line of the seal frame located in the first quadrant of the plane coordinate system. _{A projection angle θ 1} on a horizontal axis of the plane coordinate system; (d) judge whether the projection angle θ ₁ is equal to zero; and (e) if it is judged that the projection angle θ _{1 is} not equal to zero, rotate clockwise according to a rotation command The seal image has a preset angle, and steps (c) and (d) are repeated until it is determined that the projection angle θ _{1 is} equal to zero. A seal identification method includes the following steps: (A) scan and obtain a seal image; (B) perform correction, removal of white borders, normalization and contrast enhancement on the seal image to obtain a seal image; (C) Perform a binarization process on the seal image to obtain a binarized image; (D) perform image erosion and image expansion processes on the binarized image to remove noise and obtain a de-noise image; (E) ) Perform a cutting process on the denoising image and a stored seal image respectively to cut into multiple blocks; (F) Calculate each block and the denoising image in sequence based on a similarity formula A similarity value between each corresponding block of the retained seal image, and then based on the similarity value, it is determined in sequence between each block of the denoising image and each corresponding block of the retained seal image Whether it is similar, where the similarity formula is:

, P is the similarity value between each block of the denoising image and each corresponding block of the saved seal image, R and Q are respectively any block of the denoising image and the saved The sum of each pixel value in the corresponding block of the seal image; and (G) when it is determined that any block of the denoising image is not similar to the corresponding block of the stored seal image, output a notification message. The seal identification method according to claim 5, wherein the step (F) includes the following steps: capturing at least one characteristic point of each block of the denoising image and each corresponding block of the stored seal image ; And sequentially compare the distribution of the at least one feature point of each block of the denoising image with each corresponding block of the remaining seal image, and then sequentially determine each block of the denoising image Whether it is similar to each corresponding block of the saved seal image. The seal identification method according to claim 5, wherein the step (F) includes the following steps: each block of the denoising image and each corresponding block of the remaining seal image are respectively thinned, The feature vectors of each block of the denoising image and each corresponding block of the stored seal image are respectively captured; each block of the denoising image and the corresponding block are calculated based on an inner product formula The relationship between each corresponding feature vector in each corresponding block of the seal image Measure the included angle, and sequentially obtain an average vector included angle between each block of the denoising image and each corresponding block of the saved seal image; and the difference between each block based on the denoising image The average vector included angle between each corresponding block of the stored seal image determines in order whether each block of the denoising image is similar to each corresponding block of the stored seal image. The seal identification method according to claim 5, wherein the conversion processing includes the following steps: (a) determining whether the seal frame in the seal image is a square; (b) if it is determined whether the seal frame in the seal image is square When it is square, define a plane coordinate system with a bottom vertex of the seal frame in the seal image as the origin; (c) Obtain the side line of the seal frame located in the first quadrant of the plane coordinate system. _{A projection angle θ 1} on a horizontal axis of the plane coordinate system; (d) judge whether the projection angle θ ₁ is equal to zero; and (e) if it is judged that the projection angle θ _{1 is} not equal to zero, rotate clockwise according to a rotation command The seal image has a preset angle, and steps (c) and (d) are repeated until it is determined that the projection angle θ _{1 is} equal to zero.

Images