Student homework automatic correcting and scoring system based on image recognition

Abstract

The invention discloses an automatic student homework correcting and scoring system based on image recognition, relates to the technical field of education, and solves the problem that an existing automatic correcting and scoring system is difficult to recognize and compare characters and standard answers after collecting and processing homework images in real time. The similarity rate and step scores are difficult to accurately calculate and automatically correct and score; meanwhile, the error type is difficult to analyze and the error rate is difficult to calculate; the total score and the error rate are difficult to comprehensively evaluate the homework level of the student. The system comprises the following modules: an image acquisition module, an image processing module, an image recognition module, an image analysis module and a job rating module. And by calculating the similarity and step scores, automatic correction and scoring are realized. Meanwhile, the error rate is calculated according to the analysis error type, and the student homework level is evaluated by integrating the total score and the error rate. In addition, the teacher can adjust the score weight to realize flexible scoring according to teaching requirements.

Description

Technical Field

[0001] The invention belongs to the field of educational technology, and in particular is an automatic marking and grading system for student homework based on image recognition. Background Art

[0002] With the continuous development of educational informatization, the student population is growing, and the workload of teachers is increasing accordingly. The present invention collects and processes student homework image data in real time, identifies the characters in the image data, and compares and analyzes it with the standard answer. According to the results of the comparison and analysis, the student homework is automatically corrected and scored; at the same time, the error types in the student homework are analyzed and the error rate of the student homework is calculated; according to the total score and error rate of the student homework, the grade of the student homework is comprehensively evaluated. Automatic correction and scoring of student homework is realized.

[0003] The existing automatic grading and marking system for student assignments has difficulty in identifying and comparing characters in images with standard answers after real-time collection and processing of student assignment image data; it is difficult to accurately calculate similarity rates and step scores for automatic grading and marking of student assignments; at the same time, it is difficult to calculate the error rate of student assignments based on the analyzed error types; in addition, it is difficult to effectively use total scores and error rates to comprehensively assess the grade of student assignments. Summary of the invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art; to this end, the present invention proposes an automatic grading and marking system for student assignments based on image recognition, which is used to solve the technical problems that the existing automatic grading and marking systems are difficult to identify and compare characters with standard answers after real-time acquisition and processing of assignment images; difficult to accurately calculate similarity rates and step scores for automatic grading and marking; at the same time, difficult to analyze error types and calculate error rates; and difficult to use total scores and error rates to comprehensively assess student assignment grades.

[0005] To solve the above problems, the first aspect of the present invention provides a student homework automatic marking and grading system based on image recognition, comprising the following modules:

[0006] Image acquisition module: real-time acquisition of image data of students’ homework;

[0007] Image processing module: pre-processing based on the collected student homework image data, including: image denoising, image enhancement, conversion into grayscale and binarization, and adjustment of image size and direction;

[0008] Image recognition module: recognize characters in image data based on processed student homework image data;

[0009] Image analysis module: compare and analyze the handwritten characters in the recognized image data with the standard answers and automatically correct and score the students' homework; analyze the types of errors in the students' homework; calculate the error rate in the students' homework according to the error types;

[0010] Assignment Grading Module: Comprehensively evaluate the student assignment grade based on the total score and error rate of the student assignment.

[0011] As a further solution of the present invention: recognizing characters in image data according to processed student homework image data comprises the following steps:

[0012] According to the preprocessed student homework image data, the characters in the student homework image data are divided into handwritten characters and printed characters, and the characters in the student homework image data are recognized and analyzed through a deep learning model;

[0013] Annotate handwritten and printed characters in student work image data;

[0014] Input the annotated handwritten characters and printed characters into the deep learning model for model training;

[0015] The student homework automatic grading system collects and processes student homework image data in real time and inputs it into the trained deep learning model to recognize handwritten and printed characters;

[0016] The recognized handwritten characters in the student's homework image data are output.

[0017] As a further solution of the present invention, comparing and analyzing the handwritten characters in the recognized image data with the standard answers and automatically correcting and grading the student's homework includes the following steps:

[0018] S1: Convert the handwritten characters in the standard answers and student homework image data into a unified format, including: removing irrelevant symbols, unifying uppercase and lowercase letters, and character encoding;

[0019] S2: Use the Levenshtein distance algorithm to compare the student answer and the standard answer, where the student answer string set is represented by X n = {X1, X2, ..., X n}, the standard answer string set is represented by Y m ={Y1, Y2, ..., Y m};

[0020] S3: Calculate the similarity μ between the student's answer and the standard answer and score the student's homework;

[0021] S4: Calculate the step score of the student's answer;

[0022] S5: The student's homework score is calculated based on the similarity μ score and the step score.

[0023] As a further solution of the present invention: the step S2 comprises the following steps:

[0024] Calculated by analytical formula:

[0025]

[0026] Get the Levenshtein distance lev between the student answer string and the standard answer string X,Y (|X|,|Y|), where |X| and |Y| represent the length of the student answer string and the standard answer string respectively; lev X,Y (n,m) is the distance between the first n characters of X and the first m characters of Y.

[0027] As a further solution of the present invention: Step S3 comprises the following steps:

[0028] Calculated by analytical formula:

[0029] Get the similarity μ between the student's answer string and the standard answer; according to the similarity μ, set different levels of thresholds and divide the scores into:

[0030] When μ ≥ 90%, the student's assignment gets full marks:

[0031] When 60% ≤ μ < 90%, the student's work is partially scored;

[0032] When μ<60%, the student's assignment will not be scored.

[0033] As a further solution of the present invention: the step S4 comprises the following steps:

[0034] Calculated by analytical formula:

[0035] The step score is F, where n is the total number of steps and f i represents the score of the ith step, w i represents the weight of the i-th step;

[0036] When μ ≥ 90% of the steps, the step score is F;

[0037] When 60% ≤ μ < 90% of the steps, the step score is

[0038] When μ < 60% of the steps, the step score is 0.

[0039] As a further solution of the present invention: the step S5 comprises the following steps:

[0040] Calculated by analytical formula:

[0041] Get the total score of the student's homework Total_score; where W μ is the weight of similarity μ, W F is the weight of the step score F.

[0042] As a further solution of the present invention: analyzing the error types in student assignments includes the following steps:

[0043] Based on the graded student assignments, the deep learning model is used to identify and analyze the types of errors in the student assignments.

[0044] Mark the a-th error in the student's homework, where the a-th error includes the first knowledge error, the second method error, the third logic error, the fourth psychological error and the fifth negligence error;

[0045] Input the labeled a-th error into the deep learning model for model training, where a∈{1, 2, …, n};

[0046] The student homework image data is graded by the student homework automatic grading system and input into the trained deep learning model to identify the ath type of error in the student homework.

[0047] As a further solution of the present invention: calculating the error rate in the student's homework according to the error type includes the following steps:

[0048] Calculated by analytical formula:

[0049] Get the error rate α in the student's homework; where w a represents the weight of the a-th error, a j represents the number of errors, n is the total number of error types, Y m Indicates the standard answer string.

[0050] As a further solution of the present invention: comprehensively judging the grade of student assignments according to the total score and error rate of the student assignments, the following steps are included:

[0051] Calculated by analytical formula:

[0052] Q=(W*Total_score+(1-W)*α)×100%

[0053] Get the comprehensive score of the student assignment Q, where the weight W of the total score is 0.8 and the weight (1-W) of the error rate α is 0.2;

[0054] According to the comprehensive score of student assignment Q, different levels of thresholds are set to divide student assignments into five grades: A, B, C, D, and E:

[0055] When Q ≥ 80%, the assignment grade is A;

[0056] When 60% < Q < 80%, the student assignment grade is B;

[0057] When 40% ≤ Q ≤ 60%, the student assignment grade is C;

[0058] When 20% < Q < 40%, the student assignment grade is D;

[0059] When Q ≤ 20%, the student assignment grade is E.

[0060] Compared with the prior art, the beneficial effects of the present invention are:

[0061] The present invention discloses an automatic marking and scoring system for student assignments based on image recognition. By collecting and processing the assignment image data of students in real time, identifying and comparing the characters in the image with the standard answers, the acquisition of comparison results is realized. According to the comparison results, the similarity and step scores are calculated, and then the automatic marking and scoring of student assignments are realized. At the same time, according to the analyzed error types, the error rate of student assignments is calculated. In addition, by comprehensively considering the total score of the assignment, the error rate, and the corresponding weights, the comprehensive evaluation of the student assignment grade is realized, and the comprehensive evaluation of the student assignment is realized. The present invention enables teachers to flexibly adjust the scoring criteria according to the nature of the assignment and the teaching objectives by adjusting the corresponding weights of the similarity score and the step score. At the same time, by objectively and fairly comprehensively evaluating the student assignment grade, the learning situation of students in the assignment can be more accurately understood, and the error types in the student assignment can be analyzed, thereby providing strong support for teaching. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0063] Figure 1 It is a schematic diagram of the system framework of the present invention;

[0064] Figure 2 It is a flowchart of the system method of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0065] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] See also Figure 1 The first aspect of the present invention provides a student homework automatic marking and grading system based on image recognition, including the following modules:

[0067] Image acquisition module: real-time acquisition of image data of students’ homework;

[0068] Image processing module: pre-processing based on the collected student homework image data, including: image denoising, image enhancement, conversion into grayscale and binarization, and adjustment of image size and direction;

[0069] Image recognition module: recognize characters in image data based on processed student homework image data;

[0070] Image analysis module: compare and analyze the handwritten characters in the recognized image data with the standard answers and automatically correct and score the students' homework; analyze the types of errors in the students' homework; calculate the error rate in the students' homework according to the error types;

[0071] Assignment Grading Module: Comprehensively evaluate the student assignment grade based on the total score and error rate of the student assignment.

[0072] Specifically, the present invention includes: an image acquisition module, an image processing module, an image recognition module, an image analysis module and a homework rating module; pre-processing the collected student homework images includes: denoising, enhancing the contrast, brightness, smoothness and edge information of the student homework images; converting the color image into a grayscale image; binarization processing is to convert the grayscale image into a black and white image; adjusting the size and direction of the image, adjusting the image to a unified standard size and adjusting the direction of the unified image. Automatic correction and grading are achieved by comparing the characters in the image data with the standard answers to calculate the similarity and step scores; calculating the error rate of the student homework according to the analysis of the error type; and at the same time, using the total score and error rate to comprehensively evaluate the student homework grade.

[0073] In one embodiment of the present invention, recognizing characters in image data based on processed student homework image data includes the following steps:

[0074] According to the preprocessed student homework image data, the characters in the student homework image data are divided into handwritten characters and printed characters, and the characters in the student homework image data are recognized and analyzed through a deep learning model;

[0075] Annotate handwritten and printed characters in student work image data;

[0076] Input the annotated handwritten characters and printed characters into the deep learning model for model training;

[0077] The student homework automatic grading system collects and processes student homework image data in real time and inputs it into the trained deep learning model to recognize handwritten and printed characters;

[0078] The recognized handwritten characters in the student's homework image data are output.

[0079] Specifically, the convolutional neural network of the deep learning model is used to input the labeled handwritten characters and printed characters into the deep learning model for training; the student homework image data that is collected and preprocessed in real time is input into the trained deep learning model, and the model can recognize handwritten characters and printed characters, and output the handwritten characters in the recognized student homework image data.

[0080] See also Figure 2 In one embodiment of the present invention, the handwritten characters in the recognized image data are compared and analyzed with the standard answers, and the student's homework is automatically corrected and graded, including the following steps:

[0081] S1: Convert the handwritten characters in the standard answers and student homework image data into a unified format, including: removing irrelevant symbols, unifying uppercase and lowercase letters, and character encoding;

[0082] S2: Use the Levenshtein distance algorithm to compare the student answer and the standard answer, where the student answer string set is represented by X n = {X1, X2, ..., X n}, the standard answer string set is represented by Y m ={Y1, Y2, ..., Y m};

[0083] S3: Calculate the similarity μ between the student's answer and the standard answer and score the student's homework;

[0084] S4: Calculate the step score of the student's answer;

[0085] S5: The student's homework score is calculated based on the similarity μ score and the step score.

[0086] Specifically, converting the handwritten characters in the standard answers and student homework image data into a unified format includes: removing non-critical punctuation and special symbols, unifying the capitalization of the text, and encoding the characters according to a unified coding system. The Levenshtein distance algorithm calculates the minimum number of steps required to transform one string into another based on the insertion, deletion, and replacement operations of characters to quantify the degree of difference between the two strings. When comparing the student answers with the standard answers, the algorithm can effectively evaluate the correctness of the student answers. Based on the comparison results, the similarity μ and the step score of the student answer are calculated to comprehensively calculate the total score of the student homework.

[0087] As a further solution of the present invention: the step S2 comprises the following steps:

[0088] Calculated by analytical formula:

[0089]

[0090] Get the Levenshtein distance lev between the student answer string and the standard answer string X,Y (|X|,|Y|), where |X| and |Y| represent the length of the student answer string and the standard answer string respectively; lev X,Y (n,m) is the distance between the first n characters of X and the first m characters of Y.

[0091] Specifically, the Levenshtein distance algorithm refers to the minimum number of editing operations required to transform two strings from one to another. Allowed editing operations include replacing one character with another, inserting a character, and deleting a character. Using the Levenshtein distance algorithm, compare the student's answer with the standard answer.

[0092] In one embodiment of the present invention, step S3 comprises the following steps:

[0093] Calculated by analytical formula:

[0094] Get the similarity μ between the student's answer string and the standard answer; according to the similarity μ, set different levels of thresholds and divide the scores into:

[0095] When μ ≥ 90%, the student's assignment gets full marks:

[0096] When 60% ≤ μ < 90%, the student's work is partially scored;

[0097] When μ<60%, the student's assignment will not be scored.

[0098] Specifically, the Levenshtein distance algorithm is used to obtain the comparison result to calculate the similarity μ between the student's answer string and the standard answer; when the Levenshtein distance levX,Y The smaller the value of (|X|,|Y|), the higher the similarity between the student's answer string and the standard answer; according to the similarity μ, different levels of thresholds are set, and the scoring is divided into: when μ≥90%, the student's homework gets full marks; when 60%≤μ<90%, the student's homework is considered to have a wrong answer but correct steps, so it can be partially scored, namely the step score; when μ<60%, the student's homework is not scored.

[0099] In one embodiment of the present invention, the step S4 comprises the following steps:

[0100] Calculated by analytical formula:

[0101] The step score is F, where n is the total number of steps and f i represents the score of the ith step, w i represents the weight of the i-th step;

[0102] When μ ≥ 90% of the steps, the step score is F;

[0103] When 60% ≤ μ < 90% of the steps, the step score is

[0104] When μ < 60% of the steps, the step score is 0.

[0105] Specifically, the step score F is calculated by multiplying the product of the score of the i-th step and the weight of the corresponding step by the percentage; finally, the percentage is multiplied to ensure that the step score of the student's homework is in percentage form. The weight of the i-th step is set and allocated by the teacher according to the difficulty of the homework. Different levels of thresholds are set according to the similarity μ, and the step score in the student's homework is divided into three parts: when μ ≥ 90%, the step score in the student's homework is the value of F; when 60% ≤ μ < 90%, the step score in the student's homework is When μ<60%, the step score in the student's assignment is 0.

[0106] In one embodiment of the present invention, the step S5 comprises the following steps:

[0107] Calculated by analytical formula:

[0108] Get the total score of the student's homework Total_score; where W μ is the weight of similarity μ, W F is the weight of the step score F.

[0109] Specifically, the similarity μ and step score F are multiplied by their corresponding weights, divided by the square root of the sum of the squares of the weights, and then multiplied by the percentage. The square root of the sum of the squares of the weights in the denominator will adjust the size of the sum accordingly to accurately calculate the total score of the student's assignment Total_score; finally, multiply by the percentage to ensure that the total score of the student's assignment is in percentage form. The total score of the student's assignment is calculated by the similarity μ and step score F between the student's answer string and the standard answer, ensuring the fairness and accuracy of the scoring.

[0110] In one embodiment of the present invention, analyzing the error types in student assignments includes the following steps:

[0111] Based on the graded student assignments, the deep learning model is used to identify and analyze the types of errors in the student assignments.

[0112] Mark the a-th error in the student's homework, where the a-th error includes the first knowledge error, the second method error, the third logic error, the fourth psychological error and the fifth negligence error;

[0113] Input the labeled a-th error into the deep learning model for model training, where a∈{1, 2, …, n};

[0114] The student homework image data is graded by the student homework automatic grading system and input into the trained deep learning model to identify the ath type of error in the student homework.

[0115] Specifically, based on the corrected and graded student homework, the convolutional neural network of the deep learning model is used to input the annotated a-th error into the deep learning model for model training, where a∈{1, 2, ..., n}; the image data of the corrected and graded student homework is input into the trained deep learning model, and the model can identify the a-th error in the student homework, and output the a-th error in the identified student homework image data, helping students to identify and correct the a-th error, thereby improving learning efficiency. The a-th error includes the first knowledge error: the lack of in-depth understanding of the knowledge point and the weak grasp of the knowledge point; the second methodological error: the misuse of problem-solving strategies or techniques; the third logical error: the jump of thinking or logical confusion in the reasoning process; the fourth psychological error: carelessness, excessive anxiety or overconfidence leading to misreading of the question; the fifth negligent error: writing errors caused by negligence.

[0116] In one embodiment of the present invention, calculating the error rate in a student's homework according to the error type includes the following steps:

[0117] Calculated by analytical formula:

[0118] Obtain the error rate α in the student's homework; where, w a represents the weight of the a-th type of error, and a j represents the number of errors, n is the total number of error types, and Y m represents the standard answer string.

[0119] Specifically, according to the occurrence frequency of errors, combined with the weight of the error and then multiplied by the percentage, calculate the comprehensive error rate α to ensure that the value of the error rate is between 0% and 100%; the occurrence frequency of errors is obtained from the ratio of the number of errors to the sum of the standard answer strings; the weight w of the a-th type of error is assigned according to the severity of the error a which is set and assigned by the teacher according to the difficulty of the homework.

[0120] In one embodiment of the present invention, the student's homework level is comprehensively evaluated according to the total score and error rate of the student's homework, including the following steps:

[0121] Calculate through the analysis formula:

[0122] Q = (W * Total_score+(1 - W) * α)×100%

[0123] Obtain the comprehensive score of the student's homework Q, where the weight W of the total score Total_score is 0.8, and the weight (1 - W) of the error rate α is 0.2;

[0124] According to the comprehensive score of the student's homework Q, set different level thresholds to divide the student's homework into five levels: A level, B level, C level, D level and E level:

[0125] When Q ≥ 80%, the homework level is A level;

[0126] When 60% < Q < 80%, the student's homework level is B level;

[0127] When 40% ≤ Q ≤ 60%, the student's homework level is C level;

[0128] When 20% < Q < 40%, the student's homework level is D level;

[0129] When Q ≤ 20%, the student's homework level is E level.

[0130] Specifically, the student assignment Q is comprehensively scored according to the sum of the total score of the student assignment multiplied by its weight and the error rate multiplied by its weight and then multiplied by a percentage, and different levels of thresholds are set to divide the student assignments into five levels: A, B, C, D and E; among them, the student assignments of level A are excellent, the student assignments of level B are good, the student assignments of level C are average, the student assignments of level D are passing, and the student assignments of level E are unpassing; the value of the comprehensive score of the student assignment Q calculated in real time is compared with the threshold to obtain the student assignment grade.

[0131] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. An automatic grading system for student assignments based on image recognition, characterized in that: Includes the following modules: Image acquisition module: real-time acquisition of image data of students’ homework; Image processing module: pre-processing based on the collected student homework image data, including: image denoising, image enhancement, conversion into grayscale and binarization, and adjustment of image size and direction; Image recognition module: recognize characters in image data based on processed student homework image data; Image analysis module: compare and analyze the handwritten characters in the recognized image data with the standard answers and automatically correct and score the students' homework; analyze the types of errors in the students' homework; calculate the error rate in the students' homework according to the error types; Assignment Grading Module: Comprehensively evaluate the student assignment grade based on the total score and error rate of the student assignment.

2. According to the image recognition-based automatic marking and grading system for student assignments of claim 1, it is characterized in that: Recognizing characters in the image data according to the processed student homework image data includes the following steps: According to the preprocessed student homework image data, the characters in the student homework image data are divided into handwritten characters and printed characters, and the characters in the student homework image data are recognized and analyzed through a deep learning model; Annotate handwritten and printed characters in student work image data; Input the annotated handwritten characters and printed characters into the deep learning model for model training; The student homework automatic grading system collects and processes student homework image data in real time and inputs it into the trained deep learning model to recognize handwritten and printed characters; The recognized handwritten characters in the student's homework image data are output.

3. According to the image recognition-based automatic marking and grading system for student assignments of claim 1, it is characterized in that: The method compares and analyzes the handwritten characters in the recognized image data with the standard answers and automatically corrects and grades the student's homework, including the following steps: S1: Convert the handwritten characters in the standard answers and student homework image data into a unified format, including: removing irrelevant symbols, unifying uppercase and lowercase letters, and character encoding; S2: Use the Levenshtein distance algorithm to compare the student answer and the standard answer, where the student answer string set is represented by X n = {X1, X2, ..., X n }, the standard answer string set is represented by Y m ={Y1, Y2, ..., Y m }; S3: Calculate the similarity μ between the student's answer and the standard answer and score the student's homework; S4: Calculate the step score of the student's answer; S5: The student's homework score is calculated based on the similarity μ score and the step score.

4. The automatic marking and grading system for student assignments based on image recognition according to claim 3 is characterized in that: The step S2 comprises the following steps: Calculated by analytical formula: Get the Levenshtein distance lev between the student answer string and the standard answer string X,Y (|X|,|Y|), where |X| and |Y| represent the length of the student answer string and the standard answer string respectively; lev X,Y (n,m) is the distance between the first n characters of X and the first m characters of Y.

5. According to the image recognition-based automatic marking and grading system for student assignments of claim 3, it is characterized in that: The step S3 comprises the following steps: Calculated by analytical formula: Get the similarity μ between the student's answer string and the standard answer; according to the similarity μ, set different levels of thresholds and divide the scores into: When μ ≥ 90%, the student's assignment gets full marks: When 60% ≤ μ < 90%, the student's work is partially scored; When μ<60%, the student's assignment will not be scored.

6. The automatic marking and grading system for student assignments based on image recognition according to claim 3 is characterized in that: The step S4 comprises the following steps: Calculated by analytical formula: The step score is F, where n is the total number of steps and f i represents the score of the ith step, w i represents the weight of the i-th step; When μ ≥ 90% of the steps, the step score is F; When 60% ≤ μ < 90% of the steps, the step score is When μ < 60% of the steps, the step score is 0.

7. The automatic marking and grading system for student assignments based on image recognition according to claim 3 is characterized in that: The step S5 comprises the following steps: Calculated by analytical formula: Get the total score of the student's homework Total_score; where W μ is the weight of similarity μ, W F is the weight of the step score F.

8. The automatic marking and grading system for student assignments based on image recognition according to claim 1 is characterized in that: Analyzing the types of errors in student assignments includes the following steps: Based on the graded student assignments, the deep learning model is used to identify and analyze the types of errors in the student assignments. Mark the a-th error in the student's homework, where the a-th error includes the first knowledge error, the second method error, the third logic error, the fourth psychological error and the fifth negligence error; Input the labeled a-th error into the deep learning model for model training, where a∈{1, 2, …, n}; Grade the image data of students' homework after being automatically graded by the students' homework automatic grading system, and input it into the trained deep learning model to identify the a-th type of error in the students' homework.

9. The automatic marking and grading system for student assignments based on image recognition according to claim 1 is characterized in that: Calculate the error rate in the students' homework according to the error type, including the following steps: Calculated by analytical formula: Get the error rate α in the student's homework; where w a represents the weight of the a-th error, a j represents the number of errors, n is the total number of error types, Y m Indicates the standard answer string.

10. The automatic marking and grading system for student assignments based on image recognition according to claim 1, characterized in that: Comprehensively evaluate the level of the students' homework according to the total score of the students' homework and the error rate, including the following steps: Calculate through the analysis formula: Q = (W * Total_score + (1 - W) * α) × 100% Obtain the comprehensively scored students' homework Q, where the weight W of the total score Total_score is 0.8, and the weight (1 - W) of the error rate α is 0.2; According to the comprehensively scored students' homework Q, set different threshold levels to divide the students' homework into five levels: A level, B level, C level, D level, and E level: When Q ≥ 80%, the homework level is A level; When 60% < Q < 80%, the students' homework level is B level; When 40% ≤ Q ≤ 60%, the students' homework level is C level; When 20% < Q < 40%, the students' homework level is D level; When Q ≤ 20%, the students' homework level is E level.

Images