A scanning recognition method and device for automatic identification of question numbers

Abstract

The present application provides a scanning recognition method and device for automatic identification of question numbers, and relates to the field of automated marking. The manufacturing method and device include: establishing a recognition template library of digital characters 0-9, and saving the template feature components of each digital character; scanning the blank answer sheet of the examinee to obtain the examinee answer sheet image, and storing the above-mentioned examinee answer sheet image; gray-processing the above-mentioned examinee answer sheet image, converting it into a grayscale image, and then binarizing the grayscale image; using a rectangular frame to select continuous objective question information in the above-mentioned examinee answer sheet image; cutting the image in the above-mentioned rectangular frame, and identifying the category and position of each element in the image in the above-mentioned rectangular frame, and identifying the values ​​of all question number elements; according to the values ​​of the above-mentioned identified question number elements, determining the arrangement direction of the question number. The scanning recognition method and device do not require a lot of manual intervention, and can greatly improve work efficiency and accuracy.

Description

Technical Field

[0001] The present invention relates to the field of automated examination paper marking, and in particular to a scanning recognition method and device for automatically identifying question numbers. Background Art

[0002] On the answer sheet currently used in the exam, there are generally two ways to sort the objective question numbers: horizontal sorting and vertical sorting. The sorting method of the objective question numbers is not fixed and unified, and there are two different sorting methods. In order to adapt to this situation, when making a scanning template in the current scanning acquisition system, the examination staff needs to use the naked eye to manually judge whether the sorting direction of the objective question numbers on the answer sheet is horizontal or vertical. According to the sorting direction of the objective question numbers, the examination staff makes the corresponding scanning template. This method of manually judging the sorting method of the objective question numbers by the naked eye and making the corresponding scanning template is not efficient, relies on manual judgment and inspection, and is prone to human errors and problems.

[0003] Based on the above situation, how to quickly and accurately determine the order of objective questions is a problem that needs to be solved in scanning and acquisition technology. Summary of the invention

[0004] The present application provides a scanning and recognition method for automatic identification of question numbers, comprising: establishing a recognition template library of digital characters 0-9, and storing the template feature components of each digital character; scanning a blank answer sheet of a candidate to obtain a candidate answer sheet image, and storing the above-mentioned candidate answer sheet image; gray-processing the above-mentioned candidate answer sheet image, converting it into a gray-scale image, and then binarizing the gray-scale image; selecting continuous objective question information in the above-mentioned candidate answer sheet image with a rectangular frame; cropping the image within the above-mentioned rectangular frame, and identifying the category and position of each element in the image within the above-mentioned rectangular frame, and identifying the values ​​of all question number elements; determining the arrangement direction of the question number according to the values ​​of the above-mentioned identified question number elements.

[0005] In one embodiment of the present application, the establishment of a recognition template library for digital characters 0-9 and the storage of template feature components of each digital character include: graying the image of the printed digital font character of the digital character, converting it into a grayscale image, and then binarizing the grayscale image, counting the pixel points whose grayscale values ​​meet the preset conditions as 1, and counting them as 0 otherwise; segmenting the binary digital image, equally dividing it in the horizontal and vertical directions, dividing it into c equal parts in the vertical direction and d equal parts in the horizontal direction, and dividing it into c*d equal-sized rectangular images Image, c and d are both integers greater than or equal to 2; according to the segmented digital image, calculate the proportion of the pixel value 1 in the digital image in the rectangular images segmented into parts, and record it as the feature component X = (X1, X2, X3, ..., Xn) of the digital image, Xn represents the proportion of the pixel value 1 in one of the rectangular images, and n is equal to c*d; sample training is performed on the ten digital characters 0-9 respectively, the average feature component of each sample of each number is calculated, and the average feature component of each number is used as the template feature component of each digital character.

[0006] In one embodiment of the present application, the objective question information includes: the question number and the rectangular filling point on the right side of the question number.

[0007] In one embodiment of the present application, the category and position of each element in the image are identified, and the category and position of each element in the image are identified through digital contours and filled point contours.

[0008] In one embodiment of the present application, the category and position of each element in the image are identified, and the category, quantity and position of each element are determined by the spacing between adjacent question numbers above, below, left and right, or the spacing between adjacent filling points in the same question.

[0009] In one embodiment of the present application, the numerical values ​​of all question number elements are identified by comparing the distance between the feature vector of the rectangular image of the question number element to be identified and the template feature component of each number from 0 to 9. The smaller the distance, the higher the similarity.

[0010] In one embodiment of the present application, determining the arrangement direction of the question numbers based on the identified question number values ​​includes: locating the first question number in the upper left corner of the candidate's answer sheet, comparing the values ​​to the right of the first question number and the values ​​below the first question number, if the value to the right is smaller, the question numbers are arranged horizontally, and if the value below is smaller, the question numbers are arranged vertically.

[0011] In one embodiment of the present application, the scanning and recognition method also includes saving the category, quantity and position of each element in the image within the above-mentioned rectangular frame, the numerical values ​​of all the above-mentioned identified question number elements, and the arrangement direction of the above-mentioned identified question numbers into the scanning template information.

[0012] The present application discloses a scanning and recognition device for automatic identification of question numbers, comprising: a high-speed scanner for scanning a candidate's answer sheet to obtain an image of the candidate's answer sheet; an image processing module connected to the high-speed scanner for binarizing the image of the candidate's answer sheet; a title selection module connected to the image processing module for selecting continuous objective question information in the image of the candidate's answer sheet with a rectangular frame; an image recognition module connected to the title selection module for cropping the image selected by the rectangular frame, and identifying the category, quantity and position of each element in the image, and identifying the numerical value of all question number elements; a digital character template library connected to the image recognition module for storing a recognition template library of digital characters 0-9; a question number arrangement module connected to the image recognition module for determining the arrangement direction of the question number according to the identified question number numerical value.

[0013] In one embodiment of the present application, the objective question information includes: a question number and a rectangular filling point on the right side of the question number.

[0014] In one embodiment of the present application, the image recognition module is used to identify the category and position of each element in the image, and can identify the category and position of each element (question number and fill point) in the image through digital outlines and fill point outlines.

[0015] In one embodiment of the present application, the category and position of each element in the above-mentioned recognition image are determined by the spacing between adjacent question numbers in the upper, lower, left, and right directions, or the spacing between adjacent filling points in the same question.

[0016] In one embodiment of the present application, the scanning and recognition device for automatic question number recognition further includes a scanning template module, which is connected to the question number arrangement module and is used to store scanning template information.

[0017] In one embodiment of the present application, the scanning template information includes: the category, quantity and position of each element, the numerical values ​​of all the identified elements with question numbers, and the arrangement direction of the identified question numbers.

[0018] The present invention provides a scanning recognition method and device for automatic identification of question numbers. By collecting the answer sheets of examinees in advance, based on the establishment of a digital character template library and the selection of digital symbol features, the printed digital symbol recognition technology is used to automatically identify the question numbers of each objective question, thereby determining the sorting method of the objective questions. It does not require a lot of manual intervention, and can greatly improve work efficiency and accuracy, effectively ensuring the seriousness, authority, and fairness of the examination, and also ensuring the smooth progress of the scanning work.

[0019] In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. As shown in the accompanying drawings, the above and other purposes, features and advantages of the present invention will be clearer. The same reference numerals in all the drawings indicate the same parts. The drawings are not deliberately scaled to the actual size, and the focus is on illustrating the main purpose of the present invention.

[0021] Figure 1 This is a diagram of a student answer sheet.

[0022] Figure 2 It is a schematic diagram of a scanning and recognition device for automatic identification of question numbers provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0025] The present application discloses a scanning and recognition method for automatic identification of question numbers, and an embodiment of the present application specifically includes the following steps.

[0026] Step S100, establishing a recognition template library for numeric characters 0-9, and storing the template feature components of each numeric character.

[0027] In one embodiment of the present application, establishing a recognition template library of digital characters 0-9 may specifically include several sub-steps:

[0028] Sub-step S101, digital graphics: grayscale the image of the printed digital font characters of the digital characters, convert it into a grayscale image, and then binarize the grayscale image. The pixel points whose grayscale values ​​meet the preset conditions are counted as 1, otherwise they are counted as 0.

[0029] In one embodiment of the present application, the binarization of the grayscale image can be specifically implemented by the im2bw(f,b) function, where f is the grayscale value of the input image pixel, and b is the set threshold value. The pixels in the image are divided into two categories according to the grayscale value: greater than or equal to b and less than b. B can be 128, for example. The pixel points of the output binary image with a value of 0 correspond to the pixel points whose grayscale value f of the input image is less than the threshold value b. The pixel points of the output binary image with a value of 1 correspond to the pixel points whose grayscale value f of the input image is greater than or equal to the threshold value b. The title text area is represented by the value 1 (white area), and the background area is represented by the value 0 (black area).

[0030] Sub-step S102, digital image segmentation: segment the binary digital image, equally divide it in the horizontal and vertical directions, divide it into c equal parts in the vertical direction, and d equal parts in the horizontal direction, dividing it into c*d rectangular images of equal size, where c and d are both integers greater than or equal to 2.

[0031] In an embodiment of the present application, the c and d parameter values ​​can be adjusted according to the business situation. Generally, the c value is 2-4 and the d value is 2. The larger the c and d values ​​are, the higher the probability of correctly recognizing the character. Assuming c=2 and d=2, the digital image is divided into 4 (2*2) rectangles.

[0032] Sub-step S103, feature selection: based on the segmented digital image, calculate the proportion of the pixel value 1 in the digital image in each partially segmented rectangular image, and record it as the feature component X = (X1, X2, X3, ..., Xn) of the digital image, where Xn represents the proportion of the pixel value 1 in one of the rectangular images, and n is equal to c*d.

[0033] In one embodiment of the present application, the number of "1"s in each rectangular image into which the digital image is divided is used as the characteristic component of the digital image. Each rectangular image is counted from left to right and from top to bottom, and the values ​​are recorded in sequence.

[0034] Sub-step S104, sample preprocessing: perform sample training on the ten numeric characters 0-9 respectively, calculate the average feature component of each sample of each number, and use the average feature component of each number as the template feature component of each numeric character.

[0035] In one embodiment of the present application, sample training is performed on the ten digits 0-9 respectively, that is, according to the above sub-steps S101, S102, and S103, so as to calculate the average feature component of each sample of each digit.

[0036] In one embodiment of the present application, sub-step S104 can be specifically, for example, to divide the samples corresponding to the ten numbers 0-9 into 10 categories: N1, N2, N3, ..., N10, and train the samples of each category respectively, and the number of samples of each category is preferably more than 30. According to the above sub-steps S101-S104, the feature vector of each sample image is obtained, and the average value of the feature vector of each sample of each category is calculated. According to the previous digital image samples, the regularity of each digital character in the feature space is roughly outlined to provide parameter information for further calculations later.

[0037] Step S200, scanning the blank answer sheet of the examinee to obtain the examinee answer sheet image, and storing the examinee answer sheet image.

[0038] Step S300, grayscale processing is performed on the title image of the examinee's answer sheet to convert it into a grayscale image, and then the grayscale image is binarized.

[0039] In this step, the grayscale processing and binarization processing can refer to the above sub-step S101.

[0040] Step S400, using a rectangular frame to select continuous objective question information in the examinee's answer sheet image.

[0041] In one embodiment of the present application, Figure 1 As shown, the objective question information includes at least the question number and the rectangular filling points on the right side of the question number. The number of rectangular filling points is generally 4, and the number of rectangular filling points for individual subjects and individual questions may be as high as 8.

[0042] Step S500, cropping the image selected by the rectangular frame, identifying the category, quantity and position of each element in the image, and identifying the values ​​of all question number elements.

[0043] In one embodiment of the present application, each element in the image includes a question number and a fill-in point.

[0044] In one embodiment of the present application, the blank parts at the top, bottom, left and right of the image may be further cropped, and the cropping directions are the horizontal direction and the vertical direction.

[0045] In one embodiment of the present application, the category and position of each element in the image can be identified by the outline of the number and the outline of the filling point to identify the category, quantity and position of each element (question number and filling point) in the image. In addition, the first element in the upper left corner is usually the question number, the spacing between question numbers adjacent to each other in the upper, lower, left and right directions is a fixed value, and the spacing between each filling point in the same question is a fixed value. These general features can be used to quickly determine the category, quantity and position of each element (question number and filling point).

[0046] In one embodiment of the present application, the identified question number element is cut out according to the rectangular image to obtain the digital character image to be identified, and then the above-mentioned sub-steps S102 and S103 are processed, and the feature vector of the obtained digital character image to be identified is represented as Y=(Y1, Y2, Y3, ..., Yn), and then the distance with each character category feature vector X=(X1, X2, X3, ..., Xn) in the digital character template library is calculated, and the distance Z between the template feature component X of each number 0-9 to be identified is compared. Zj=|X1-Y1|*|X2-Y2|*|X3-Y3|*...*|Xn-Yn|, j=0,1,2,3,...,9. Calculate 10 Z values, take the minimum value, judge which category of characters has a higher matching rate, and then identify the digital character to which the Y to be identified belongs. Through this method, the values ​​of all question number elements can be further identified.

[0047] Step S600, determining the arrangement direction of the question numbers according to the identified question number values.

[0048] In one embodiment of the present application, to determine the arrangement direction of the question numbers, the first question number m in the upper left corner of the examinee's answer sheet can be located, and the value to the right of m and the value below m can be compared. If the value to the right is smaller, the question numbers are arranged horizontally. If the value below is smaller, the question numbers are arranged vertically.

[0049] Step S700, saving the category, quantity and position of each element in the image within the rectangular frame, the values ​​of all identified question number elements, and the arrangement direction of the identified question numbers into the scanning template information.

[0050] In one embodiment of the present application, the above-mentioned scanning template information may be stored in a database server for comparison with the information to be identified.

[0051] This application discloses a scanning and recognition device for automatic identification of question numbers, see Figure 2 , Figure 2 It is a schematic diagram of an embodiment of a scanning and identification device disclosed in this application.

[0052] The scanning and recognition device includes a high-speed scanner 1 for scanning an examinee's answer sheet to obtain an image of the examinee's answer sheet; an image processing module 2, connected to the high-speed scanner 1, for performing binarization processing on the image of the examinee's answer sheet; a title selection module 3, connected to the image processing module 2, for selecting continuous objective question information in the image of the examinee's answer sheet with a rectangular frame; an image recognition module 4, connected to the title selection module 3, for cropping the image selected by the rectangular frame, and identifying the category, quantity and position of each element in the image, and identifying the numerical value of all question number elements; a digital character template library 5, connected to the image recognition module 4, for storing a recognition template library of digital characters 0-9; a question number arrangement module 6, connected to the image recognition module 4, for determining the arrangement direction of the question number according to the recognized question number numerical value.

[0053] In one embodiment of the present application, the scanning and identification device further includes: a scanning template module 7, connected to the question number arrangement module 6, for storing scanning template information. The scanning template information may include the category, quantity and position of each element in the identified image, the question number value, the arrangement direction of the question number, etc. The scanning template information can be used to compare with the information to be identified.

[0054] In one embodiment of the present application, the title selection module 3 is used to select continuous objective question information in the test-taker's answer sheet image with a rectangular frame, and the objective question information at least includes the question number and the rectangular filling point on the right side of the question number. The number of rectangular filling points is generally 4, and the number of rectangular filling points for individual subjects and individual questions may be as high as 8.

[0055] In one embodiment of the present application, the image recognition module 4 is used to crop the image selected by the rectangular frame, and can further crop the blank parts of the top, bottom, left and right of the image, and the cropping directions are horizontal and vertical.

[0056] In one embodiment of the present application, the image recognition module 4 recognizes the category and position of each element in the image, and can recognize the category and position of each element (question number and filling point) in the image through the outline of the number and the outline of the filling point. In addition, the first element in the upper left corner is usually the question number, and the spacing between question numbers adjacent to each other in the upper, lower, left and right directions is a fixed value. The spacing between adjacent filling points in the same question is a fixed value. These general features can be used to quickly determine the category and position of each element (question number and filling point).

[0057] In one embodiment of the present application, the image recognition module 4 recognizes the values ​​of all question number elements. For example, the recognized question number elements can be cut out according to the rectangular image to obtain the digital character image to be recognized, and then the above-mentioned sub-steps S102 and S103 are processed. The feature vector of the digital character image to be recognized is represented as Y=(Y1, Y2, Y3, ..., Yn), and then the distance with each character category feature vector X=(X1, X2, X3, ..., Xn) in the digital character template library is calculated, and the distance Z between the template feature component X of each number 0-9 to be recognized Y is compared. Zj=|X1-Y1|*|X2-Y2|*|X3-Y3|*...*|Xn-Yn|, j=0,1,2,3,...,9. Calculate 10 Z values, take the minimum value, judge which category of characters has a higher matching rate, and then recognize the digital character to which the Y to be recognized belongs. Through this method, the values ​​of all question number elements can be further recognized.

[0058] In one embodiment of the present application, the digital character template library 5 is used to store the recognition template library of digital characters 0-9. The recognition template library storing the digital characters 0-9 is established in the following manner:

[0059] Step S100, establishing a recognition template library for numeric characters 0-9.

[0060] In one embodiment of the present application, establishing a recognition template library of digital characters 0-9 may specifically include several sub-steps:

[0061] Sub-step S101, digital graphics: grayscale the image of the printed digital font characters of the digital characters, convert it into a grayscale image, and then binarize the grayscale image. Pixel points whose grayscale values ​​meet the preset conditions are counted as 1, otherwise they are counted as 0.

[0062] In one embodiment of the present application, it can be specifically implemented by the im2bw(f,b) function, where f is the grayscale value of the input image pixel, and b is the set threshold value. The pixels in the image are divided into two categories according to the grayscale value: greater than or equal to b and less than b. The pixel points of the output binary image are 0, which corresponds to the pixel points whose grayscale value f of the input image is less than the threshold b. The pixel points of the output binary image are 1, which corresponds to the pixel points whose grayscale value f of the input image is greater than or equal to the threshold b. The title text area is represented by the value 1 (white area), and the background area is represented by the value 0 (black area).

[0063] Sub-step S102, digital image segmentation: segment the binary digital image, equally divide it in the horizontal and vertical directions, divide it into c equal parts in the vertical direction, and d equal parts in the horizontal direction, dividing it into c*d rectangular images of equal size, where c and d are both integers greater than or equal to 2.

[0064] In an embodiment of the present application, the c and d parameter values ​​can be adjusted according to the business situation. Generally, the c value is 2-4 and the d value is 2. The larger the c and d values ​​are, the higher the probability of correctly recognizing the character. Assuming c=2 and d=2, the digital image is divided into 4 (2*2) rectangles.

[0065] Sub-step S103, feature selection: based on the segmented digital image, calculate the proportion of the pixel value 1 in the digital image in each partially segmented rectangular image, and record it as the feature component X = (X1, X2, X3, ..., Xn) of the digital image, where Xn represents the proportion of the pixel value 1 in one of the rectangular images, and n is equal to c*d.

[0066] In one embodiment of the present application, the number of "1"s in each rectangular image into which the digital image is divided is used as the characteristic component of the digital image. Each rectangular image is counted from left to right and from top to bottom, and the values ​​are recorded in sequence.

[0067] Sub-step S104, sample preprocessing: perform sample training on the ten numeric characters 0-9 respectively, repeat the above sub-steps S101, S102, S103, calculate the average feature component of each sample of each number, and use the average feature component of each number as the template feature component of each number.

[0068] In one embodiment of the present application, sub-step S104 can be specifically, for example, to divide the samples corresponding to the ten numbers 0-9 into 10 categories: N1, N2, N3, ..., N10, and train the samples of each category respectively, and the number of samples of each category is preferably more than 30. According to the above sub-steps S101-S104, the feature vector of each sample image is obtained, and the average value of the feature vector of each sample of each category is calculated. According to the previous digital image samples, the regularity of each digital character in the feature space is roughly outlined to provide parameter information for further calculations later.

[0069] The present invention provides a scanning recognition method and device for automatic identification of question numbers. By collecting the answer sheets of examinees in advance, based on the establishment of a digital character template library and the selection of digital symbol features, the printed digital symbol recognition technology is used to automatically identify the question numbers of each objective question, thereby determining the sorting method of the objective questions. It does not require a lot of manual intervention, and can greatly improve work efficiency and accuracy, effectively ensuring the seriousness, authority, and fairness of the examination, and also ensuring the smooth progress of the scanning work.

[0070] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0071] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, or the positions or positional relationships in which the product of the invention is usually placed when in use. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as limiting the present invention. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0072] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A scanning recognition method for automatic identification of question numbers, characterized in that: include: Establish a recognition template library for digital characters 0-9 and save the template feature components of each digital character; Scanning a blank answer sheet of a candidate to obtain a candidate answer sheet image, and storing the candidate answer sheet image; The above-mentioned examinee answer sheet image is gray-processed to be converted into a gray-scale image, and then the gray-scale image is binarized; In the above-mentioned image of the examinee's answer sheet, use a rectangular frame to select continuous objective question information; Cut the image in the rectangular frame, identify the category, quantity and position of each element in the image in the rectangular frame, and identify the values ​​of all the question number elements; According to the values ​​of the identified question number elements, determining the arrangement direction of the question numbers; The step of establishing a recognition template library for digital characters 0-9 and storing the template feature components of each digital character includes: Grayscale processing is performed on the image of the printed digital font characters of the digital characters, and the image is converted into a grayscale image, and then the grayscale image is binarized, and the pixel points whose grayscale values ​​meet the preset conditions are counted as 1, otherwise they are counted as 0; The binary digital image is segmented equally in the horizontal and vertical directions, into c equal parts in the vertical direction and d equal parts in the horizontal direction, into c*d equal-sized rectangular images, where c and d are both integers greater than or equal to 2; According to the segmented digital image, the proportion of the pixel value 1 in the digital image in each of the segmented rectangular images is calculated and recorded as the characteristic component X=(X1, X2, X3, ..., Xn) of the digital image, where Xn represents the proportion of the pixel value 1 in one of the rectangular images, and n is equal to c*d; Sample training is performed on the ten numeric characters 0-9 respectively, and the average feature component of each sample of each number is calculated, and the average feature component of each number is used as the template feature component of each numeric character.

2. The scanning and recognition method for automatic question number recognition according to claim 1, characterized in that: The objective question information includes: Fill in the question number and the rectangle to the right of the question number.

3. The scanning and recognition method for automatic question number recognition according to claim 1, characterized in that: Identify the category and position of each element in the image, and identify the category and position of each element in the image through digital contours and filled point contours.

4. The scanning and recognition method for automatic question number recognition according to claim 1, characterized in that: Identify the category and position of each element in the image, and determine the category and position of each element through the spacing between adjacent question numbers, or the spacing between adjacent filling points in the same question.

5. The scanning and recognition method for automatic question number recognition according to claim 1, characterized in that: The numerical values ​​of all question number elements are identified by comparing the distance between the feature vector of the rectangular image of the question number element to be identified and the template feature component of each number from 0 to 9. The smaller the distance, the higher the similarity.

6. The scanning and recognition method for automatic question number recognition according to claim 1, characterized in that: According to the identified question number value, the arrangement direction of the question number is determined including: Locate the first question number in the upper left corner of the candidate's answer sheet, and compare the value to the right of the first question number with the value below the first question number. If the value on the right is smaller, the question numbers are arranged horizontally. If the value below is smaller, the question numbers are arranged vertically.

7. The scanning and recognition method for automatic question number recognition according to claim 1, characterized in that: include: The category, quantity and position of each element in the image within the rectangular frame, the values ​​of all the identified question number elements, and the arrangement direction of the identified question numbers are saved in the scanning template information.

8. A scanning and recognition device for automatic identification of question numbers, characterized in that: include: A high-speed scanner is used to scan the answer sheet of the examinee and obtain the image of the answer sheet of the examinee; An image processing module, connected to the high-speed scanner, for performing binarization processing on the image of the examinee's answer sheet; A title selection module, connected to the image processing module, is used to select continuous objective question information in the candidate's answer sheet image with a rectangular frame; The image recognition module is connected to the above-mentioned title selection module, and is used to crop the image selected by the rectangular frame, and identify the category, quantity and position of each element in the image, and identify the value of all the title elements; A digital character template library, connected to the above-mentioned image recognition module, is used to store a recognition template library of digital characters 0-9; A question number arrangement module, connected to the image recognition module, is used to determine the arrangement direction of the question numbers according to the recognized question number values; Grayscale processing is performed on the image of the printed digital font characters of the digital characters, and the image is converted into a grayscale image, and then the grayscale image is binarized, and the pixel points whose grayscale values ​​meet the preset conditions are counted as 1, otherwise they are counted as 0; The binary digital image is segmented equally in the horizontal and vertical directions, into c equal parts in the vertical direction and d equal parts in the horizontal direction, into c*d equal-sized rectangular images, where c and d are both integers greater than or equal to 2; According to the segmented digital image, the proportion of the pixel value 1 in the digital image in each of the segmented rectangular images is calculated and recorded as the characteristic component X=(X1, X2, X3, ..., Xn) of the digital image, where Xn represents the proportion of the pixel value 1 in one of the rectangular images, and n is equal to c*d; Sample training is performed on the ten numeric characters 0-9 respectively, and the average feature component of each sample of each number is calculated, and the average feature component of each number is used as the template feature component of each numeric character.

9. The scanning and recognition device for automatic identification of question numbers according to claim 8, characterized in that: The objective question information includes: Fill in the question number and the rectangle to the right of the question number.

Images