An intelligent teaching management method and system based on big data mining

By using big data mining technology to analyze the compositional elements and colors of paintings, the smoothness of lines, the balance of composition, and the color matching of paintings can be quantitatively evaluated. This solves the problems of objectivity and low efficiency in traditional painting teaching and enables personalized teaching management and quality improvement.

CN118710462BActive Publication Date: 2024-11-15DERUIHUA EDUCATION TECH (BEIJING) CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411195606.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-29
Publication Date
2024-11-15
Estimated Expiration
2044-08-29

AI Technical Summary

Technical Problem

Traditional painting teaching methods lack objectivity and are inefficient, making it difficult to provide personalized feedback and guidance, which leads to a decline in the quality and efficiency of painting teaching.

Method used

By using big data mining technology and element recognition and style recognition models, we can extract compositional elements and analyze colors in paintings, quantitatively evaluate line smoothness, compositional balance and compositional proportion, and combine color space conversion and clustering algorithms to analyze color matching, thereby establishing a style recognition model to achieve style recognition and evaluation of artworks.

Benefits of technology

It achieves objectivity and efficiency in painting instruction, provides personalized improvement suggestions and suitable learning courses, and enhances teaching quality and efficiency.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118710462B_ABST
    Figure CN118710462B_ABST
Patent Text Reader

Abstract

The present invention provides an intelligent teaching management method and system based on big data mining, which pre-processes student paintings to obtain standardized work images, extracts composition elements from the standardized work images through an element recognition model to obtain work composition element data, performs composition analysis on the composition element data to obtain composition element distribution feature data, performs color analysis on the standardized work images to obtain work color matching feature data, performs style matching through a style recognition model based on the work composition element distribution feature data and the work color matching feature data to obtain the work style, compares and analyzes the work composition element distribution feature data and the work color matching feature data according to preset style standard indicators based on the work style to obtain a work evaluation result, provides specific improvement suggestions for students based on the work evaluation result and pushes adapted learning courses, thereby realizing objective and efficient teaching management.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of educational technology, and in particular to an intelligent teaching management method and system based on big data mining. Background Art

[0002] Painting instruction, as an essential component of modern education, aims to cultivate students' drawing skills, aesthetic appreciation, and artistic expression. Traditional painting instruction relies on the teacher's subjective judgment and experience. Teachers evaluate and guide students' paintings based on their personal experience and aesthetic standards, resulting in a lack of objectivity and consistency in evaluation. Furthermore, as the number of students increases, it becomes difficult for teachers to provide real-time, personalized feedback and suggestions when faced with a large number of student works. This reduces the efficiency of managing and guiding student works, further impacting student growth. Therefore, providing students with more objective, accurate, and personalized painting instruction and promoting improvements in both quality and efficiency has become a pressing issue. Summary of the Invention

[0003] (1) Technical problems to be solved

[0004] The purpose of the present invention is to provide an intelligent teaching management method and system based on big data mining to solve the problems of lack of objectivity and low efficiency in painting teaching and to achieve more objective and efficient teaching management.

[0005] (2) Technical solution

[0006] To achieve the above objectives, the present invention provides, on the one hand, an intelligent teaching management method based on big data mining, the method comprising:

[0007] The student's paintings are obtained and preprocessed to obtain standardized images of the paintings.

[0008] The standardized work image is subjected to composition element extraction through an element recognition model to obtain work composition element data, and the composition element data is subjected to composition analysis to obtain composition element distribution feature data.

[0009] The standardized work images are subjected to color analysis to obtain the color matching feature data of the works. The style of the works is obtained by performing style matching through a style recognition model based on the distribution feature data of the composition elements and the color matching feature data of the works. The style recognition model is a mapping relationship between the distribution feature data of the composition elements, the color matching feature data of the works and the style of the works established based on the collected historical big data of the paintings.

[0010] According to the style of the work, the distribution feature data of the composition elements and the color matching feature data of the work are compared and analyzed according to the preset style standard indicators to obtain the work evaluation results. Based on the work evaluation results, specific improvement suggestions are provided to students and appropriate learning courses are pushed.

[0011] Furthermore, the method of extracting composition elements from the standardized work image through an element recognition model to obtain work composition element data, and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes:

[0012] The texture feature map is obtained by performing texture detection on the standardized work image through the gray level co-occurrence matrix.

[0013] The texture feature map is used to identify straight lines and curves in the image through the Hough transform algorithm to obtain line feature data.

[0014] The texture feature map is segmented by the watershed algorithm to obtain a texture region group, and the texture region group is matched with a preset pattern to obtain pattern feature data.

[0015] The line feature data and the pattern feature data are used as composition element data, and the composition element data are subjected to multi-level composition analysis to obtain composition element distribution feature data.

[0016] Furthermore, the method of using the line feature data and the pattern feature data as composition element data and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes:

[0017] Line smoothness data is obtained by performing line smoothness evaluation based on line feature data in the composition element data through Fourier transform.

[0018] The composition balance data is obtained by position distribution analysis based on the pattern feature data in the composition element data.

[0019] Composition ratio data is obtained by performing ratio analysis based on pattern feature data in the composition element data.

[0020] The line smoothness data, composition balance data, and composition ratio data are summarized to obtain the composition element distribution feature data.

[0021] Furthermore, the method of obtaining line smoothness data by performing line smoothness evaluation through Fourier transform based on line feature data in the composition element data includes:

[0022] Obtain the length of the line, and obtain the number of data points according to a preset data point acquisition quantity rule based on the length of the line. The data point acquisition quantity rule is a mapping relationship between the line length and the number of data points established based on the ratio of the line length to the unit length.

[0023] The coordinate system is established with the intersection of the left and lower boundaries of the image as the origin. The coordinates of the data points on the line are obtained at uniform intervals according to the number of data points obtained. The data points are converted from the spatial domain to the frequency domain using Fourier transform to obtain the frequency components. for:

[0024] .

[0025] in is the data point index, It is The complex number formed by the coordinates of the data points, is the total number of data points, is an imaginary unit, is the index of the frequency component, ranging from arrive , It is frequency components.

[0026] Obtain the amplitude spectrum of the frequency component according to the frequency component for:

[0027] .

[0028] in It is The real part of the frequency component, It is The imaginary part of the frequency component, It is The magnitude spectrum of the frequency components.

[0029] According to the preset frequency component cutoff ratio, the frequency component before the cutoff of the frequency component index is selected as the low-frequency component, and the frequency component after the cutoff of the frequency component index is selected as the high-frequency component. The line smoothness is obtained according to the low-frequency component and the high-frequency component:

[0030] .

[0031] in is the maximum frequency index of the low-frequency component, is the minimum frequency index of the high-frequency component, It is Line smoothness of the line.

[0032] Obtain the line smoothness data of the entire image based on the line smoothness of each line for:

[0033] .

[0034] in is the total number of lines, is the line smoothness data.

[0035] Furthermore, the method of obtaining composition balance data by position distribution analysis based on pattern feature data in composition element data includes:

[0036] Get the pixel coordinates of each pattern in the pattern feature data, and calculate the geometric center of the pattern based on the pixel coordinates for:

[0037] ;

[0038] .

[0039] in It's a pattern Pixel area, Is a point in the region The horizontal axis, Is a point in the region The vertical coordinate, is the number of pixels in the pattern, It's a pattern The horizontal coordinate of the geometric center, It's a pattern The vertical coordinate of the geometric center.

[0040] According to the geometric center of each pattern, the geometric center distance of the pattern is:

[0041] .

[0042] in It's a pattern The horizontal coordinate of the geometric center, It's a pattern The vertical coordinate of the geometric center, It's a pattern The horizontal coordinate of the geometric center, It's a pattern The vertical coordinate of the geometric center, It's a pattern Geometric center and pattern Geometric center distance.

[0043] The shortest distance between each pattern and its geometric center is selected to obtain each adjacent pattern distance data set, and the pattern distance average and pattern distance standard deviation are calculated based on the average and standard deviation of each adjacent pattern distance data set.

[0044] The pattern distance standard deviation is compared with the pattern distance average value. When the difference between the pattern distance standard deviation and the pattern distance average value is greater than the preset difference threshold, the distribution result is the first characterization result. When the difference between the pattern distance standard deviation and the pattern distance average value is less than or equal to the preset difference threshold, the distribution result is the second characterization result.

[0045] The image of the work is divided into thirds to obtain the coordinates of the intersection of the thirds lines. When the distribution result is the first characterization result, the pattern with the largest distance between adjacent patterns is selected as the visually prominent pattern, and the degree of conformity between the visually prominent pattern and the intersection of the thirds lines is calculated as the composition balance data. When the distribution result is the second characterization result, the mean degree of conformity between the geometric center of each pattern and the intersection of the thirds lines is calculated as the composition balance data.

[0046] Furthermore, the method of dividing the work image by the rule of thirds to obtain the coordinates of the intersection points of the third lines, selecting the pattern with the largest distance between adjacent patterns as the visually prominent pattern when the distribution result is the first characterization result, calculating the degree of conformity between the visually prominent pattern and the intersection points of the third lines as the composition balance data, and calculating the mean degree of conformity between the geometric center of each pattern and the intersection points of the third lines as the composition balance data when the distribution result is the second characterization result includes:

[0047] The work image is divided into nine equal areas using the rule of thirds to obtain four intersection points of the third lines, and the coordinates of the intersection points of the third lines are obtained based on the length and width of the work image.

[0048] When the distribution result is the first characterization result, the pattern with the largest distance between adjacent patterns is selected as the visually prominent pattern. The distance between the geometric center of the visually prominent pattern and each three-line intersection is calculated, and the intersection with the smallest distance value is selected as the target intersection. The conformity between the visually prominent pattern and the three-line intersection is obtained by comparing the distance between the geometric center of the pattern and the target intersection with the preset maximum distance. for:

[0049] .

[0050] in is the distance between the geometric center of the visually salient pattern and the target intersection point, is the preset maximum distance, It is the degree of conformity between the visually prominent pattern and the intersection of the third lines.

[0051] When the distribution result is the second characterization result, the distance between each pattern geometric center and each three-line intersection is calculated and the intersection with the smallest distance value is selected as its corresponding target intersection point. The distance between each pattern geometric center and its corresponding target intersection point is compared with the preset maximum distance to obtain the conformity of each pattern with the three-line intersection point. The mean of the conformity between each pattern geometric center and the three-line intersection point is calculated based on the mean of the conformity between each pattern geometric center and the three-line intersection point. for:

[0052] .

[0053] in is the number of patterns, It is The distance between the intersection of a pattern and its corresponding target, It is the mean of the conformity between the geometric center of each pattern and the intersection of the three-point line.

[0054] Furthermore, the method of obtaining composition ratio data by performing ratio analysis based on pattern feature data in composition element data includes:

[0055] The pattern sizes are sorted according to the number of pixels in the pattern feature data, and the first preset patterns are obtained as the main patterns according to the preset number of main patterns. The width of the main pattern is obtained by subtracting the horizontal coordinates of the leftmost and rightmost pixels of the main pattern, and the height of the main pattern is obtained by subtracting the vertical coordinates of the topmost and bottommost pixels of the main pattern.

[0056] Get the pattern width ratio based on the main pattern width and main pattern height and pattern height ratio for:

[0057] ;

[0058] .

[0059] in It is The width of the main pattern, It is The width of the main pattern, It is The height of the main pattern, It is The height of the main pattern, It's a pattern With pattern The width ratio, It's a pattern With pattern height ratio.

[0060] Furthermore, the method of performing color analysis on the standardized work image to obtain the work color matching feature data, and performing style matching using a style recognition model based on the work composition element distribution feature data and the work color matching feature data to obtain the work style includes:

[0061] The image RGB values ​​are extracted through the pixel traversal algorithm, and the image RGB values ​​are converted into image HSV values ​​through the color space conversion algorithm. The image HSV values ​​are clustered through the K-means clustering algorithm to obtain the center value of the main color cluster of the image and the proportion of the main color cluster as the color matching feature data of the work.

[0062] Inputting the composition ratio data and the color matching feature data of the work composition element distribution feature data into the style recognition model for style matching to obtain the style of the work, wherein the style of the work includes realistic style, cartoon style, anime style, impressionist style, abstract style, and Chinese style; the steps of establishing the style recognition model include:

[0063] Collect images of paintings and extract composition ratio data and color matching feature data from the composition element distribution feature data. Label each work with style labels to obtain original data. Use 70% of the original data as training data to train the style recognition model through the neural network algorithm.

[0064] The input layer receives the distribution feature data of composition elements and color matching feature data, the middle layer uses the ReLU activation function, and the output layer uses the softmax activation function to set 6 neurons to output the probability of the style of the work.

[0065] The cross entropy loss function is used to train the style recognition model using the training dataset according to the forward propagation and backpropagation algorithms through the Adam optimizer.

[0066] 30% of the original data is used as validation data to verify the style recognition model and the model is adjusted based on the validation results.

[0067] Furthermore, the method of comparing and analyzing the composition element distribution feature data and the color matching feature data of the work according to the preset style standard indicators to obtain a work evaluation result based on the work style, and providing students with specific improvement suggestions and pushing adapted learning courses based on the work evaluation result includes:

[0068] Style standard indicators are set based on the composition element distribution characteristic data and color matching characteristic data of historical excellent works in each style, including the line smoothness, composition balance, composition ratio, image main color cluster range value and color cluster proportion of each style of works.

[0069] According to the style of the work, the distribution characteristic data of the composition elements and the color matching characteristic data of the work are compared with the style standard indicators to obtain the work evaluation results. Based on the deviation items in the work evaluation results, specific improvement suggestions are provided to students and adapted learning courses are pushed.

[0070] Based on the same inventive concept, the present invention provides an intelligent teaching management system based on big data mining, the system including: an image preprocessing module, a composition analysis module, a color and style analysis module, and a feedback module, and the modules are connected in sequence.

[0071] The image preprocessing module is used to obtain students' paintings and preprocess the paintings to obtain standardized work images.

[0072] The composition analysis module is used to extract composition elements from the standardized work image through the element recognition model to obtain work composition element data, and perform composition analysis on the composition element data to obtain composition element distribution feature data.

[0073] The color and style analysis module is used to perform color analysis on the standardized work image to obtain the work color matching feature data, and to obtain the work style by performing style matching based on the work composition element distribution feature data and the work color matching feature data through a style recognition model. The style recognition model is a mapping relationship between the composition element distribution feature data, the work color matching feature data and the work style established based on the collected historical big data of the painting works.

[0074] The feedback module is used to compare and analyze the composition element distribution feature data and the color matching feature data of the work according to the preset style standard indicators according to the style of the work to obtain the work evaluation results, provide students with specific improvement suggestions and push adapted learning courses based on the work evaluation results.

[0075] (3) Beneficial effects

[0076] Compared with the existing technology, the present invention has the following beneficial effects: through the element recognition model combined with multi-level composition analysis, quantitative evaluation of the line smoothness, composition balance and composition proportion of the painting work is achieved; through color space conversion and clustering algorithm, the color is analyzed to extract the main color clusters of the work to achieve color matching feature evaluation; through the style recognition model trained based on big data, the style of the work is recognized and the distribution feature data of the composition elements and the color matching feature data of the work are compared with the style standard indicators to obtain the evaluation results, thereby achieving more objective and efficient painting teaching management as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0077] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0078] Figure 1 This is a flowchart of an intelligent teaching management method based on big data mining according to Example 1 of the present invention;

[0079] Figure 2 This is a schematic diagram of the module composition of an intelligent teaching management system based on big data mining according to Example 2 of the present invention. DETAILED DESCRIPTION

[0080] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention. Example 1

[0081] like Figure 1 As shown, this embodiment provides an intelligent teaching management method based on big data mining, the method comprising:

[0082] S10. Obtain the student's painting and pre-process the painting to obtain a standardized work image; the pre-processing includes image cropping, scaling, denoising and standardization, which are all conventional image pre-processing methods in the industry.

[0083] S20. Extract composition elements from the standardized work image through an element recognition model to obtain composition element data of the work, and perform composition analysis on the composition element data to obtain composition element distribution feature data; the composition element data includes line and pattern feature data; the composition element distribution feature data includes line smoothness data, composition balance data, and composition ratio data.

[0084] S30. Perform color analysis on the standardized work image to obtain color matching feature data of the work, and perform style matching using a style recognition model based on the distribution feature data of the composition elements and the color matching feature data of the work to obtain the style of the work. The style recognition model is a mapping relationship between the distribution feature data of the composition elements, the color matching feature data of the work and the style of the work established based on the collected historical big data of the painting works.

[0085] S40. Compare and analyze the composition element distribution feature data and the color matching feature data of the work according to the preset style standard indicators according to the style of the work to obtain the work evaluation results, provide students with specific improvement suggestions and push appropriate learning courses based on the work evaluation results.

[0086] It should be noted that the method of extracting composition elements from a standardized work image through an element recognition model to obtain work composition element data, and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes:

[0087] S21. Perform texture detection on the standardized work image using a gray-level co-occurrence matrix to obtain a texture feature map.

[0088] For example, the standardized work image is a landscape painting, and the texture of the trunks and branches of trees, the rock texture of the mountains, and the ripple features of the river are extracted through the gray-level co-occurrence matrix GLCM to obtain a texture feature map.

[0089] S22. Using the Hough transform algorithm to identify straight lines and curves in the texture feature map to obtain line feature data.

[0090] For example, the texture feature map is used through the Hough transform algorithm to identify the lines in the image, such as mountain boundary lines, tree texture lines, and river line feature data.

[0091] S23 , segmenting the detected texture features in the texture feature map using a watershed algorithm to obtain a texture region group, and matching the texture region group with a preset pattern to obtain pattern feature data.

[0092] For example: the texture feature map is segmented into independent texture areas such as tree areas, flower areas, mountain areas, river areas, rock areas, and human areas through the watershed algorithm, and then these texture areas are matched with preset pattern templates such as commonly used tree, flower, ripple and other template patterns to obtain pattern feature data.

[0093] S24: Using the line feature data and the pattern feature data as composition element data, and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data.

[0094] It should be noted that the method of using line feature data and pattern feature data as composition element data and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes:

[0095] S241. Line smoothness data is obtained by performing a line smoothness evaluation through Fourier transform based on the line feature data in the composition element data. Smooth lines are more in line with the aesthetics of nature and art, and also reflect the student's basic painting skills. The lines of the works of students with good painting skills are smooth and fluent, while the lines of the works of students with poor painting skills are significantly worse than those of students with good painting skills due to unstable hands. Therefore, the evaluation of line smoothness can reflect the level of the painting work to a certain extent.

[0096] S242. Composition balance data is obtained through position distribution analysis based on the pattern feature data in the composition element data. Composition balance reflects the reasonable layout of each pattern in space, making the entire picture appear neither too crowded nor too empty. By evaluating composition balance, we can judge the student's mastery of composition principles and reflect the visual coordination of their paintings.

[0097] S243. Perform proportion analysis based on the pattern feature data in the composition element data to obtain composition proportion data. The proportional relationship between the elements in the work directly affects the style of the picture. Different proportions can reflect different painting styles. The style of the student's work can be identified through proportion analysis. For example, the proportions of people and animals in the realistic style are close to those in the real world, while the cartoon style has shorter and rounder patterns, so the sizes of people and animals in the cartoon style are closer to those in the realistic style, and the proportions are closer to 1.

[0098] S244 , summarizing the line smoothness data, the composition balance data, and the composition ratio data to obtain composition element distribution feature data.

[0099] It should be noted that the method of obtaining line smoothness data by performing line smoothness evaluation through Fourier transform based on line feature data in the composition element data includes:

[0100] S2411. Obtain the length of the line, and obtain the number of data points according to the preset data point acquisition number rule based on the length of the line. The data point acquisition number rule is a mapping relationship between the line length and the number of data points established based on the ratio of the line length to the unit length; for example, a unit length of 1 mm corresponds to 1 data point. If the line length is 20 mm, the number of data points is 20.

[0101] S2412. Establish a coordinate system with the intersection of the left and lower boundaries of the image as the origin. Obtain the coordinates of the data points on the line at uniform intervals according to the number of data points obtained. Use Fourier transform to convert the data points from the spatial domain to the frequency domain to obtain the frequency components. for:

[0102] .

[0103] in is the data point index, It is The complex number formed by the coordinates of the data points, is the total number of data points, is an imaginary unit, is the index of the frequency component, ranging from arrive , It is frequency components.

[0104] For example: Assume the curve has 5 data points ,but , , , , ,when hour, = + =10+13 , similarly we can get , , , .

[0105] S2413, obtain the amplitude spectrum of the frequency component according to the frequency component for:

[0106] .

[0107] in It is The real part of the frequency component, It is The imaginary part of the frequency component, It is The magnitude spectrum of the frequency components.

[0108] For example: When , , , , Time available , , similarly we can get , , .

[0109] S2414: Select the frequency component before the cutoff of the frequency component index as the low-frequency component according to the preset frequency component cutoff ratio, and select the frequency component after the cutoff of the frequency component index as the high-frequency component. The line smoothness is obtained based on the low-frequency component and the high-frequency component:

[0110] .

[0111] in is the maximum frequency index of the low-frequency component, is the minimum frequency index of the high-frequency component, It is Line smoothness of the line.

[0112] The low-frequency component reflects the main structure and smooth changes of the line, and the high-frequency component reflects the rapid changes of the line, such as sharp turns and jitters of the line; for example, if there are a total of 5 frequency components, when the preset frequency component cutoff ratio is set to 60%, the first 3 frequency components of the frequency component index are selected as low-frequency components, and the last 2 frequency components of the frequency component index are selected as high-frequency components, then the line smoothness of the aforementioned line is .

[0113] S2415: Obtain line smoothness data of the entire image based on the line smoothness of each line. for:

[0114] .

[0115] in is the total number of lines, is the line smoothness data.

[0116] For example: If the image has a total of 5 lines, their smoothness is , , , , , then the line smoothness data of the entire image . The larger the value, the smoother the line. When it is greater than 1, it indicates that the low-frequency component is dominant and the smooth change of the line is significantly greater than the sharp change. Less than 1 indicates that high-frequency components dominate and the lines have many sharp turning points and jitters. The evaluation of line smoothness can reflect the students' drawing level.

[0117] It should be noted that the method of obtaining composition balance data by position distribution analysis based on pattern feature data in composition element data includes:

[0118] S2421. Obtain the pixel coordinates of each pattern in the pattern feature data, and calculate the geometric center of the pattern based on the pixel coordinates. for:

[0119] ;

[0120] .

[0121] in It's a pattern Pixel area, Is a point in the region The horizontal axis, Is a point in the region The vertical coordinate, is the number of pixels in the pattern, It's a pattern The horizontal coordinate of the geometric center, It's a pattern The vertical coordinate of the geometric center.

[0122] S2422. The geometric center distance of each pattern is obtained based on the geometric center of each pattern:

[0123] .

[0124] in It's a pattern The horizontal coordinate of the geometric center, It's a pattern The vertical coordinate of the geometric center, It's a pattern The horizontal coordinate of the geometric center, It's a pattern The vertical coordinate of the geometric center, It's a pattern Geometric center and pattern Geometric center distance.

[0125] S2423. Select the shortest distance between each pattern and its geometric center to obtain a data set of adjacent pattern distances for each pattern, and calculate a pattern distance average and a pattern distance standard deviation based on the average and standard deviation of the data set of adjacent pattern distances for each pattern.

[0126] S2424. Compare the pattern distance standard deviation with the pattern distance average value. When the difference between the pattern distance standard deviation and the pattern distance average value is greater than a preset difference threshold, the distribution result is a first characterization result. When the difference between the pattern distance standard deviation and the pattern distance average value is less than or equal to the preset difference threshold, the distribution result is a second characterization result. The first characterization result indicates that the pattern is visually prominent, and the second characterization result indicates that the pattern is evenly distributed.

[0127] S2425. Divide the image of the work by the rule of thirds to obtain the coordinates of the intersection of the thirds lines. When the distribution result is the first characterization result, select the pattern with the largest distance between adjacent patterns as the visually prominent pattern, calculate the degree of conformity between the visually prominent pattern and the intersection of the thirds lines as the composition balance data, and when the distribution result is the second characterization result, calculate the mean degree of conformity between the geometric center of each pattern and the intersection of the thirds lines as the composition balance data.

[0128] It should be noted that the method of dividing the image of the work by the rule of thirds to obtain the coordinates of the intersection points of the third lines, selecting the pattern with the largest distance between adjacent patterns as the visually prominent pattern when the distribution result is the first characterization result, calculating the degree of conformity between the visually prominent pattern and the intersection points of the third lines as the composition balance data, and calculating the mean degree of conformity between the geometric center of each pattern and the intersection points of the third lines as the composition balance data when the distribution result is the second characterization result includes:

[0129] S24251. Divide the image of the work into nine equal areas using the rule of thirds to obtain four intersection points of three-point lines. Obtain the coordinates of the intersection points of three-point lines based on the length and width of the work image. For example, if the length of the work image is L and the width is W, then the coordinates of the four intersection points of three-point lines are .

[0130] S24252. When the distribution result is the first characterization result, the pattern with the largest distance between adjacent patterns is selected as the visually prominent pattern. The distance between the geometric center of the visually prominent pattern and each three-line intersection is calculated, and the intersection with the smallest distance is selected as the target intersection. The degree of conformity between the visually prominent pattern and the three-line intersection is obtained by comparing the distance between the geometric center of the pattern and the target intersection with the preset maximum distance. for:

[0131] .

[0132] in is the distance between the geometric center of the visually salient pattern and the target intersection point, is the preset maximum distance, It is the conformity between the visually prominent pattern and the intersection of the thirds line; the preset maximum distance, for example, in this embodiment, takes half the length of the canvas diagonal as the preset maximum distance. The greater the distance between the geometric center of the pattern and the target intersection, the smaller the conformity value, indicating that the composition balance is worse.

[0133] S24253. When the distribution result is the second characterization result, the distance between each pattern geometric center and each three-line intersection is calculated and the intersection with the smallest distance value is selected as its corresponding target intersection point. The distance between each pattern geometric center and its corresponding target intersection point is compared with the preset maximum distance to obtain the conformity between each pattern and the three-line intersection point. The mean of the conformity between each pattern geometric center and the three-line intersection point is calculated to obtain the mean conformity between each pattern geometric center and the three-line intersection point. for:

[0134] .

[0135] in is the number of patterns, It is The distance between the intersection of a pattern and its corresponding target, It is the mean of the conformity between the geometric center of each pattern and the intersection of the three-point line.

[0136] For example: The length of the work image is mm, width is mm, then the coordinates of the four three-point intersection points are If there are three patterns and the difference between the pattern distance standard deviation and the pattern distance average is less than the preset difference threshold, that is, the three patterns are evenly distributed and there is no visually prominent pattern. At this time, the distribution result is the second characterization result. Calculate the distance between the geometric center of each pattern and each three-point line intersection and select the intersection with the smallest distance value as its corresponding target intersection. If the geometric center of pattern 1 is , then the distance from pattern 1 to the first intersection is Similarly, the distances from pattern 1 to the other three intersection points are , so the shortest distance is , select the first intersection as its target intersection, , if the geometric center of pattern 2 is , the geometric center of pattern 3 is , similarly we can calculate , , take half of the diagonal length of the canvas as the preset maximum distance, , then the mean value of the degree of conformity between the geometric center of each pattern and the intersection of the three-point line is .

[0137] It should be noted that the method of performing ratio analysis on the pattern feature data in the composition element data to obtain the composition ratio data includes:

[0138] S2431. Sort the patterns by size according to the number of pixels in the pattern feature data, obtain the first preset number of patterns as the main pattern according to the preset number of main patterns, obtain the width of the main pattern by subtracting the horizontal coordinates of the leftmost and rightmost pixels of the main pattern, and obtain the height of the main pattern by subtracting the vertical coordinates of the topmost and bottommost pixels of the main pattern.

[0139] S2432: Obtain pattern width ratio based on main pattern width and main pattern height and pattern height ratio for:

[0140] ;

[0141] .

[0142] in It is The width of the main pattern, It is The width of the main pattern, It is The height of the main pattern, It is The height of the main pattern, It's a pattern With pattern The width ratio, It's a pattern With pattern height ratio.

[0143] The number of preset main patterns is set according to the total number of patterns. When the total number of patterns is small, the corresponding number of preset main patterns is small. When the total number of patterns is large, the corresponding number of preset main patterns is large. For example, in this embodiment, the preset number of main patterns is set to 3, that is, the first 3 patterns are obtained as the main patterns according to the pattern size for proportional analysis.

[0144] It should be noted that the method of performing color analysis on the standardized work image to obtain the work color matching feature data, and performing style matching using a style recognition model based on the work composition element distribution feature data and the work color matching feature data to obtain the work style includes:

[0145] S31. Extract the RGB values ​​of the image through the pixel traversal algorithm, convert the image RGB values ​​into image HSV values ​​through the color space conversion algorithm, cluster the image HSV values ​​through the K-means clustering algorithm to obtain the center values ​​of the main color clusters of the image and the proportions of the main color clusters as the color matching feature data of the work; cluster the image HSV values ​​through the K-means clustering algorithm to obtain the center values ​​of the main color clusters of the image and the proportions of the main color clusters, for example, color cluster 1: H:30, S:65, V:72, accounting for 35%, color cluster 2: 92, S:70, V:56, accounting for 21%, color cluster 3: H:205, S:44, V:78, accounting for 20%, color cluster 4: H:6, S:3, V:96, accounting for 7%, color cluster 5: H:5, S:12, V:26, accounting for 17%.

[0146] S32: Inputting the composition ratio data and the color matching feature data of the work composition element distribution feature data into a style recognition model for style matching to obtain the work style, wherein the work styles include realistic style, cartoon style, anime style, impressionist style, abstract style, and Chinese style. The steps of establishing the style recognition model include:

[0147] A1. Collect images of paintings and extract composition ratio data and color matching feature data from the composition element distribution feature data. Label each work with style labels to obtain raw data. Use 70% of the raw data as training data to train the style recognition model using a neural network algorithm.

[0148] A2. The input layer receives the distribution feature data of composition elements and color matching feature data. The middle layer uses the ReLU activation function, and the output layer uses the softmax activation function to set 6 neurons to output the probability of the style of the work.

[0149] A3. Use the cross entropy loss function to train the style recognition model using the training dataset according to the forward propagation and backpropagation algorithms through the Adam optimizer.

[0150] A4. Use 30% of the original data as validation data to verify the style recognition model and adjust the model based on the validation results.

[0151] It should be noted that the method of comparing and analyzing the composition element distribution feature data and the color matching feature data of the work according to the preset style standard indicators to obtain the work evaluation results according to the work style, and providing students with specific improvement suggestions and pushing adapted learning courses based on the work evaluation results includes:

[0152] S41. Set style standard indicators based on the composition element distribution feature data and color matching feature data of historical excellent works in each style, including line smoothness, composition balance, composition ratio, image main color cluster range value and color cluster proportion of each style work.

[0153] S42. Compare the distribution feature data of the composition elements and the color matching feature data of the work with the style standard indicators according to the style of the work to obtain the work evaluation results, and provide students with specific improvement suggestions and push adapted learning courses based on the deviation items in the work evaluation results; the improvement suggestions include but are not limited to line practice, color matching improvement, composition technique optimization, style detail enhancement, etc.; the adapted learning courses include line teaching, composition teaching, color teaching, style-specific teaching, etc. For example: the realistic style standard is greater than 1.2, the composition balance is 0.3-0.5, the ratio of people to objects in the composition is the real proportion data, the main color cluster range values ​​and color cluster proportions of the image include: Color cluster 1, namely earth color: H:20-40, S:50-70, V:60-80, accounting for 30%-50%, Color cluster 2, namely natural green: H:80-100, S:60-80, V:40-60, accounting for 20%-40%, Color cluster 3, namely sky blue: H:200-220, S:40-60, V:70-90, accounting for 10%-30%, Color cluster 4, highlight color: H:0-20, S:0-20, V:90-100, accounting for 5%-15%, Color cluster 5, shadow color: H:0-20, S:0-20, V:10- 30, accounting for 5%-15%. If the line smoothness data of the student's work is 0.9, the composition balance data is 0.5, the proportion of people and objects in the composition ratio conforms to the realistic proportion, the main color clusters of the image are color cluster 1: H:30, S:65, V:72, accounting for 35%, color cluster 2: 92, S:70, V:56, accounting for 21%, color cluster 3: H:205, S:44, V:78, accounting for 20%, color cluster 4: H:6, S:3, V:96, accounting for 7%, color cluster 5: H:5, S:12, V:26, accounting for 17%, then after comparative analysis, the student's deviation from the realistic style is mainly due to the deviation in line smoothness and the deviation in the proportion of color cluster 5. It is recommended to strengthen the control and fluency of lines and the control of shadow color ratio, and promote courses related to line teaching and color teaching. Example 2

[0154] Based on the same inventive concept, Figure 2 As shown, this embodiment provides an intelligent teaching management system based on big data mining, which includes: an image preprocessing module, a composition analysis module, a color and style analysis module, and a feedback module, and the modules are connected in sequence.

[0155] The image preprocessing module is used to obtain students' paintings and preprocess the paintings to obtain standardized work images.

[0156] The composition analysis module is used to extract composition elements from the standardized work image through the element recognition model to obtain work composition element data, and perform composition analysis on the composition element data to obtain composition element distribution feature data.

[0157] The color and style analysis module is used to perform color analysis on the standardized work image to obtain the work color matching feature data, and to obtain the work style by performing style matching based on the work composition element distribution feature data and the work color matching feature data through a style recognition model. The style recognition model is a mapping relationship between the composition element distribution feature data, the work color matching feature data and the work style established based on the collected historical big data of the painting works.

[0158] The feedback module is used to compare and analyze the composition element distribution feature data and the color matching feature data of the work according to the preset style standard indicators according to the style of the work to obtain the work evaluation results, provide students with specific improvement suggestions and push adapted learning courses based on the work evaluation results.

[0159] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in Example 1 of the method, and will not be elaborated on here.

[0160] Finally, it should be noted that although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An intelligent teaching management method based on big data mining, characterized in that: The method comprises: Obtaining students' paintings, and preprocessing the paintings to obtain standardized work images; Extract composition elements from the standardized work image through an element recognition model to obtain composition element data of the work, and perform composition analysis on the composition element data to obtain composition element distribution feature data; Performing color analysis on the standardized work images to obtain the work color matching feature data, and performing style matching through a style recognition model based on the work composition element distribution feature data and the work color matching feature data to obtain the work style, wherein the style recognition model is a mapping relationship between the composition element distribution feature data, the work color matching feature data and the work style established based on the collected historical big data of the painting works; According to the style of the work, the distribution feature data of the composition elements and the color matching feature data of the work are compared and analyzed according to the preset style standard indicators to obtain the work evaluation results, and specific improvement suggestions are provided to students based on the work evaluation results and adapted learning courses are pushed; The method of extracting composition elements from a standardized work image through an element recognition model to obtain work composition element data, and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes: The texture feature map is obtained by performing texture detection on the standardized work image through the gray-level co-occurrence matrix; The texture feature map is used to identify straight lines and curves in the image through the Hough transform algorithm to obtain line feature data; The texture feature map is segmented by a watershed algorithm to obtain a texture region group, and the texture region group is matched with a preset pattern to obtain pattern feature data; The line feature data and the pattern feature data are used as composition element data, and the composition element data are subjected to multi-level composition analysis to obtain composition element distribution feature data.

2. According to claim 1, an intelligent teaching management method based on big data mining is characterized in that: The method of using line feature data and pattern feature data as composition element data and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes: Line smoothness data is obtained by performing line smoothness evaluation through Fourier transformation according to line feature data in the composition element data; According to the pattern feature data in the composition element data, the composition balance data is obtained by position distribution analysis; Performing ratio analysis according to pattern feature data in the composition element data to obtain composition ratio data; The line smoothness data, composition balance data, and composition ratio data are summarized to obtain the composition element distribution feature data.

3. The intelligent teaching management method based on big data mining according to claim 2 is characterized in that: The method for obtaining line smoothness data by performing line smoothness evaluation through Fourier transform according to line feature data in the composition element data comprises: Obtaining the length of the line, and obtaining the number of data points according to a preset data point acquisition number rule based on the length of the line, wherein the data point acquisition number rule is a mapping relationship between the line length and the number of data points established according to the ratio of the line length to the unit length; The coordinate system is established with the intersection of the left and lower boundaries of the image as the origin. The coordinates of the data points on the line are obtained at uniform intervals according to the number of data points obtained. The data points are converted from the spatial domain to the frequency domain using Fourier transform to obtain the frequency components. for: ; in is the data point index, It is The complex number formed by the coordinates of the data points, is the total number of data points, is an imaginary unit, is the index of the frequency component, ranging from arrive , It is frequency components; According to the frequency components, the amplitude spectrum of the frequency components is obtained for: ; in It is The real part of the frequency component, It is The imaginary part of the frequency component, It is The amplitude spectrum of the frequency components; According to the preset frequency component cutoff ratio, the frequency component before the cutoff of the frequency component index is selected as the low-frequency component, and the frequency component after the cutoff of the frequency component index is selected as the high-frequency component. The line smoothness is obtained according to the low-frequency component and the high-frequency component: ; in is the maximum frequency index of the low-frequency component, is the minimum frequency index of the high-frequency component, It is Line smoothness of the lines; Get the line smoothness data of the entire image based on the line smoothness of each line for: ; in is the total number of lines, is the line smoothness data.

4. The intelligent teaching management method based on big data mining according to claim 2 is characterized in that: The method for obtaining composition balance data by position distribution analysis based on pattern feature data in composition element data comprises: Get the pixel coordinates of each pattern in the pattern feature data, and calculate the geometric center of the pattern based on the pixel coordinates for: ; ; in It's a pattern Pixel area, It is a point in the region The horizontal axis of It is a point in the region The vertical coordinate of is the number of pixels in the pattern, It's a pattern The geometric center abscissa, It's a pattern The vertical coordinate of the geometric center; According to the geometric center of each pattern, the geometric center distance of the pattern is: ; in It's a pattern The geometric center abscissa, It's a pattern The vertical coordinate of the geometric center, It's a pattern The geometric center abscissa, It's a pattern The vertical coordinate of the geometric center, It's a pattern Geometric Center and Pattern Geometric center distance; The shortest distance between each pattern and its geometric center is selected to obtain a data set of adjacent pattern distances for each pattern, and a mean value and a standard deviation of pattern distances are calculated based on the mean value and the standard deviation of the data set of adjacent pattern distances for each pattern; The pattern distance standard deviation is compared with the pattern distance average value, and when the difference between the pattern distance standard deviation and the pattern distance average value is greater than a preset difference threshold, the distribution result is obtained as a first characterization result; when the difference between the pattern distance standard deviation and the pattern distance average value is less than or equal to the preset difference threshold, the distribution result is obtained as a second characterization result; The image of the work is divided into thirds by the method of thirds to obtain the coordinates of the intersection of the thirds lines. When the distribution result is the first characterization result, the pattern with the largest distance between adjacent patterns is selected as the visually prominent pattern, and the degree of conformity between the visually prominent pattern and the intersection of the thirds lines is calculated as the composition balance data. When the distribution result is the second characterization result, the mean degree of conformity between the geometric center of each pattern and the intersection of the thirds lines is calculated as the composition balance data.

5. The intelligent teaching management method based on big data mining according to claim 4 is characterized in that: The method of dividing the work image by the method of thirds to obtain the coordinates of the intersection of the thirds lines, selecting the pattern with the largest distance between adjacent patterns as the visually prominent pattern when the distribution result is the first characterization result, calculating the conformity between the visually prominent pattern and the intersection of the thirds lines as the composition balance data, and calculating the mean conformity between the geometric center of each pattern and the intersection of the thirds lines as the composition balance data when the distribution result is the second characterization result includes: Divide the work image into nine equal areas using the rule of thirds to obtain four intersection points of the third lines, and obtain the coordinates of the intersection points of the third lines based on the length and width of the work image; When the distribution result is the first characterization result, the pattern with the largest distance between adjacent patterns is selected as the visually prominent pattern, the distance between the geometric center of the visually prominent pattern and each three-line intersection is calculated, and the intersection with the smallest distance value is selected as the target intersection point. The conformity between the visually prominent pattern and the three-line intersection point is obtained by comparing the distance between the geometric center of the pattern and the target intersection point with the preset maximum distance. for: ; in is the distance between the geometric center of the visually salient pattern and the target intersection point, is the preset maximum distance, It is the degree of conformity between the visually prominent pattern and the intersection of the third lines; When the distribution result is the second characterization result, the distance between each pattern geometric center and each three-line intersection is calculated and the intersection with the smallest distance value is selected as its corresponding target intersection point. The distance between each pattern geometric center and its corresponding target intersection point is compared with the preset maximum distance to obtain the conformity between each pattern and the three-line intersection point. The mean of the conformity between each pattern geometric center and the three-line intersection point is calculated based on the mean of the conformity between each pattern geometric center and the three-line intersection point. for: ; in is the number of patterns, It is The distance between a pattern and its corresponding target intersection point, It is the mean of the conformity between the geometric center of each pattern and the intersection of the three-point line.

6. The intelligent teaching management method based on big data mining according to claim 2 is characterized in that: The method for obtaining composition ratio data by performing ratio analysis according to pattern feature data in composition element data comprises: Sort the patterns by size according to the number of pixels in the pattern feature data, obtain the first preset patterns as the main patterns according to the preset number of main patterns, subtract the horizontal coordinates of the leftmost and rightmost pixels of the main pattern to obtain the width of the main pattern, and subtract the vertical coordinates of the uppermost and lowermost pixels of the main pattern to obtain the height of the main pattern; Get the pattern width ratio based on the main pattern width and the main pattern height and pattern height ratio for: ; ; in It is The width of the main pattern, It is The width of the main pattern, It is The height of the main pattern, It is The height of the main pattern, It's a pattern With pattern The width ratio, It's a pattern With pattern height ratio.

7. The intelligent teaching management method based on big data mining according to claim 6 is characterized in that: The method of performing color analysis on the standardized work image to obtain the work color matching feature data, and performing style matching through a style recognition model according to the work composition element distribution feature data and the work color matching feature data to obtain the work style includes: The image RGB value is extracted through the pixel traversal algorithm, and the image RGB value is converted into the image HSV value through the color space conversion algorithm. The image HSV value is clustered by the K-means clustering algorithm to obtain the center value of the main color cluster of the image and the proportion of the main color cluster as the color matching feature data of the work; Input the composition ratio data and the color matching feature data of the work in the composition element distribution feature data into the style recognition model for style matching to obtain the style of the work, wherein the style of the work includes realistic style, cartoon style, animation style, impressionist style, abstract style, and Chinese style; the steps of establishing the style recognition model include: Collect images of paintings and extract composition ratio data and color matching feature data from the composition element distribution feature data. Label each work with style labels to obtain original data. Use 70% of the original data as training data to train the style recognition model through a neural network algorithm. The input layer receives the distribution feature data of composition elements and color matching feature data, the middle layer uses the ReLU activation function, and the output layer uses the softmax activation function to set the probability of 6 neurons outputting the style of the work; The cross entropy loss function is used to train the style recognition model using the training dataset according to the forward propagation and back propagation algorithms through the Adam optimizer; 30% of the original data is used as verification data to verify the style recognition model and the model is adjusted based on the verification results.

8. The intelligent teaching management method based on big data mining according to claim 1 is characterized in that: The method of comparing and analyzing the distribution feature data of composition elements and the color matching feature data of the work according to the preset style standard index to obtain the work evaluation result according to the style of the work, and providing specific improvement suggestions to students and pushing adapted learning courses according to the work evaluation result includes: The style standard indicators are set according to the composition element distribution characteristic data and the color matching characteristic data of the historical excellent works in each style, including the line smoothness, composition balance, composition ratio, image main color cluster range value and color cluster proportion of each style of works; According to the style of the work, the distribution feature data of the composition elements and the color matching feature data of the work are compared with the style standard indicators to obtain the work evaluation results. According to the deviation items in the work evaluation results, specific improvement suggestions are provided to students and adapted learning courses are pushed.

9. An intelligent teaching management system based on big data mining, used to execute the method described in any one of claims 1 to 8, characterized in that: The system comprises: an image preprocessing module, a composition analysis module, a color and style analysis module, and a feedback module, wherein the modules are connected in sequence; The image preprocessing module is used to obtain students' paintings and preprocess the paintings to obtain standardized work images; The composition analysis module is used to extract composition elements from the standardized work image through the element recognition model to obtain work composition element data, and to perform composition analysis on the composition element data to obtain composition element distribution feature data; The color and style analysis module is used to perform color analysis on the standardized work image to obtain the work color matching feature data, and to perform style matching through a style recognition model based on the work composition element distribution feature data and the work color matching feature data to obtain the work style. The style recognition model is a mapping relationship between the composition element distribution feature data, the work color matching feature data and the work style established based on the collected historical big data of the painting works; The feedback module is used to compare and analyze the composition element distribution feature data and the color matching feature data of the work according to the preset style standard indicators according to the style of the work to obtain the work evaluation results, and provide students with specific improvement suggestions and push adapted learning courses according to the work evaluation results; The method of extracting composition elements from a standardized work image through an element recognition model to obtain work composition element data, and performing multi-level composition analysis on the composition element data to obtain composition element distribution feature data includes: The texture feature map is obtained by performing texture detection on the standardized work image through the gray-level co-occurrence matrix; The texture feature map is used to identify straight lines and curves in the image through the Hough transform algorithm to obtain line feature data; The texture feature map is segmented by a watershed algorithm to obtain a texture region group, and the texture region group is matched with a preset pattern to obtain pattern feature data; The line feature data and the pattern feature data are used as composition element data, and the composition element data are subjected to multi-level composition analysis to obtain composition element distribution feature data.