Multiple-choice Question Localization Processing System and Method Based on Computer Vision

Abstract

The present invention discloses a multiple-choice question positioning processing system and method based on computer vision, including the following steps: cutting the picture of the multiple-choice questions from the picture of the test paper according to the size and style of the test paper, performing graying, binarization and noise removal processing on the scanned image of the test paper to obtain a preprocessed image; on the preprocessed image, extracting three features related to the filling area of the multiple-choice questions, including: the actual filling of the student, the box of the multiple-choice question filling area, and the line segment of the multiple-choice question filling area box; based on the extracted feature information, calculating the starting coordinates of the multiple-choice question filling area, where the starting point of the filling area is the upper left corner position of the first option; performing boundary adjustment, interference elimination and image correction processing on the calculated filling area to obtain the final filling area positioning result. The present invention has the characteristics of high automation degree and accurate positioning.

Description

Technical Field

[0001] The present invention relates to the technical field of computer vision, and particularly to a multiple-choice question positioning processing system and method based on computer vision. Background Art

[0002] Traditional test paper marking mainly relies on manual work, which is not only time-consuming and laborious, but also has certain subjectivity. With the rapid development of computer vision technology, more and more application scenarios begin to involve automated test paper marking, especially the automatic marking of objective questions such as multiple-choice questions and fill-in-the-blank questions. However, in actual applications, there are often some variables in the physical state of the test paper and the answering behavior of students, such as poor printing quality of the test paper, tilting and deformation of the scanned image, and writing interference during students' answering. These factors pose challenges to the positioning and recognition of multiple-choice questions.

[0003] Existing technical solutions usually adopt the positioning mark method, which sets positioning marks on the multiple-choice question test paper for the alignment and calibration of multiple-choice questions in the automatic marking system. However, this technical solution shows limitations when dealing with various different formats of test papers. Especially when the test paper format, the position or size of the positioning mark changes, or in the case of no positioning mark, the existing automatic marking system may not work properly, resulting in a decrease in marking accuracy or the inability to perform automatic marking. Therefore, it is necessary to design a multiple-choice question positioning processing system and method based on computer vision with high automation and accurate positioning. Summary of the Invention

[0004] The purpose of the present invention is to provide a multiple-choice question positioning processing system and method based on computer vision to solve the problems raised in the above background art.

[0005] To solve the above technical problems, the present invention provides the following technical solution: A multiple-choice question positioning processing method based on computer vision, including the following steps:

[0006] A. According to the size and style of the test paper, cut out the picture of the multiple-choice question from the test paper picture, and perform grayscale conversion, binarization, and noise removal processing on the scanned image of the test paper to obtain a preprocessed image;

[0007] B. On the preprocessed image, extract three features related to the filling area of the multiple-choice question, including: the actual filling of the student, the box of the multiple-choice question filling area, and the line segment of the multiple-choice question filling area box;

[0008] C. Based on the extracted feature information, calculate the starting coordinates of the filling area of the multiple-choice question;

[0009] D. Perform boundary adjustment, interference elimination, and image correction processing on the calculated filling area to obtain the final positioning result of the filling area.

[0010] According to the above technical solution, step B specifically includes:

[0011] Students' actual filling: Use the OTSU algorithm to convert the grayscale image into a binary image, use the morphological opening operation to eliminate the filling frame and the ABCD letters of the multiple-choice questions other than the black filled items, then use the contour search algorithm to find the contour of the filled area, match the search contour length greater than the preset value to find the filled area, and finally calculate the x, y coordinates of the first multiple-choice question based on the known filling. If there is no suitable filling, the calculated x or y value is invalid;

[0012] Multiple-choice question filling area box: Use the adaptive threshold algorithm to convert the grayscale image into a binary image, and then use the contour search algorithm to find the contour of the filling area box. According to the matching search, the part with a contour length greater than the preset value is the search for the filling area. Finally, based on the known filling area, calculate the x, y coordinates of the first multiple-choice question; if there is no suitable filling, the calculated x or y value is invalid.

[0013] Line segments of the filling area box of multiple-choice questions: Through morphological operations, the line segment features in the image are extracted, including horizontal line segments and vertical line segments. Then, the line segment detection algorithm is used to find the line segments of the filling area box. According to the position and length information of the line segments, the position and range of the filling area box are determined, and the x, y coordinates of the first multiple-choice question are calculated; if there is no suitable filling, the calculated x or y value is invalid.

[0014] According to the above technical solution, the calculation of the starting point coordinates of the filling area of ​​the multiple-choice question in step C specifically includes:

[0015] For each multiple-choice question starting point coordinate obtained by each feature extraction method, a credibility value is attached.

[0016] Among the starting point coordinates obtained by the three feature extraction methods, the coordinates with the highest credibility are selected as the final starting point coordinates of the filling area to achieve comprehensive value selection.

[0017] According to the above technical solution, the credibility value calculation method includes:

[0018] Step S1: Collect and preprocess data, including cleaning data, removing duplicate values, and processing missing values;

[0019] Step S2: According to the definition and quantification method of the evaluation index, score each indicator and get S 1 , S 2 , S 3 ;

[0020] Step S3: Assign weight W according to the importance of the evaluation indicator 1 , W 2 , W 3 , where W1 +W 2 +W 3 = 1;

[0021] Step S4: Use the credibility calculation formula, substitute the quantization value and weight into the formula, and calculate the credibility value C;

[0022] Step S5: Output the credibility value C.

[0023] According to the above technical solution, the method for comprehensive value taking includes:

[0024] Step A1: Set the starting coordinates obtained by the student's actual filling as (x 1 , y 1 ), and the attached credibility output value is C 1 ;

[0025] Set the starting coordinates obtained by the multiple-choice question filling area box as (x 2 , y 2 ), and the attached credibility output value is C 2 ;

[0026] Set the starting coordinates obtained by the multiple-choice question filling area box line segment as (x 3 , y 3 ), and the attached credibility output value is C 3 ;

[0027] Set the credibility threshold T. When the credibility C ≥ T, the corresponding coordinates are regarded as valid coordinates;

[0028] Step A2: Screen the valid coordinates according to the credibility threshold T: When C 1 ≥ T, then (x 1 , y 1 ) are valid coordinates. When C 2 ≥ T, then (x 2 , y 2 ) are valid coordinates. When C 3 ≥ T, then (x 3 , y 3 ) are valid coordinates. Output the screened valid coordinates (x i , y i ), where i is a certain value among 1, 2, or 3, and capture the number of valid coordinates;

[0029] Step A3: When the number of captured valid coordinates is 1, directly select this coordinate as the final starting coordinate of the filling area;

[0030] When the number of captured valid coordinates is greater than 1, compare the credibility values and determine the final coordinates (x final , y final ), where (xfinal , y final ) = argC i , where (x i , y i ) ∈ valid coordinates, that is, select the coordinate with the highest credibility as the starting coordinate of the final filling area;

[0031] When the number of captured valid coordinates is 0, the multiple-choice question positioning processing step of the current test paper is ended, and a prompt signal is output to remind the staff to perform manual positioning.

[0032] According to the above technical solution, in the step S4, the credibility calculation formula is:

[0033] C = W 1 * S 1 + W 2 * S 2 + W 3 * S 3

[0034] In the formula, where C represents the credibility value, W 1 , W 2 , W 3 respectively represent the weights of the credibility of the information source, the credibility of the information content, and the relevance of the target, and S 1 , S 2 , S 3 respectively represent the quantization values of the credibility of the information source, the credibility of the information content, and the relevance evaluation index of the target.

[0035] According to the above technical solution, the multiple-choice question positioning processing system based on computer vision includes an image preprocessing module, a feature extraction module, a starting coordinate calculation module, and a post-processing module. The image preprocessing module, the feature extraction module, the starting coordinate calculation module, and the post-processing module are electrically connected to each other; among them,

[0036] The image preprocessing module is used to perform grayscale conversion, binarization, and noise removal processing on the scanned test paper image;

[0037] The feature extraction module is used to extract three features related to the multiple-choice question filling area on the preprocessed image, including the actual filling of students, the box of the multiple-choice question filling area, and the line segment of the multiple-choice question filling area box;

[0038] The starting coordinate calculation module is used to calculate the starting coordinate of the multiple-choice question filling area based on the extracted feature information;

[0039] The post-processing module is used to perform boundary adjustment, interference elimination, and image correction processing on the calculated filling area and obtain the final filling area positioning result.

[0040] According to the above technical solution, the feature extraction module includes a student actual filling extraction sub-module, a multiple-choice question filling area box extraction sub-module, and a multiple-choice question filling area box line segment extraction sub-module; among them,

[0041] The student actual filling extraction sub-module is used to extract the actual filling features of the student;

[0042] The multiple-choice question filling area box extraction sub-module is used to extract the multiple-choice question filling area box features;

[0043] The multiple-choice question filling area box line segment extraction sub-module is used to extract the multiple-choice question filling area box line segment features.

[0044] According to the above technical solution, the starting point coordinate calculation module includes a credibility evaluation unit, and the credibility evaluation unit is used to attach a credibility value to the starting point coordinates of multiple-choice questions obtained by each feature extraction method. The credibility evaluation unit further includes a data preprocessing sub-unit, a quantization scoring sub-unit, a weight assignment sub-unit, and a credibility calculation sub-unit, and the data preprocessing sub-unit, the quantization scoring sub-unit, the weight assignment sub-unit, and the credibility calculation sub-unit are electrically connected to each other; among them,

[0045] The data preprocessing sub-unit is used to collect and preprocess data;

[0046] The quantization scoring sub-unit is used to score each index according to the definition and quantization method of the evaluation index;

[0047] The weight assignment sub-unit is used to assign weight values according to the importance of the evaluation index;

[0048] The credibility calculation sub-unit is used to calculate the credibility value.

[0049] According to the above technical solution, the starting point coordinate calculation module further includes a comprehensive value-taking unit, and the comprehensive value-taking unit is used for:

[0050] Setting the starting point coordinates obtained by the three feature extraction methods and their corresponding credibility values;

[0051] Setting a credibility threshold to screen valid coordinates;

[0052] Selecting the final starting point coordinate of the filling area according to the number of captured valid coordinates.

[0053] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: Through the analysis of three different characteristics of the multiple-choice question filling area, the present invention can accurately locate the starting point of the multiple-choice questions without additional multiple-choice question positioning marks in the case of tilting, deformation, and interference of the scanned image, achieving the improvement of the accuracy and robustness of the multiple-choice question positioning and recognition, thereby greatly improving the recognition rate of multiple-choice questions, reducing the requirements of the recognition program for the test paper format, and being able to adapt to test papers of different formats in various regions and schools. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention.

[0055] In the accompanying drawings:

[0056] Figure 1 FIG. is a flowchart of the steps of a multiple-choice question positioning processing method based on computer vision provided in Embodiment 1 of the present invention;

[0057] Figure 2 FIG. is a schematic diagram of the module composition of a multiple-choice question positioning processing system based on computer vision provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0058] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0059] Embodiment 1

[0060] Please refer to Figure 1 , the present invention provides the following technical solution: A multiple-choice question positioning processing method based on computer vision, including the following steps:

[0061] A. According to the size and style of the test paper, cut out the picture of the multiple-choice questions from the test paper picture, and perform graying, binarization, and noise removal processing on the test paper scanned image to obtain a preprocessed image;

[0062] B. On the preprocessed image, extract three characteristics related to the multiple-choice question filling area, including: the actual filling of students, the box of the multiple-choice question filling area, and the line segment of the box of the multiple-choice question filling area;

[0063] C. Based on the extracted feature information, calculate the starting point coordinates of the multiple-choice question filling area, where the starting point of the filling area is the upper left corner position of the first option;

[0064] D. Perform boundary adjustment, interference elimination, and image correction on the calculated filled area to obtain the final filled area positioning result.

[0065] Step B specifically includes:

[0066] Actual filling by students: Use the OTSU algorithm to convert the grayscale image into a binary image, eliminate the filling box lines and the ABCD letters of multiple-choice questions outside the already filled items through morphological opening operation, then use the contour search algorithm to find the contour of the filled area, and the part where the length of the matching search contour is greater than the preset value is the searched filled area. Finally, according to the known filling, calculate the x and y coordinates of the first multiple-choice question. If there is no appropriate filling, the calculated x or y value is invalid;

[0067] Box of the filled area of multiple-choice questions: Use the adaptive threshold algorithm to convert the grayscale image into a binary image, then use the contour search algorithm to find the contour of the filled area box. The part where the length of the matching search contour is greater than the preset value is the searched filled area. Finally, according to the known filled area, calculate the x and y coordinates of the first multiple-choice question; if there is no appropriate filling, the calculated x or y value is invalid.

[0068] Line segments of the box of the filled area of multiple-choice questions: Extract the line segment features in the image through morphological operations, including horizontal line segments and vertical line segments, then use the line segment detection algorithm to find the line segments of the filled area box. According to the position and length information of the line segments, determine the position and range of the filled area box, and calculate the x and y coordinates of the first multiple-choice question; if there is no appropriate filling, the calculated x or y value is invalid, where the line segment detection algorithm includes but is not limited to the Hough transform.

[0069] The specific steps for calculating the starting coordinates of the filled area of multiple-choice questions in Step C include:

[0070] For the starting coordinates of multiple-choice questions obtained by each feature extraction method, a credibility value is attached; the credibility value can be comprehensively evaluated based on factors such as the integrity of the contour, size consistency, and positional relationship.

[0071] Among the starting coordinates obtained by the three feature extraction methods, select the coordinate with the highest credibility as the final starting coordinate of the filled area to achieve comprehensive value taking; if the coordinate obtained by a certain method is invalid, such as no appropriate contour or line segment is found, set its credibility to 0 and exclude it from the final selection.

[0072] The calculation method of the credibility value includes:

[0073] Step S1: Collect data and preprocess it, including cleaning data, removing duplicate values, and handling missing values;

[0074] Step S2: Score each indicator according to the definition and quantification method of the evaluation indicators to obtain S 1 、S 2 、S 3 ; where the quantification range is from 0 to 1, where 1 represents completely credible and 0 represents completely incredible, and S 1 is the scoring value based on the authority and historical accuracy of the information source; S 2 is the scoring value based on the accuracy, timeliness, and integrity of the information content; S 3 is to calculate the correlation between the data and the target using the correlation measurement method and convert it into a quantification value from 0 to 1;

[0075] Step S3: Allocate weights W 1 、W 2 、W 3 , where W 1 +W 2 +W 3 = 1; when the importance of the evaluation indicator is higher, the corresponding weight value W is larger;

[0076] Step S4: Use the credibility calculation formula, substitute the quantification value and weight into the formula, and calculate the credibility value C;

[0077] Step S5: Output the credibility value C; Through the above steps, the credibility of the starting coordinates of the multiple-choice questions obtained by each feature extraction method can be calculated more accurately, and the most reliable coordinates can be comprehensively selected as the starting coordinates of the final filling area, thereby effectively improving the accuracy and robustness of the multiple-choice question positioning process.

[0078] The method of comprehensive value extraction includes:

[0079] Step A1: Set the starting coordinates actually filled by the student as (x 1 , y 1 ), and the attached credibility output value is C 1 ;

[0080] Set the starting coordinates of the multiple-choice question filling area box as (x 2 , y 2 ), and the attached credibility output value is C 2 ;

[0081] Set the starting coordinates of the multiple-choice question filling area box line segment as (x 3 , y 3 ), and the attached credibility output value is C 3 ;

[0082] Set the credibility threshold T. When the credibility C ≥ T, the corresponding coordinates are regarded as valid coordinates;

[0083] Step A2: Screen valid coordinates according to the credibility threshold T: When C 1 ≥T, then (x 1 , y 1 ) is a valid coordinate. When C 2 ≥T, then (x 2 , y 2 ) is a valid coordinate. When C 3 ≥T, then (x 3 , y 3 ) is a valid coordinate. Output the screened valid coordinates (x i , y i ), where i is a certain value among 1, 2, or 3, and capture the number of valid coordinates;

[0084] Step A3: When the number of captured valid coordinates is 1, directly select this coordinate as the starting coordinate of the final filling area;

[0085] When the number of captured valid coordinates is greater than 1, compare the credibility values and determine the final coordinates (x final , y final ), where (x final , y final ) = argC i , where (x i , y i ) ∈ valid coordinates, that is, select the coordinate with the highest credibility as the starting coordinate of the final filling area;

[0086] When the number of captured valid coordinates is 0, end the multiple-choice question positioning processing step of the current test paper and output a prompt signal to remind the staff to perform manual positioning.

[0087] In step S4, the credibility calculation formula is:

[0088] C = W 1 * S 1 + W 2 * S 2 + W 3 * S 3

[0089] In the formula, where C represents the credibility value, W 1 , W 2 , W 3 respectively represent the weights of the credibility of the information source, the credibility of the information content, and the relevance of the target, and S 1 , S 2 , S 3 respectively represent the quantification values of the credibility of the information source, the credibility of the information content, and the relevance evaluation index of the target.

[0090] In the above steps, through the analysis of three different characteristics of the multiple-choice question filling area, in the case of the scanned image being tilted, deformed, or having interference, no additional multiple-choice question positioning marks are required, and the starting point of the multiple-choice questions can be accurately located, achieving the effect of improving the accuracy and robustness of multiple-choice question positioning and recognition, thereby greatly increasing the recognition rate of multiple-choice questions, reducing the requirements of the recognition program for the test paper format, and being able to adapt to test papers of different formats in various regions and schools.

[0091] Embodiment 2

[0092] Please refer to Figure 2 , the present invention also provides a computer vision-based multiple-choice question positioning processing system for implementing the method in Embodiment 1. The system includes an image preprocessing module, a feature extraction module, a starting point coordinate calculation module, and a post-processing module. The image preprocessing module, the feature extraction module, the starting point coordinate calculation module, and the post-processing module are electrically connected to each other; wherein,

[0093] The image preprocessing module is used to perform grayscale conversion, binarization, and noise removal processing on the scanned test paper image;

[0094] The feature extraction module is used to extract three features related to the multiple-choice question filling area on the preprocessed image, including the actual filling of students, the square of the multiple-choice question filling area, and the line segment of the square of the multiple-choice question filling area;

[0095] The starting point coordinate calculation module is used to calculate the starting point coordinates of the multiple-choice question filling area based on the extracted feature information;

[0096] The post-processing module is used to perform boundary adjustment, interference elimination, and image correction processing on the calculated filling area and obtain the final filling area positioning result.

[0097] The feature extraction module includes a student actual filling extraction sub-module, a multiple-choice question filling area square extraction sub-module, and a multiple-choice question filling area square line segment extraction sub-module; wherein,

[0098] The student actual filling extraction sub-module is used to extract the actual filling features of students;

[0099] The multiple-choice question filling area square extraction sub-module is used to extract the square features of the multiple-choice question filling area;

[0100] The multiple-choice question filling area square line segment extraction sub-module is used to extract the line segment features of the square of the multiple-choice question filling area.

[0101] The starting point coordinate calculation module includes a credibility evaluation unit. The credibility evaluation unit is used to attach a credibility value to the starting point coordinates of multiple-choice questions obtained by each feature extraction method. The credibility evaluation unit further includes a data preprocessing subunit, a quantization scoring subunit, a weight assignment subunit, and a credibility calculation subunit, and the data preprocessing subunit, the quantization scoring subunit, the weight assignment subunit, and the credibility calculation subunit are electrically connected to each other; wherein,

[0102] The data preprocessing subunit is used to collect and preprocess data;

[0103] The quantization scoring subunit is used to score each index according to the definition and quantization method of the evaluation index;

[0104] The weight assignment subunit is used to assign weight values according to the importance of the evaluation index;

[0105] The credibility calculation subunit is used to calculate the credibility value.

[0106] The starting point coordinate calculation module further includes a comprehensive value-taking unit. The comprehensive value-taking unit is used for:

[0107] Setting the starting point coordinates obtained by three feature extraction methods and their corresponding credibility values;

[0108] Setting a credibility threshold to screen valid coordinates;

[0109] Selecting the final starting point coordinates of the filling area according to the number of captured valid coordinates;

[0110] Compared with the existing method of using positioning marks, which shows limitations when dealing with various different formats of test papers, the system may not work properly, the accuracy of marking papers is reduced, or automatic marking cannot be carried out. Through advanced image processing and intelligent algorithms, the accuracy and robustness of multiple-choice question positioning and recognition are improved. When facing the above challenges, the system can still accurately identify the answers to multiple-choice questions, thereby improving the automation and accuracy of exam evaluation.

[0111] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "including", "comprising" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements not only includes those elements, but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device.

[0112] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A method for positioning and processing multiple-choice questions based on computer vision, characterized in that: The following steps are involved: A. According to the size and style of the test paper, cut out the picture of the multiple-choice questions from the test paper picture, grayscale, binarize and remove noise from the scanned image of the test paper to obtain the pre-processed image; B. Extract three features related to the filling area of ​​the multiple-choice questions from the pre-processed image, including: the actual filling of the students, the box of the filling area of ​​the multiple-choice questions, and the line segments of the box of the filling area of ​​the multiple-choice questions; C. Based on the extracted feature information, calculate the starting point coordinates of the filling area of ​​the multiple-choice question; D. Perform boundary adjustment, interference elimination and image correction on the calculated filling area to obtain the final filling area positioning result; The calculation of the starting point coordinates of the filling area of ​​the multiple-choice question in step C specifically includes: For each multiple-choice question starting point coordinate obtained by each feature extraction method, a credibility value is attached; Among the starting point coordinates obtained by the three feature extraction methods, the coordinates with the highest credibility are selected as the final starting point coordinates of the filling area to achieve comprehensive value selection.

2. The method for locating and processing multiple-choice questions based on computer vision according to claim 1, characterized in that: Step B specifically includes: Students' actual filling: Use the OTSU algorithm to convert the grayscale image into a binary image, use the morphological opening operation to eliminate the filling frame and the ABCD letters of the multiple-choice questions other than the black filled items, then use the contour search algorithm to find the contour of the filled area, match the search contour length greater than the preset value to find the filled area, and finally calculate the x, y coordinates of the first multiple-choice question based on the known filling. If there is no suitable filling, the calculated x or y value is invalid; Multiple-choice question filling area box: Use the adaptive threshold algorithm to convert the grayscale image into a binary image, then use the contour search algorithm to find the contour of the filling area box, and search for the part with a contour length greater than the preset value based on the matching search to find the filling area. Finally, based on the known filling area, calculate the x, y coordinates of the first multiple-choice question; if there is no suitable filling, the calculated x or y value is invalid; Line segments of the filling area box of multiple-choice questions: Through morphological operations, the line segment features in the image are extracted, including horizontal line segments and vertical line segments. Then, the line segment detection algorithm is used to find the line segments of the filling area box. According to the position and length information of the line segments, the position and range of the filling area box are determined, and the x, y coordinates of the first multiple-choice question are calculated; if there is no suitable filling, the calculated x or y value is invalid.

3. The method for locating and processing multiple-choice questions based on computer vision according to claim 1, characterized in that: The credibility value calculation method includes: Step S1: Collect and preprocess data, including cleaning data, removing duplicate values, and processing missing values; Step S2: According to the definition and quantification method of the evaluation indicators, score each indicator to obtain S1, S2, and S3; Step S3: According to the importance of the evaluation indicators, weights W1, W2, and W3 are assigned, where W1+W2+W3=1; Step S4: Use the credibility calculation formula, substitute the quantized value and the weight into the formula, and calculate the credibility value C; Step S5: Output the credibility value C.

4. The method for locating and processing multiple-choice questions based on computer vision according to claim 1, characterized in that: The comprehensive value methods include: Step A1: Set the starting point coordinates obtained by the student's actual filling to (x1, y1), and the accompanying credibility output value to C1; Set the starting point coordinates of the multiple choice question filling area box to (x2, y2), and the accompanying credibility output value to C2; Set the starting point coordinates of the box segment of the multiple-choice question filling area to (x3, y3), and the accompanying credibility output value to C3; Set the credibility threshold T. When the credibility C ≥ T, the corresponding coordinates are considered valid coordinates. Step A2: Filter valid coordinates according to the credibility threshold T: when C1 ≥ T, (x1, y1) is a valid coordinate, when C2 ≥ T, (x2, y2) is a valid coordinate, when C3 ≥ T, (x3, y3) is a valid coordinate, and output the filtered valid coordinates (x i ,y i ), where i is a value of 1, 2, or 3, capturing the number of valid coordinates; Step A3: When the number of captured valid coordinates is 1, the coordinate is directly selected as the final starting coordinate of the filling area; When the number of captured valid coordinates is greater than 1, the credibility value is compared and the final coordinate (x final ,y final ), where (x final ,x final )=argC i , where (x i ,y i )∈valid coordinates, that is, the coordinate with the highest credibility is selected as the final starting coordinate of the filling area; When the number of captured valid coordinates is 0, the multiple-choice question positioning processing step of the current test paper is terminated, and a prompt signal is output to remind the staff to perform manual positioning.

5. The method for locating and processing multiple-choice questions based on computer vision according to claim 3, characterized in that: In step S4, the credibility calculation formula is: C=W1*S1+W2*S2+W3*S3 In the formula, C represents the credibility value, W1, W2, and W3 represent the weights of the credibility of the information source, the credibility of the information content, and the relevance of the target, respectively, and S1, S2, and S3 represent the quantitative values ​​of the credibility of the information source, the credibility of the information content, and the relevance evaluation indicators of the target, respectively.

6. A multiple-choice question positioning processing system based on computer vision, characterized in that: The system includes an image preprocessing module, a feature extraction module, a starting point coordinate calculation module and a post-processing module, wherein the image preprocessing module, the feature extraction module, the starting point coordinate calculation module and the post-processing module are electrically connected to each other; wherein, The image preprocessing module is used to perform grayscale, binarization and noise removal processing on the scanned image of the test paper; The feature extraction module is used to extract three features related to the multiple-choice question filling area on the pre-processed image, including the student's actual filling, the multiple-choice question filling area box and the multiple-choice question filling area box line segment; The starting point coordinate calculation module is used to calculate the starting point coordinates of the filling area of ​​the multiple-choice question based on the extracted feature information; The post-processing module is used to perform boundary adjustment, interference elimination and image correction processing on the calculated filling area and obtain the final filling area positioning result; The starting point coordinate calculation module includes a credibility evaluation unit, which is used to attach a credibility value to the starting point coordinates of each multiple-choice question obtained by the feature extraction method. The credibility evaluation unit further includes a data preprocessing subunit, a quantization subunit, a weight allocation subunit and a credibility calculation subunit. The data preprocessing subunit, the quantization subunit, the weight allocation subunit and the credibility calculation subunit are electrically connected to each other; wherein, The data preprocessing subunit is used to collect and preprocess data; The quantitative scoring subunit is used to score each indicator according to the definition and quantification method of the evaluation indicator; The weight allocation subunit is used to allocate weight values ​​according to the importance of the evaluation index; The credibility calculation subunit is used to calculate the credibility value; The starting point coordinate calculation module further includes a comprehensive value obtaining unit, which is used to: Set the starting point coordinates obtained by the three feature extraction methods and their corresponding credibility values; Set the credibility threshold to filter valid coordinates; Based on the number of captured valid coordinates, the final starting coordinates of the filling area are selected.

7. The multiple choice question positioning processing system based on computer vision according to claim 6, characterized in that: The feature extraction module includes a student actual filling extraction submodule, a multiple-choice question filling area box extraction submodule, and a multiple-choice question filling area box line segment extraction submodule; wherein, The student actual filling extraction submodule is used to extract the student's actual filling features; The multiple choice question filling area box extraction submodule is used to extract the features of the multiple choice question filling area box; The multiple-choice question filling area box line segment extraction submodule is used to extract the box line segment features of the multiple-choice question filling area.

Images