Test paper correction score proofreading method and device, storage medium and terminal

Abstract

The invention discloses an examination paper correction score proofreading method and device, a storage medium and a terminal, relates to the technical field of image recognition, and mainly aims to solve the problem of low examination paper correction score proofreading accuracy. The method mainly comprises the steps that a test paper image of test paper to be corrected is collected, target detection is carried out on the test paper image, correction frames in the test paper to be corrected are obtained, and the correction frames comprise a test paper total score correction frame and question correction frames corresponding to different questions respectively; performing character recognition on the correction frame, determining a total score of examination paper correction and scores of all the questions according to a recognition result, and calculating to obtain a total score of the examination paper according to the scores of all the questions; and if the total score of the test paper is not matched with the total score of the test paper correction, generating an abnormal correction result of the test paper to be corrected. The method is mainly used for correcting the score of the test paper.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and in particular to a method and device for checking test paper marking scores, a storage medium and a terminal. Background Art

[0002] In the process of evaluating teaching results, test paper grading and score proofreading are key links. Although the development of artificial intelligence has promoted the popularization of automated grading systems, manual test paper grading still occupies an important position in many educational scenarios because manual test paper grading can not only handle complex subjective questions, but also provide students with more targeted feedback through teachers' professional judgment. However, due to its reliance on manual labor, manual test paper grading faces the problem of a high error rate in score statistics.

[0003] In order to solve the problem of high error rate in manual examination marking, the existing method mainly adopts manual cross-validation to check the scores. However, this verification method requires a lot of manpower, which is costly and time-consuming, resulting in low efficiency of examination score proofreading. Summary of the invention

[0004] In view of this, the present invention provides a method and device for proofreading test paper grading scores, a storage medium, and a terminal, the main purpose of which is to solve the problem of low test paper score proofreading efficiency in existing manual test paper grading.

[0005] According to one aspect of the present invention, a method for checking test paper marking scores is provided, comprising:

[0006] Collecting an image of the test paper to be proofread, and performing target detection on the test paper image to obtain correction frames in the test paper to be proofread, wherein the correction frames include a correction frame for the total score of the test paper and question correction frames corresponding to different questions;

[0007] Performing character recognition on the correction box, determining the total score of the test paper correction and the score of each question according to the recognition result, and calculating the total score of the test paper according to the scores of all the questions;

[0008] If the total score of the test paper does not match the total score of the test paper correction, a proofreading exception result of the test paper to be proofread is generated.

[0009] Furthermore, the result of the target detection also includes a question frame, and the target detection is performed on the test paper image to obtain a correction frame in the test paper to be proofread, including:

[0010] Performing target detection on the test paper image to obtain question frame coordinate information and correction frame coordinate information;

[0011] For any title frame, the intersection IOU value between the title frame and the correction frame is calculated according to the coordinate information of the title frame and the coordinate information of the correction frame, wherein the intersection IOU value is the ratio of the overlapping area between the title frame and the correction frame to the area of ​​the correction frame;

[0012] Determine the correction box whose intersection IOU value is greater than the intersection IOU threshold as the title correction box of the title corresponding to the title box;

[0013] The correction box whose intersection IOU values ​​with all the question boxes are less than the intersection IOU threshold is determined as the correction box for the total score of the test paper.

[0014] Furthermore, the question correction box includes a score correction box and / or a symbol correction box, and the method further includes:

[0015] In the case where the score correction box or the symbol correction box corresponding to any question is missing, the question image is cropped according to the coordinate information of the question box to obtain the question image;

[0016] Binarizing the question image to obtain an initial connected domain block image of the question, and denoising the initial connected domain block image according to a denoising area threshold to obtain a target connected domain block image;

[0017] Obtaining the horizontal coordinate boundary value of the correction box of the previous question and the vertical coordinate boundary value of the question, and cropping the target connected domain block image according to the horizontal coordinate boundary value and the vertical coordinate boundary value;

[0018] The cropped connected domain block image is recognized by a correction trace recognition model to obtain the correction symbol of the question, or the cropped connected domain block image is recognized by a score recognition model to obtain the correction score of the question.

[0019] Furthermore, the recognition result includes the total score of the test paper, and the correction symbol and / or correction score of each of the questions, and the character recognition of the correction box is performed to determine the total score of the test paper and the score of each of the questions according to the recognition result, including:

[0020] For any of the questions, when the recognition result only includes a correction symbol, if the correction symbol is a check mark, the score of the question is configured as the score of the question; if the correction symbol is a wrong check mark, the score of the question is configured as zero;

[0021] In the case where the recognition result includes half correct and half wrong answers and a correction score, if the correction score is a positive number, the correction score is configured as the score of the question; if the correction score is a negative number, the sum of the correction score and the score of the question is configured as the score of the question.

[0022] Furthermore, the method further comprises:

[0023] In the case that the recognition result does not satisfy the score symbol matching strategy, the image of the symbol correction box is segmented, and the segmented image is recognized by the correction trace recognition model to obtain a secondary recognition result of the symbol correction box, wherein the segmented symbol matching strategy includes that the correction symbol is a correct or incorrect number, and the recognition result of the correction box does not include a correction score, and the correction symbol is half correct and half incorrect, and the recognition result of the correction box includes a correction score;

[0024] If the secondary recognition result of the symbol correction box still does not satisfy the score symbol matching strategy, segment the image of the score correction box, and recognize the segmented image through the score recognition model to obtain the secondary recognition result of the score correction box;

[0025] The score of the question is determined according to the secondary recognition result of the score correction box.

[0026] Furthermore, the step of determining the score of the question according to the secondary recognition result of the score correction box includes:

[0027] If the absolute value of the correction score in the secondary recognition result of the score correction box is less than or equal to the score of the question, the correction score is configured as the score of the question;

[0028] If the absolute value of the correction score in the secondary recognition result of the score correction box is greater than the score of the question, the candidate correction score in the secondary recognition result of the score correction box is retrieved, and the candidate correction score whose absolute value is less than or equal to the score of the question and whose confidence is the largest is configured as the score of the question.

[0029] Furthermore, the target detection result also includes a total score frame of the test paper and question score frames corresponding to different questions, and the recognition result of the question score frame includes the scores of different questions and the total score of the test paper. The method also includes:

[0030] In the case of abnormal test paper scores, abnormal questions in the test paper are identified, and text extraction is performed on the test paper image to obtain test paper text information, wherein the abnormal questions include questions with missing question score boxes and questions with score recognition confidence less than a preset score recognition confidence threshold;

[0031] Identify the related test papers matching the test paper text information from the test paper text library, and identify the related questions corresponding to the abnormal questions from the related test paper texts according to the question text similarity;

[0032] Retrieving the score of the related question as the expected score of the abnormal question, and performing test paper score verification according to the expected score, and if the test paper score verification result is passed, updating the expected score to the score of the abnormal question;

[0033] If the test paper score verification result is not passed, return to the step of extracting text from the test paper image and extracting text from the test paper image, and if the test paper score verification result is still not passed, configure the score of the abnormal question as an invalid score;

[0034] Among them, the test paper score abnormality includes that there is at least one question score box missing in the result of the target detection, or the test paper to be proofread fails the test paper score verification, and the test paper score verification includes that if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is passed, and if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is failed.

[0035] According to another aspect of the present invention, there is provided a proofreading device for marking test papers, comprising:

[0036] The acquisition module is used to acquire the test paper image of the test paper to be proofread, and perform target detection on the test paper image to obtain the correction frame in the test paper to be proofread, wherein the correction frame includes the test paper total score correction frame and the question correction frames corresponding to different questions respectively;

[0037] A recognition module is used to perform character recognition on the correction box, determine the total score of the test paper correction and the score of each question according to the recognition result, and calculate the total score of the test paper according to the scores of all the questions;

[0038] The matching module is used to generate a proofreading abnormality result of the test paper to be proofread if the total score of the test paper does not match the total score of the test paper correction.

[0039] Furthermore, the acquisition module includes:

[0040] A detection unit, used to perform target detection on the test paper image to obtain question frame coordinate information and correction frame coordinate information;

[0041] a calculation unit, for calculating, for any title frame, an intersection IOU value between the title frame and the correction frame according to the title frame coordinate information and the correction frame coordinate information, wherein the intersection IOU value is a ratio of an overlapping area between the title frame and the correction frame to an area of ​​the correction frame;

[0042] A first determination module is used to determine the correction box whose intersection IOU value is greater than the intersection IOU threshold as the title correction box of the title corresponding to the title box;

[0043] The second determination module is used to determine the correction box whose intersection IOU values ​​with all the question boxes are less than the intersection IOU threshold as the test paper total score correction box.

[0044] Furthermore, the device also includes:

[0045] A first cropping module is used to crop the question image according to the coordinate information of the question frame to obtain the question image when the score correction frame or the symbol correction frame corresponding to any question is missing;

[0046] A processing module, used for performing binarization processing on the question image to obtain an initial connected domain block image of the question, and performing denoising processing on the initial connected domain block image according to a denoising area threshold to obtain a target connected domain block image;

[0047] A second cropping module is used to obtain the horizontal coordinate boundary value of the correction box of the previous question and the vertical coordinate boundary value of the question, and to crop the target connected domain block image according to the horizontal coordinate boundary value and the vertical coordinate boundary value;

[0048] The recognition module is also used to recognize the cropped connected domain block image through a correction trace recognition model to obtain the correction symbol of the question, or to recognize the cropped connected domain block image through a score recognition model to obtain the correction score of the question.

[0049] Furthermore, the identification module includes:

[0050] A first configuration unit is configured to, for any of the questions, when the recognition result only includes a correction symbol, if the correction symbol is a check mark, configure the score of the question as the score of the question; if the correction symbol is a wrong mark, configure the score of the question as zero;

[0051] The second configuration unit is used to configure the correction score as the score of the question if the recognition result includes half correct and half wrong answers and a correction score. If the correction score is a positive number, the correction score is configured as the score of the question; if the correction score is a negative number, the sum of the correction score and the score of the question is configured as the score of the question.

[0052] Furthermore, the device also includes:

[0053] A symbol secondary recognition module is used for segmenting the image of the symbol correction box when the recognition result does not meet the score symbol matching strategy, and recognizing the segmented image through the correction trace recognition model to obtain a secondary recognition result of the symbol correction box, wherein the symbol matching strategy includes that the correction symbol is a correct or incorrect number, and the recognition result of the correction box does not include a correction score, and the correction symbol is half correct and half incorrect, and the recognition result of the correction box includes a correction score;

[0054] A score secondary recognition module, which is used to segment the image of the score correction box if the secondary recognition result of the symbol correction box still does not meet the score symbol matching strategy, and recognize the segmented image through a score recognition model to obtain a score correction box secondary recognition result;

[0055] The third determination module is used to determine the score of the question according to the secondary recognition result of the score correction box.

[0056] Furthermore, the third determination module includes:

[0057] A third configuration unit is used to configure the correction score as the score of the question if the absolute value of the correction score in the secondary recognition result of the score correction box is less than or equal to the score of the question;

[0058] The fourth configuration unit is used to retrieve the candidate correction scores in the secondary recognition results of the score correction box if the absolute value of the correction score in the secondary recognition results of the score correction box is greater than the score of the question, and configure the candidate correction score whose absolute value is less than or equal to the score of the question and whose confidence is the largest as the score of the question.

[0059] Furthermore, the device also includes:

[0060] A text extraction module is used to identify abnormal questions in the test paper when the test paper score is abnormal, and perform text extraction on the test paper image to obtain test paper text information, wherein the abnormal questions include questions with missing question score boxes and questions with score recognition confidence less than a preset score recognition confidence threshold;

[0061] The test paper search module is used to identify related test papers matching the test paper text information from the test paper text library, and identify related questions corresponding to abnormal questions from the related test paper text according to the similarity of the question text;

[0062] A first score verification module is used to retrieve the score of the related question as the expected score of the abnormal question, and perform test paper score verification according to the expected score. If the test paper score verification result is passed, the expected score is updated to the score of the abnormal question;

[0063] A second score verification module is used for returning to the step of extracting text from the test paper image and extracting text from the test paper image if the test paper score verification result is not passed, and if the test paper score verification result is still not passed, configuring the score of the abnormal question as an invalid score;

[0064] Among them, the test paper score abnormality includes that there is at least one question score box missing in the result of the target detection, or the test paper to be proofread fails the test paper score verification, and the test paper score verification includes that if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is passed, and if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is failed.

[0065] According to another aspect of the present invention, a storage medium is provided, wherein at least one executable instruction is stored in the storage medium, and the executable instruction enables a processor to execute operations corresponding to the above-mentioned method for proofreading test paper grading scores.

[0066] According to another aspect of the present invention, there is provided a terminal, comprising: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus;

[0067] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned method for proofreading the test paper marking scores.

[0068] By means of the above technical solution, the technical solution provided by the embodiment of the present invention has at least the following advantages:

[0069] The present invention provides a method and device for proofreading test paper correction scores, a storage medium, and a terminal. The embodiment of the present invention acquires a test paper image of a test paper to be proofread, and performs target detection on the test paper image to obtain correction frames in the test paper to be proofread, wherein the correction frames include a test paper total score correction frame and a question correction frame corresponding to different questions respectively; character recognition is performed on the correction frames, and the test paper correction total score and the scores of each question are determined according to the recognition result, and the test paper total score is calculated according to the scores of all the questions; if the test paper total score does not match the test paper correction total score, an abnormal proofreading result of the test paper to be proofread is generated, thereby avoiding the dependence of paper test paper correction score verification on manual labor, greatly reducing the time for test paper correction score verification, and ensuring the accuracy of test paper score proofreading, thereby effectively improving the proofreading efficiency of test paper correction score.

[0070] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0072] Figure 1 A flow chart of a method for checking test paper grading scores provided by an embodiment of the present invention is shown;

[0073] Figure 2 A flow chart of another method for checking test paper scores provided by an embodiment of the present invention is shown;

[0074] Figure 3 A block diagram showing the composition of a proofreading device for marking test papers provided by an embodiment of the present invention is shown;

[0075] Figure 4 A schematic diagram of the structure of a terminal provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0076] The exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0077] Aiming at the problem that the existing manual examination paper marking and score checking efficiency is low, the embodiment of the present invention provides a method for checking the examination paper marking and score, such as Figure 1 As shown, the method includes:

[0078] 101. Collect an image of the test paper to be proofread, and perform object detection on the image of the test paper to obtain a correction frame in the test paper to be proofread.

[0079] In an embodiment of the present invention, the test paper to be proofread is a paper test paper corrected by a teacher, and the test paper includes a correction frame, which includes a correction frame for the total score of the test paper and a correction frame for the title corresponding to different questions, and the correction frame is uniformly filled in with red handwriting. The title correction frame can specifically be a symbol correction frame for the teacher to fill in the correct number and the wrong number, or a score correction frame for the teacher to fill in the deduction score or the score score, and can also include a score correction frame and a symbol correction frame at the same time. The test paper image can be a photo or a color scanned image of the test paper to be proofread. After the test paper image is collected, the test paper image is preprocessed, such as scaling to a preset size, standardizing the image, etc., and then the target detection model is used to perform target detection on the preprocessed test paper image, and the correction frame in the test paper to be proofread (the coordinates and angle information of the minimum circumscribed rectangular frame of the correction frame) is output. Among them, the target detection model can be an R-CenterNet model, a yolov5 model, etc., and the embodiment of the present invention is not specifically limited.

[0080] 102. Perform character recognition on the correction box, determine the total score of the test paper correction and the score of each question according to the recognition result, and calculate the total score of the test paper according to the scores of all the questions.

[0081] In an embodiment of the present invention, after the correction frame is detected, character recognition is performed on the image in the correction frame to obtain the total score of the test paper in the correction frame of the total score of the test paper, and the correction symbol and / or score in the correction frame of each question. The score of each question is determined according to the correction symbol and / or score of each question, and then the scores of all questions are added up to obtain the total score of the test paper to be proofread. Among them, the character recognition process can specifically include: cutting each correction frame image based on the coordinates and angle information of the minimum circumscribed rectangular frame of the correction frame, and then preprocessing the collected image, such as scaling to a preset size, standardizing the image, etc., and then using the image recognition model to recognize the preprocessed correction frame image to obtain the total score of the test paper and the score of each question. Among them, the image recognition model can be a convolutional recurrent neural network (CRNN), or it can be other neural network models that can be used for image recognition, and the embodiment of the present invention does not make specific limitations.

[0082] 103. If the total score of the test paper does not match the total score of the corrected test paper, a proofreading exception result of the test paper to be proofread is generated.

[0083] In the embodiment of the present invention, after determining the total score of the test paper, the total score of the test paper is compared with the total score of the test paper correction. If the two are equal, it indicates that the total score of the test paper correction is correct, and the proofreading is passed; if the two are not equal, it indicates that the total score of the test paper correction is not equal to the total score of all questions, and there may be a score calculation error, and then the proofreading abnormal result of the test paper to be proofread is generated. This abnormal result can be sent to the terminal device to notify manual intervention to recheck the score of the current test paper.

[0084] It should be noted that by collecting test paper images of manually marked paper papers, extracting and identifying the marking boxes in the test paper images based on target detection and image recognition technology, and proofreading the question correction scores based on the comparison result of the sum of the scores of all questions and the total score of the test paper corrections, it is possible to realize automatic and batch proofreading of multiple test papers to be verified, which greatly reduces the manual intervention in the test paper proofreading process and the time consumption of test paper verification. At the same time, it ensures the accuracy of the test paper score proofreading, thereby effectively improving the efficiency of the test paper score proofreading.

[0085] In an embodiment of the present invention, for further explanation and limitation, as Figure 2 As shown, the step of performing target detection on the test paper image includes:

[0086] 201. Perform target detection on the test paper image to obtain question frame coordinate information and correction frame coordinate information.

[0087] 202. For any title frame, calculate the intersection IOU value between the title frame and the correction frame according to the coordinate information of the title frame and the coordinate information of the correction frame.

[0088] 203. Determine the correction box whose intersection IOU value is greater than the intersection IOU threshold as the title correction box of the title corresponding to the title box.

[0089] 204. Determine the correction box whose intersection IOU values ​​with all the question boxes are less than the intersection IOU threshold as the test paper total score correction box.

[0090] In an embodiment of the present invention, multiple correction boxes are obtained during target detection and extraction, and the subordinate relationship between the title and the correction box cannot be accurately determined, that is, the title and the correction box do not correspond. In the process of target detection, the title box of the title is extracted, that is, the result of target detection also includes the title box. Furthermore, for each title box, the intersection IOU value between it and different correction boxes is calculated respectively, and the correction boxes and title boxes whose intersection IOU values ​​are greater than the intersection IOU threshold are associated together, so that different correction boxes match the corresponding titles. Among them, the intersection IOU (Intersection over Union) value is the ratio of the overlapping area of ​​the title box and the correction box to the area of ​​the correction box, that is, the quotient of the overlapping area of ​​the title box and the correction box divided by the area of ​​the correction box. The intersection IOU threshold can be 60%, 70%, or it can be customized according to actual application requirements, and the embodiment of the present invention does not make specific limitations. The title box is the minimum circumscribed rectangular box covering the corresponding title text.

[0091] It should be noted that, unlike the online graded test paper (where the correspondence between the question and the correction box is clear), there is no correspondence between the question and the correction box in the paper image of the paper test paper. The skew or wrinkle of the paper will cause the correspondence between the correction box and the question to be misaligned. By extracting the question box in the test paper image and judging the correspondence between the correction box and the question box based on the intersection IOU value between the question box and the correction box, the correction box and the question are accurately corresponded, thereby improving the accuracy of object detection.

[0092] In an embodiment of the present invention, for further explanation and limitation, the method further includes:

[0093] In the case where the score correction box or the symbol correction box corresponding to any question is missing, the question image is cropped according to the coordinate information of the question box to obtain the question image;

[0094] Binarizing the question image to obtain an initial connected domain block image of the question, and denoising the initial connected domain block image according to a denoising area threshold to obtain a target connected domain block image;

[0095] Obtaining the horizontal coordinate boundary value of the correction box of the previous question and the vertical coordinate boundary value of the question, and cropping the target connected domain block image according to the horizontal coordinate boundary value and the vertical coordinate boundary value;

[0096] The cropped connected domain block image is recognized by a correction trace recognition model to obtain the correction symbol of the question, or the cropped connected domain block image is recognized by a score recognition model to obtain the correction score of the question.

[0097] In an embodiment of the present invention, the question correction box includes a score correction box and / or a symbol correction box. During the target detection process, there is a situation where the target detection model fails to detect, resulting in the missing of the score correction box or the symbol correction box. After matching the question box with the correction box, if it is found that the score correction box or the symbol correction box of a certain question is missing, it is necessary to extract the connected domain block image in the correction box of the current question, and identify the connected domain block image through the corresponding model to obtain the recognition result of the correction box. Specifically, if the missing is the symbol correction box, the cropped connected domain block image is identified using the correction trace recognition model, and if the missing is the score correction box, the cropped connected domain block image is identified using the score recognition model. Among them, the correction trace recognition model can be MobileNetV3, and the score recognition model can be a CRNN model, which is not specifically limited in the embodiment of the present invention. The cropped connected domain block image extraction specifically includes: cropping the image of the question from the test paper image according to the coordinate information of the question frame to obtain the question image; binarizing the question image to obtain a binary image, extracting the connected domain in the binary image, removing the noise block, and obtaining the target connected domain block image. Then, the horizontal coordinate boundary of the correction box of the previous question of the current question and the vertical coordinate boundary of the current question are used as range constraints to further crop the target connected domain block image to obtain the cropped connected domain block image. For example, the symbol correction box in the test paper is on the left side of the score correction box, and the question correction box of question A is missing. Then the rightmost coordinate of the symbol correction box of the question above the question is taken as the left boundary, the rightmost coordinate of the score correction box is taken as the right boundary, the upper limit coordinate of the question box of question A is taken as the upper boundary, and the lower limit coordinate of the question box is taken as the lower boundary. Traverse the connected domain blocks in the rectangular area of ​​the left boundary, upper boundary, right boundary and lower boundary obtained in the fourth step, and crop the connected domain blocks to obtain the cropped connected domain block image for identification. Among them, the binarization processing can adopt OTSU binarization, or other binarization methods can be adopted; to remove noise blocks, blocks with an area smaller than an area threshold (1 / 2 the average area of ​​the correction box) can be regarded as noise blocks, or a custom area threshold can be used for denoising, which is not specifically limited in the embodiment of the present invention.

[0098] It should be noted that the secondary search and recognition process of the missing correction box can be performed multiple times according to the preset search times. When the search times reach the preset search times and still cannot be found, if the missing is the symbol correction box, the current question correction is determined to be invalid, and the score box score is set to 0; if the missing is the score correction box, the recognition result of the score correction box is configured as blank, and the current question score box recognition result is processed as blank. By performing a secondary search for the missing correction box, the impact of the target detection model detection failure on the proofreading process can be greatly reduced, ensuring the success rate and accuracy of the proofreading process.

[0099] In one embodiment of the present invention, for further explanation and limitation, the step of performing character recognition on the correction box and determining the total score of the test paper correction and the score of each question according to the recognition result includes:

[0100] For any of the questions, when the recognition result only includes a correction symbol, if the correction symbol is a check mark, the score of the question is configured as the score of the question; if the correction symbol is a wrong check mark, the score of the question is configured as zero;

[0101] In the case where the recognition result includes half correct and half incorrect answers and a correction score, and in the case where the recognition result only includes a correction score, if the correction score is a positive number, the correction score is configured as the score of the question; if the correction score is a negative number, the sum of the correction score and the score of the question is configured as the score of the question.

[0102] In the embodiment of the present invention, the recognition result includes the total score of the test paper correction, and the correction symbol and / or correction score of each question. The correction symbol includes a correct number, a wrong number and a half-correct and half-wrong number, and the correction score includes a positive number and a negative number. When the correction symbol is a correct number and a wrong number, and the score of the current question is 0 points or a full score, the recognition result of the score correction box should be empty. When the correction symbol is half right and half wrong, and the correction symbol is not recognized, and only the correction score is recognized, the correction score is the score that the correction teacher determines to be obtained or the score that should be deducted for this question, and the correction score should be a positive number or a negative number. When the correction score is a positive number, the correction score is the score of the current question, and when the correction score is a negative number, the correction score is the deduction of the current question, and the score of the question is added to the negative number to obtain the score of this question. Wherein, the score of the question can be pre-set, or it can be obtained by identifying the target detection and score recognition model, and the embodiment of the present invention does not make specific restrictions.

[0103] It should be noted that in order to improve the efficiency of grading, the way of grading the test papers is often different for different students. For example, for students with good academic performance, generally the scores are relatively high. When grading, the method of deducting points is often used for grading, that is, when grading, negative scores are filled in the grading score box on the right side of the wrong question. For students with average grades, the deduction and scoring grading methods may be used alternately according to the scores of specific questions. For students with poor academic performance, teachers generally use the scoring method for grading. By configuring the joint judgment logic of the score correction box and the symbol correction box, from the perspective of grading, the restrictions on the teacher's grading method are greatly reduced, allowing teachers to flexibly adopt different grading methods for different questions and papers with different scores. From the perspective of proofreading, the compatibility of different types of grading is greatly improved, thereby greatly improving the applicability of test paper proofreading.

[0104] In an embodiment of the present invention, for further explanation and limitation, the method further includes:

[0105] In the case where the recognition result does not satisfy the score symbol matching strategy, the image of the symbol correction frame is segmented using the Sobel operator, and the segmented image is recognized using the correction trace recognition model to obtain a secondary recognition result of the symbol correction frame;

[0106] If the secondary recognition result of the symbol correction box still does not meet the score symbol matching strategy, the image of the score correction box is segmented by using the Sobel operator, and the segmented image is recognized by the score recognition model to obtain the secondary recognition result of the score correction box;

[0107] The score of the question is determined according to the secondary recognition result of the score correction box.

[0108] In the embodiment of the present invention, the score symbol matching strategy includes that the correction symbol is a correct or wrong number, and the recognition result of the correction box does not include the correction score, and the correction symbol is half correct and half wrong, and the recognition result of the correction box includes the correction score, that is, when the recognition is correct (the symbol and the score match), when the correction symbol is a correct or wrong number, the recognition result of the correction box should not include the correction score, and when the correction symbol is half correct and half wrong, the recognition result should include the correction score. If the above conditions are not met, it means that the recognition result of at least one of the symbol correction box and the score correction box is abnormal, and the correction box needs to be recognized again. In the secondary recognition process of the correction box, the image of the symbol correction box is segmented first, and the correction trace recognition model is used to recognize the segmented image. If the secondary recognition result of the symbol correction box meets the above conditions, that is, when the recognition result does not include the correction score, the secondary recognition result of the symbol correction box is correct or wrong; when the recognition result includes the correction score, the secondary recognition result of the symbol correction box is half correct and half wrong, then the score of the question can be determined based on the secondary recognition result of the symbol correction box. If the secondary recognition result of the symbol correction box still does not meet the score symbol matching strategy, the image of the score correction box is segmented, and the score recognition model is used to recognize the secondary recognition result of the score correction box on the segmented image, so as to determine the score of the current question based on the secondary recognition result of the score correction box. Among them, the image segmentation method can be a Sobel operator or other image segmentation algorithms, which are not specifically limited in the embodiments of the present invention.

[0109] It should be noted that the cross-check of the correction symbols and correction scores through the score symbol matching strategy can accurately identify the matching anomalies between the correction symbols and the correction scores, so as to make timely corrections. At the same time, the secondary recognition of the symbol correction box is given priority, which also fully considers the characteristics of the symbol correction box being difficult to recognize and having a higher recognition error rate than the score correction box, and minimizes the image processing amount of the secondary recognition, thereby improving the correction efficiency of the correction recognition anomaly.

[0110] In one embodiment of the present invention, for further explanation and limitation, the step of determining the score of the question according to the secondary recognition result of the score correction box includes:

[0111] If the absolute value of the correction score in the secondary recognition result of the score correction box is less than or equal to the score of the question, the correction score is configured as the score of the question;

[0112] If the absolute value of the correction score in the secondary recognition result of the score correction box is greater than the score of the question, the candidate correction score in the secondary recognition result of the score correction box is retrieved, and the candidate correction score whose absolute value is less than or equal to the score of the question and whose confidence is the largest is configured as the score of the question.

[0113] In the embodiment of the present invention, the correction score can be a positive number or a negative number, but its absolute value will never exceed the score of the current question. Therefore, if the absolute value of the correction score obtained by the secondary recognition of the score correction box is less than or equal to the score of the question, it indicates that the recognition is correct, and the correction score can be configured as the score of the question; if the absolute value of the correction score obtained by the secondary recognition of the score correction box is greater than the score of the question, it indicates that the recognition is abnormal. Since the output of the score recognition model includes multiple recognition results corresponding to different confidence levels during the image recognition process, when the recognition result with the highest confidence level, that is, the correction score recognition is abnormal, multiple candidate corrections with absolute values ​​less than or equal to the score of the question are screened out from the remaining recognition results, and the candidate correction score with the highest confidence level is taken as the correct recognition result of the correction score box, and the score of the current question is configured with this candidate correction score.

[0114] It should be noted that the above-mentioned verification of the recognition result of the score correction box is not limited to the secondary recognition process, but can also be used for character recognition of the correction box, and the process of determining the total score of the test paper correction and the score of each question according to the recognition result, which is not repeated in this application. After the cross-validation of the symbol correction box and the score correction box, the secondary recognition result of the score correction box is further verified, which can avoid proofreading errors caused by abnormal score recognition, thereby further improving the accuracy of score recognition and proofreading.

[0115] In an embodiment of the present invention, for further explanation and limitation, the method further includes:

[0116] In the case of abnormal test paper scores, abnormal questions in the test paper are identified, and text is extracted from the test paper image to obtain test paper text information;

[0117] Identify the related test papers matching the test paper text information from the test paper text library, and identify the related questions corresponding to the abnormal questions from the related test paper texts according to the question text similarity;

[0118] Retrieving the score of the related question as the expected score of the abnormal question, and performing test paper score verification according to the expected score, and if the test paper score verification result is passed, updating the expected score to the score of the abnormal question;

[0119] In the case that the test paper score verification result is failed, return to the step of extracting text from the test paper image, and if the test paper score verification result is still failed, the score of the abnormal question is configured as an invalid score.

[0120] In the embodiment of the present invention, the result of the target detection also includes a total score frame of the test paper and a question score frame corresponding to different questions, and the recognition result of the question score frame includes the scores of different questions and the total score of the test paper. The target detection method of the question score frame is the same as the target detection method of the correction frame. The recognition of the question score frame adopts the same score recognition model as the score correction frame, and the embodiment of the present invention will not be repeated here. After obtaining the target detection result, the question score frame of each question is also traversed to see if it is missing. If the question score frame of one or more questions is missing, the test paper score is determined to be abnormal, and the questions with missing question score frames are regarded as abnormal questions. In addition, after the question score frame and the total score frame of the test paper are identified, the test paper score verification is also performed, that is, it is determined whether the sum of the scores of all questions in the test paper to be proofread is equal to the total score of the test paper. If it is equal, the score verification passes, and if it is not equal, the score verification fails. When the score check fails, the test paper score is determined to be abnormal, and the questions with score recognition confidence less than the preset score recognition confidence threshold are determined as abnormal questions. The test paper image of the current test paper is recognized by optical character recognition (OCR) to obtain the test paper text information, so as to match the related test paper with the highest content overlap from the test paper text library according to the test paper text information. Specifically, the matching can be performed by calculating the hash value or text similarity of the two test paper texts.

[0121] After determining the associated test paper, calculate the text similarity between the abnormal question and each question in the associated test paper, take the question with the highest text similarity as the associated question of the abnormal question, and retrieve the question score of the associated question as the expected score of the abnormal question, and perform the test paper score verification based on the expected score and the scores of other questions in the current test paper. If the verification passes, the expected score is used as the score of the abnormal question; if the verification fails, return to the step of identifying the abnormal question in the test paper and extracting text from the test paper image. If the verification result is still recognition, the score of the abnormal question is configured as an invalid score. In the case of an invalid question score, if the correction box recognition result includes the correction score, and the score is a positive number, the score of this question can still be identified. In other cases, it cannot be extracted.

[0122] It should be noted that the object detection and image recognition methods can avoid presetting the score of the questions, further improve the convenience of the test paper score proofreading, and reduce the configuration time in the early stage of the test paper proofreading. By checking the score of the questions, the automatic proofreading and correction of the score of the questions can be realized, which greatly improves the accuracy of the recognition of the score of the questions, thereby improving the efficiency of the score proofreading.

[0123] The present invention provides a method for proofreading test paper correction scores. The embodiment of the present invention collects a test paper image of a test paper to be proofread, and performs target detection on the test paper image to obtain correction frames in the test paper to be proofread, wherein the correction frames include a test paper total score correction frame and question correction frames corresponding to different questions respectively; character recognition is performed on the correction frames, and the test paper correction total score and the scores of each question are determined according to the recognition result, and the test paper total score is calculated according to the scores of all the questions; if the test paper total score does not match the test paper correction total score, an abnormal proofreading result of the test paper to be proofread is generated, thereby avoiding the dependence of paper test paper correction score verification on manual labor, greatly reducing the time for test paper correction score verification, and ensuring the accuracy of test paper score proofreading, thereby effectively improving the proofreading efficiency of test paper correction score.

[0124] Furthermore, as a response to the above Figure 1 The embodiment of the present invention provides a proofreading device for marking test papers, such as Figure 3 As shown, the device comprises:

[0125] The acquisition module 31 is used to acquire the test paper image of the test paper to be proofread, and perform target detection on the test paper image to obtain the correction frame in the test paper to be proofread, wherein the correction frame includes the test paper total score correction frame and the question correction frames corresponding to different questions respectively;

[0126] The recognition module 32 is used to perform character recognition on the correction box, determine the total score of the test paper correction and the score of each question according to the recognition result, and calculate the total score of the test paper according to the scores of all the questions;

[0127] The matching module 33 is used to generate a proofreading abnormality result of the test paper to be proofread if the total score of the test paper does not match the total score of the test paper correction.

[0128] Furthermore, the acquisition module 31 includes:

[0129] A detection unit, used to perform target detection on the test paper image to obtain question frame coordinate information and correction frame coordinate information;

[0130] a calculation unit, for calculating, for any title frame, an intersection IOU value between the title frame and the correction frame according to the title frame coordinate information and the correction frame coordinate information, wherein the intersection IOU value is a ratio of an overlapping area between the title frame and the correction frame to an area of ​​the correction frame;

[0131] A first determination module is used to determine the correction box whose intersection IOU value is greater than the intersection IOU threshold as the title correction box of the title corresponding to the title box;

[0132] The second determination module is used to determine the correction box whose intersection IOU values ​​with all the question boxes are less than the intersection IOU threshold as the test paper total score correction box.

[0133] Furthermore, the device also includes:

[0134] A first cropping module is used to crop the question image according to the coordinate information of the question frame to obtain the question image when the score correction frame or the symbol correction frame corresponding to any question is missing;

[0135] A processing module, used for performing binarization processing on the question image to obtain an initial connected domain block image of the question, and performing denoising processing on the initial connected domain block image according to a denoising area threshold to obtain a target connected domain block image;

[0136] A second cropping module is used to obtain the horizontal coordinate boundary value of the correction box of the previous question and the vertical coordinate boundary value of the question, and to crop the target connected domain block image according to the horizontal coordinate boundary value and the vertical coordinate boundary value;

[0137] The recognition module 32 is further used to recognize the cropped connected domain block image through a correction trace recognition model to obtain the correction symbol of the question, or to recognize the cropped connected domain block image through a score recognition model to obtain the correction score of the question.

[0138] Furthermore, the identification module includes:

[0139] A first configuration unit is configured to, for any of the questions, when the recognition result only includes a correction symbol, if the correction symbol is a check mark, configure the score of the question as the score of the question; if the correction symbol is a wrong mark, configure the score of the question as zero;

[0140] The second configuration unit is used to configure the correction score as the score of the question if the recognition result includes half correct and half wrong answers and a correction score. If the correction score is a positive number, the correction score is configured as the score of the question; if the correction score is a negative number, the sum of the correction score and the score of the question is configured as the score of the question.

[0141] Furthermore, the device also includes:

[0142] A symbol secondary recognition module is used for segmenting the image of the symbol correction box when the recognition result does not meet the score symbol matching strategy, and recognizing the segmented image through the correction trace recognition model to obtain a secondary recognition result of the symbol correction box, wherein the symbol matching strategy includes that the correction symbol is a correct or incorrect number, and the recognition result of the correction box does not include a correction score, and the correction symbol is half correct and half incorrect, and the recognition result of the correction box includes a correction score;

[0143] A score secondary recognition module, for segmenting the image of the score correction box if the secondary recognition result of the symbol correction box still fails to meet the score symbol matching strategy, and identifying the segmented image through a score recognition model to obtain a score correction box secondary recognition result;

[0144] The third determination module is used to determine the score of the question according to the secondary recognition result of the score correction box.

[0145] Furthermore, the third determination module includes:

[0146] A third configuration unit is used to configure the correction score as the score of the question if the absolute value of the correction score in the secondary recognition result of the score correction box is less than or equal to the score of the question;

[0147] The fourth configuration unit is used to retrieve the candidate correction scores in the secondary recognition results of the score correction box if the absolute value of the correction score in the secondary recognition results of the score correction box is greater than the score of the question, and configure the candidate correction score whose absolute value is less than or equal to the score of the question and whose confidence is the largest as the score of the question.

[0148] Furthermore, the device also includes:

[0149] A text extraction module is used to identify abnormal questions in the test paper when the test paper score is abnormal, and perform text extraction on the test paper image to obtain test paper text information, wherein the abnormal questions include questions with missing question score boxes and questions with score recognition confidence less than a preset score recognition confidence threshold;

[0150] The test paper search module is used to identify related test papers matching the test paper text information from the test paper text library, and identify related questions corresponding to abnormal questions from the related test paper text according to the similarity of the question text;

[0151] A first score verification module is used to retrieve the score of the related question as the expected score of the abnormal question, and perform test paper score verification according to the expected score. If the test paper score verification result is passed, the expected score is updated to the score of the abnormal question;

[0152] A second score verification module is used for returning to the step of extracting text from the test paper image and extracting text from the test paper image if the test paper score verification result is not passed, and if the test paper score verification result is still not passed, configuring the score of the abnormal question as an invalid score;

[0153] Among them, the test paper score abnormality includes that there is at least one question score box missing in the result of the target detection, or the test paper to be proofread fails the test paper score verification, and the test paper score verification includes that if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is passed, and if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is failed.

[0154] The present invention provides a proofreading device for test paper correction scores. The embodiment of the present invention acquires a test paper image of a test paper to be proofread, and performs target detection on the test paper image to obtain correction frames in the test paper to be proofread, wherein the correction frames include a test paper total score correction frame and question correction frames corresponding to different questions respectively; character recognition is performed on the correction frames, and the test paper correction total score and the scores of each question are determined according to the recognition result, and the test paper total score is calculated according to the scores of all the questions; if the test paper total score does not match the test paper correction total score, an abnormal proofreading result of the test paper to be proofread is generated, thereby avoiding the dependence of paper test paper correction score verification on manual labor, greatly reducing the time of test paper correction score verification, and ensuring the accuracy of test paper score proofreading, thereby effectively improving the proofreading efficiency of test paper correction score.

[0155] According to one embodiment of the present invention, a storage medium is provided, wherein the storage medium stores at least one executable instruction, and the computer executable instruction can execute the method for checking the examination paper correction score in any of the above method embodiments.

[0156] Figure 4 A schematic diagram of the structure of a terminal provided according to an embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the terminal.

[0157] like Figure 4 As shown, the terminal may include: a processor (processor) 402 , a communication interface (Communications Interface) 404 , a memory (memory) 406 , and a communication bus 408 .

[0158] The processor 402 , the communication interface 404 , and the memory 406 communicate with each other via a communication bus 408 .

[0159] The communication interface 404 is used for network communication with other devices such as a client or other servers.

[0160] The processor 402 is used to execute the program 410, and specifically can execute the relevant steps in the above-mentioned embodiment of the method for checking the test paper marking score.

[0161] Specifically, the program 410 may include program codes, which include computer operation instructions.

[0162] The processor 402 may be a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiment of the present invention. The one or more processors included in the terminal may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0163] The memory 406 is used to store the program 410. The memory 406 may include a high-speed RAM memory, and may also include a non-volatile memory (non-volatile memory), such as at least one disk memory.

[0164] The program 410 may be specifically configured to enable the processor 402 to perform the following operations:

[0165] Collecting an image of the test paper to be proofread, and performing target detection on the image of the test paper to obtain correction frames in the test paper to be proofread, wherein the correction frames include a correction frame for the total score of the test paper and question correction frames corresponding to different questions;

[0166] Performing character recognition on the correction box, determining the total score of the test paper correction and the score of each question according to the recognition result, and calculating the total score of the test paper according to the scores of all the questions;

[0167] If the total score of the test paper does not match the total score of the test paper correction, a proofreading exception result of the test paper to be proofread is generated.

[0168] Obviously, those skilled in the art should understand that the above modules or steps of the present invention can be implemented by a general computing device, they can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be implemented by a program code executable by a computing device, so that they can be stored in a storage device and executed by the computing device, and in some cases, the steps shown or described can be executed in a different order than here, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0169] 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 method for proofreading test paper grading, characterized in that: include: Collecting an image of the test paper to be proofread, and performing target detection on the test paper image to obtain correction frames in the test paper to be proofread, wherein the correction frames include a correction frame for the total score of the test paper and correction frames for different questions respectively; Performing character recognition on the correction box, determining the total score of the test paper correction and the score of each question according to the recognition result, and calculating the total score of the test paper according to the scores of all the questions; If the total score of the test paper does not match the total score of the test paper correction, a proofreading exception result of the test paper to be proofread is generated.

2. The method according to claim 1, characterized in that The result of the target detection also includes a question frame, and the target detection is performed on the test paper image to obtain the correction frame in the test paper to be proofread, including: Performing target detection on the test paper image to obtain question frame coordinate information and correction frame coordinate information; For any title frame, the intersection IOU value between the title frame and the correction frame is calculated according to the coordinate information of the title frame and the coordinate information of the correction frame, wherein the intersection IOU value is the ratio of the overlapping area between the title frame and the correction frame to the area of ​​the correction frame; Determine the correction box whose intersection IOU value is greater than the intersection IOU threshold as the title correction box of the title corresponding to the title box; The correction box whose intersection IOU values ​​with all the question boxes are less than the intersection IOU threshold is determined as the correction box for the total score of the test paper.

3. The method according to claim 2, characterized in that The question correction box includes a score correction box and / or a symbol correction box, and the method further includes: In the case where the score correction box or the symbol correction box corresponding to any question is missing, the question image is cropped according to the coordinate information of the question box to obtain the question image; Binarizing the question image to obtain an initial connected domain block image of the question, and denoising the initial connected domain block image according to a denoising area threshold to obtain a target connected domain block image; Obtaining the horizontal coordinate boundary value of the correction box of the previous question and the vertical coordinate boundary value of the question, and cropping the target connected domain block image according to the horizontal coordinate boundary value and the vertical coordinate boundary value; The cropped connected domain block image is recognized by a correction trace recognition model to obtain the correction symbol of the question, or the cropped connected domain block image is recognized by a score recognition model to obtain the correction score of the question.

4. The method according to claim 1, characterized in that The recognition result includes the total score of the test paper, and the correction symbol and / or correction score of each question. The character recognition of the correction box is performed, and the total score of the test paper and the score of each question are determined according to the recognition result, including: For any of the questions, when the recognition result only includes a correction symbol, if the correction symbol is a check mark, the score of the question is configured as the score of the question; if the correction symbol is a wrong check mark, the score of the question is configured as zero; In the case where the recognition result includes half correct and half wrong answers and a correction score, if the correction score is a positive number, the correction score is configured as the score of the question; if the correction score is a negative number, the sum of the correction score and the score of the question is configured as the score of the question.

5. The method according to claim 4, characterized in that The method further comprises: In the case that the recognition result does not satisfy the score symbol matching strategy, the image of the symbol correction box is segmented, and the segmented image is recognized by the correction trace recognition model to obtain a secondary recognition result of the symbol correction box, wherein the segmented symbol matching strategy includes that the correction symbol is a correct or incorrect number, and the recognition result of the correction box does not include a correction score, and the correction symbol is half correct and half incorrect, and the recognition result of the correction box includes a correction score; If the secondary recognition result of the symbol correction box still does not satisfy the score symbol matching strategy, segment the image of the score correction box, and recognize the segmented image through the score recognition model to obtain the secondary recognition result of the score correction box; The score of the question is determined according to the secondary recognition result of the score correction box.

6. The method according to claim 5, characterized in that Determining the score of the question according to the secondary recognition result of the score correction box includes: If the absolute value of the correction score in the secondary recognition result of the score correction box is less than or equal to the score of the question, the correction score is configured as the score of the question; If the absolute value of the correction score in the secondary recognition result of the score correction box is greater than the score of the question, the candidate correction score in the secondary recognition result of the score correction box is retrieved, and the candidate correction score whose absolute value is less than or equal to the score of the question and whose confidence is the largest is configured as the score of the question.

7. The method according to any one of claims 1 to 6, characterized in that The target detection result also includes a total score frame of the test paper and a question score frame corresponding to different questions, and the recognition result of the question score frame includes the scores of different questions and the total score of the test paper. The method also includes: In the case of abnormal test paper scores, abnormal questions in the test paper are identified, and text extraction is performed on the test paper image to obtain test paper text information, wherein the abnormal questions include questions with missing question score boxes and questions with score recognition confidence less than a preset score recognition confidence threshold; Identify the related test papers matching the test paper text information from the test paper text library, and identify the related questions corresponding to the abnormal questions from the related test paper texts according to the question text similarity; Retrieving the score of the related question as the expected score of the abnormal question, and performing test paper score verification according to the expected score, and if the test paper score verification result is passed, updating the expected score to the score of the abnormal question; If the test paper score verification result is not passed, return to the step of extracting text from the test paper image and extracting text from the test paper image, and if the test paper score verification result is still not passed, configure the score of the abnormal question as an invalid score; Among them, the test paper score abnormality includes that there is at least one question score box missing in the result of the target detection, or the test paper to be proofread fails the test paper score verification, and the test paper score verification includes that if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is passed, and if the sum of the scores of all questions is equal to the total score of the test paper, then the verification is failed.

8. A proofreading device for marking test papers, characterized in that: include: The acquisition module is used to acquire the test paper image of the test paper to be proofread, and perform target detection on the test paper image to obtain the correction frame in the test paper to be proofread, wherein the correction frame includes the test paper total score correction frame and the question correction frames corresponding to different questions respectively; A recognition module is used to perform character recognition on the correction box, determine the total score of the test paper correction and the score of each question according to the recognition result, and calculate the total score of the test paper according to the scores of all the questions; The matching module is used to generate a proofreading abnormality result of the test paper to be proofread if the total score of the test paper does not match the total score of the test paper correction.

9. A storage medium storing at least one executable instruction, wherein the executable instruction enables a processor to execute operations corresponding to the method for checking test paper marking scores as described in any one of claims 1 to 7.

10. A terminal, comprising: A processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and the executable instruction enables the processor to perform operations corresponding to the method for proofreading test paper marking scores as described in any one of claims 1-7.

Images