A method for automatically evaluating the correctness of handwritten Chinese character structure based on template matching

Abstract

The application relates to a handwriting Chinese character structure right-wrong automatic evaluation method based on template matching, which comprises the following steps: S1, collecting template Chinese characters by taking correctly written Chinese characters as the template Chinese characters; S2, carrying out data preprocessing based on the collected template Chinese characters; S3, extracting features from the preprocessed template Chinese characters and automatically classifying the features to obtain necessary features and ignorable features, and storing the necessary features and the ignorable features as feature templates; S4, obtaining handwriting Chinese characters to be tested and carrying out data preprocessing; S5, extracting features from the preprocessed handwriting Chinese characters to be tested; S6, comparing the features of the handwriting Chinese characters to be tested obtained in the step S5 with the feature templates obtained in the step S3; and S7, feeding back a test result according to a feature comparison result. The application realizes automatic feature classification and handwriting Chinese character structure right-wrong evaluation, develops a computer-aided handwriting Chinese character practice system, and obtains a good test result.

Description

Technical Field

[0001] This invention relates to the technical field of evaluating the structure of handwritten Chinese characters using computer technology, and in particular to an automatic evaluation method for the correctness of handwritten Chinese character structure based on template matching. Background Technology

[0002] Chinese characters, as the crystallization of thousands of years of Chinese civilization, serve as an excellent carrier and link for inheriting and disseminating Chinese culture. With the increasing popularity of Chinese worldwide, the learning of Chinese characters is receiving more and more attention. Traditional methods of teaching Chinese characters heavily rely on teachers' manual assessment, which is labor-intensive and time-consuming, failing to provide rapid feedback to learners. In recent years, with the rapid development of information technology, many computer-aided Chinese character writing teaching systems have been implemented.

[0003] Unlike Western alphabetic writing systems, Chinese characters not only emphasize the aesthetics of their overall layout but also adhere to writing standards regarding stroke order, stroke placement, and other aspects. Chinese characters comprise four levels, from highest to lowest: the character itself, its components, strokes, and stroke shapes. Based on the number of components, Chinese characters can be further divided into single-component characters and multi-component characters. Single-component characters are also called simple structures, while multi-component characters exhibit 12 different structural relationships between their components, such as top-bottom and left-right arrangements. In the process of learning Chinese characters, learners not only make mistakes at the stroke level (stroke order, stroke placement, etc.) but also exhibit various errors or non-standardizations in the structure of the character's components.

[0004] Currently, various handwritten Chinese character evaluation systems exist, such as An Weihua et al.'s "Automatic Evaluation Method and Device for Correctness of Chinese Character Writing Based on Digital Handwriting Devices," Jin Lianwen et al.'s "A Method for Evaluating the Layout of Handwritten Chinese Characters" and "A Method for Evaluating the Quality of Chinese Character Writing Based on Confidence," and Huang Feng et al.'s "A Method for Evaluating the Neatness of Handwritten Chinese Characters Based on the Annotation of Important Writing Features," etc. Existing handwritten Chinese character evaluation systems either focus on the overall aesthetics of the character layout or emphasize errors and irregularities in strokes, stroke order, and stroke style, with few methods addressing the evaluation of the correctness of the structural components of handwritten Chinese characters. Summary of the Invention

[0005] This invention addresses the aforementioned problems by providing an automatic evaluation method for the correctness of handwritten Chinese character structure based on template matching. This invention solves the problem of the lack of evaluation for the correctness and standardization of Chinese character component structure in existing computer-aided Chinese character teaching systems. It can automatically evaluate the correctness of the component structure of Chinese characters written by learners and provide feedback on errors or non-standardizations, enabling learners to complete handwritten Chinese character practice in unattended environments.

[0006] To achieve the above objectives, the specific solution of the present invention is as follows:

[0007] An automatic evaluation method for the structural correctness of handwritten Chinese characters based on template matching includes the following steps:

[0008] S1. Collect template Chinese characters by using correctly written Chinese characters as template characters;

[0009] S2. Based on the collected template Chinese characters, perform data preprocessing;

[0010] S3. Extract features from the preprocessed template Chinese characters and automatically classify the features to obtain necessary features and negligible features. Store the necessary features and negligible features as feature templates.

[0011] S4. Obtain the handwritten Chinese characters to be tested and perform data preprocessing;

[0012] S5. Extract features from the preprocessed handwritten Chinese characters to be tested;

[0013] S6. Compare the features of the handwritten Chinese characters to be tested obtained in step S5 with the feature template obtained in step S3.

[0014] S7. Based on the feature comparison results, provide feedback on the test results.

[0015] Preferably, the data preprocessing in step S2 includes data augmentation, data standardization, and interpolation.

[0016] Preferably, the feature extraction in step S3 includes the calculation of component features, and the specific calculation method is as follows:

[0017] S31. Suppose a Chinese character has n strokes. Based on the number of strokes n, the point set data of the Chinese character is represented as follows:

[0018]

[0019] S32. Calculate the bounding rectangle of Chinese characters.

[0020] First calculate the line(X) of each stroke. min X max Y min Y max Then, by calculating Box(X) in all strokes of the Chinese character... min X max Y min Y max ),Right now,

[0021] Box x_max =Max(line1X) max， line2X max ...line n X max ),

[0022] Box X _ min =Min(line) r X min， line2X min ...line n X min ),

[0023] Box Y_max =Max(line1Y) max， line2Y max ...line n Y max ),

[0024] Box Y_min =Max(line1Y) min line2Y min ...line n Y max );

[0025] S33, the outer rectangle of the calculation component;

[0026] S34, Centroid Calculation;

[0027] When writing Chinese characters, the writing speed of each stroke is not consistent, which leads to inaccuracy in calculating the centroid of the stroke. It is necessary to unify the dot density of each stroke.

[0028] Calculate the length of the strokes:

[0029]

[0030] Let density = 10, then the number of points in a stroke is:

[0031]

[0032] but

[0033]

[0034]

[0035] After isodense calculation, the new stroke representation is as follows:

[0036]

[0037] Calculate the centroid M of the uniformly dense strokes, and use it as the centroid of the current stroke:

[0038]

[0039]

[0040] Furthermore, the feature extraction in step S3 also includes feature selection based on a preset feature category, specifically using the following strategy:

[0041] Feature Category 1: The space occupied by the outer frame of the component in each cell of the whole character grid, with 1 for occupied and 0 for unoccupied;

[0042] Feature Category 2: The space occupied by the center of the component's outer frame in each cell of the whole character grid; occupies 1 space, and does not occupy 0 space.

[0043] Feature category 3: The space occupied by the component's centroid in each grid cell of the whole character; occupies 1 space and does not occupy 0 space.

[0044] Feature Category 4: The space occupied by the component strokes in each grid of the whole character; occupies 1 space and does not occupy 0 space.

[0045] Feature category 5: Whether the boundaries of the component outline and the Chinese character outline are in contact; 1 indicates contact, 0 indicates no contact.

[0046] Feature Category 6: The position of the centroid of a component relative to the centroid of another component, represented by quadrants. The quadrant in which the centroid of a component is located is denoted as 1, and the quadrant in which the centroid of a component is not located is denoted as 0. For example, if the centroid of the other component is set as the origin, the quadrant in which the centroid of the current component is located is determined by the angle relationship between the vectors of the current centroid of the component and the centroid of the other component set as the origin and the X-axis. If the current centroid of the component is above the X-axis, the angle is the angle between the two vectors. If the current centroid of the component is below the X-axis, the angle is 360° minus the angle between the two vectors.

[0047] Feature category 7: Whether the bounding boxes of the two components contain each other, 1 for containing and 0 for not containing;

[0048] Feature category 8: Whether the overlapping length of the outer frames of the two components in the x-axis and y-axis directions is greater than 2 / 3 of the length and height of the outer frame of the smaller component; if greater, it is 1, otherwise it is 0.

[0049] Feature category 9: Whether the overlapping area of ​​the outer frames of the two components is greater than 2 / 3 of the area of ​​the outer frame of the smaller component, if greater, is 1, otherwise is 0; whether it is greater than 1 / 3 of the area of ​​the outer frame of the larger component, if greater, is 1, otherwise is 0;

[0050] Feature category 10: Whether the length (height) ratio of the outer frame of the two components is greater than 3 or less than 1 / 3, if it meets the condition, it is 1; otherwise, it is 0.

[0051] Feature Category 11: The outer frame of the component in the top left, middle, and bottom right grids of the nine-grid pattern has other components in the positive and negative directions of the x and y axes. If there are other components, it is 1; otherwise, it is 0.

[0052] Preferably, feature category 6 includes two types of quadrants: four quadrants or eight quadrants. For the four quadrants, the first quadrant is [0°, 90°), the second quadrant is [90°, 180°), the third quadrant is [180°, 270°), and the fourth quadrant is [270°, 360°); for the eight quadrants, the first quadrant is [0°, 45°), the second quadrant is [45°, 90°), the third quadrant is [90°, 135°), the fourth quadrant is [135°, 180°), the fifth quadrant is [180°, 225°), the sixth quadrant is [225°, 270°), the seventh quadrant is [270°, 315°), and the eighth quadrant is [315°, 360°).

[0053] Preferably, in step S3, feature classification is automatically performed to obtain essential features and ignorable features. The specific method is as follows:

[0054] Compare the corresponding features of different template Chinese character samples corresponding to the same Chinese character. If the feature values of all samples are equal, mark this feature as an essential feature of the corresponding template Chinese character. If different samples have different feature values, mark this feature as an ignorable feature.

[0055] Preferably, step S7 includes:

[0056] If all essential features match, return correct. If there are unmatched essential features, return the corresponding component and prompt the practitioner that the structure where the component is located is incorrect.

[0057] Using the method for evaluating the correctness of the structure of handwritten Chinese character components of the present invention, first design features for different types of component structure errors, collect correct template Chinese characters and extract the features of the template Chinese characters, automatically classify the template features by comparing multiple correctly structured samples, extract the features of the handwritten Chinese characters to be tested, combine the classification of the template features and compare with the template, and feedback the comparison result to the practitioner. If there is a structure error, prompt the practitioner with the incorrect component structure. The present invention realizes automatic feature classification and evaluation of the correctness of the structure of handwritten Chinese characters, develops a computer-aided handwritten Chinese character practice system, and obtains good test results. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 is a flowchart of the present invention;

[0059] Figure 2 is a template Chinese character of the present invention taking "Nie" as an example;

[0060] Figure 3 is a display of the evaluation result of the present invention taking "Nie" as an example. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0061] To facilitate the understanding and implementation of the present invention by those of ordinary skill in the art, the embodiments of the present invention will be further described below in conjunction with the accompanying drawings.

[0062] Referring to Figures 1 to 3 , the present invention provides a method for automatically evaluating the correctness of the structure of handwritten Chinese characters based on template matching, including the following steps:

[0063] S1. Use the correctly written Chinese characters as template Chinese characters and collect the template Chinese characters.

[0064] For example, Figure 2 among the 4 template Chinese characters "Nie" in

[0065] , "Ear" and "Double" are in the upper and lower structures, and the two "You" are in the left and right structures. Different parts are marked with different colors. For example, "Ear" is marked blue, the left "You" is marked red, and the right "You" is marked green.

[0066] The data preprocessing includes data augmentation, constructing more template Chinese characters with correct component structures based on the collected correctly written template Chinese characters. The data preprocessing also includes standardization, interpolation, etc. of handwritten Chinese character data.

[0067] S3. Extract features from the preprocessed template Chinese characters and automatically classify the features to obtain essential features and ignorable features, and store the essential features and ignorable features as a feature template.

[0068] Among them, the feature extraction in step S3 includes the calculation of component features. The specific calculation method is as follows:

[0069] S31. Assume that a Chinese character has n strokes. According to the number of strokes n, the point set data of the Chinese character is represented as:

[0070]

[0071] S32. Calculate the bounding rectangle of the Chinese character

[0072] First, calculate the line(X min , X max , Y min , Y max ) of each stroke, and then calculate the Box(X min , X max , Y min , Y max ) of all strokes in the Chinese character, that is,

[0073] Box x_max = Max(line1X max , line2X max...line n X max ),

[0074] Box x_min =Min(line1X min ,line2X min …line n X min ),

[0075] Box Y_max =Max(line1Y max ,line2Y max …line n Y max ),

[0076] Box Y_min =Max(line1Y min ,line2Y min …line n Y min );

[0077] S33. Calculate the external rectangular box of the component;

[0078] For example: The Chinese character "明" is composed of the components "日" and "月". Among them, the component "日" is written with 4 strokes from 1 - 4, and the component "月" is written with 4 strokes from 5 - 8. Therefore, the calculation of the external rectangular box of the component "日" is as follows:

[0079] Box 日X_max =Max(line1X max ,line2X max ...lire4X max );

[0080] Box 日X_min =Min(lire1X min ,line2X min ...line4X min );

[0081] Box 日Y_max =Max(line1Y max ,line2Y max ...lire4Y max );

[0082] Box 日Y_min =Max(line1Y min ,lire2Y min ...line4Y min );

[0083] The bounding rectangle of component "moon" is calculated as follows:

[0084] Box 月X_max =Max(line5X) max ,line6X max ...line8X max );

[0085] Box 月X_min =Min(line5X) min ,line6X min ...line8X min );

[0086] Box 月Y_max =Max(line5Y) max line6Y max ...line8Y max );

[0087] Box 月Y_min =Max(line5Y) min lie6Y min ...line8Y min );

[0088] S34, Centroid Calculation;

[0089] When writing Chinese characters, the writing speed of each stroke is not consistent, which leads to inaccuracy in calculating the centroid of the stroke. It is necessary to unify the dot density of each stroke.

[0090] Calculate the length of the strokes:

[0091]

[0092] Let density = 10, then the number of points in a stroke is:

[0093]

[0094] but

[0095]

[0096]

[0097] After isodense calculation, the new stroke representation is as follows:

[0098]

[0099] Calculate the centroid M of the uniformly dense strokes, and use it as the centroid of the current stroke:

[0100]

[0101]

[0102] Furthermore, the feature extraction in step S3 also includes feature selection based on a preset feature category, specifically using the following strategy:

[0103] The corresponding features of different template Chinese character samples corresponding to the same Chinese character are compared. If the feature value of all samples is equal, the feature is marked as a necessary feature of the corresponding template Chinese character. If different samples have different feature values, the feature is marked as an ignorable feature.

[0104] Furthermore, step S3, automatic feature classification, also includes feature selection based on feature categories, with the specific strategy as follows:

[0105] Feature Category 1: The space occupied by the outer frame of the component in each cell of the whole character grid, with 1 for occupied and 0 for unoccupied;

[0106] Feature Category 2: The space occupied by the center of the component's outer frame in each cell of the whole character grid; occupies 1 space, and does not occupy 0 space.

[0107] Feature category 3: The space occupied by the component's centroid in each grid cell of the whole character; occupies 1 space and does not occupy 0 space.

[0108] Feature Category 4: The space occupied by the component strokes in each grid of the whole character; occupies 1 space and does not occupy 0 space.

[0109] Feature category 5: Whether the boundaries of the component outline and the Chinese character outline are in contact; 1 indicates contact, 0 indicates no contact.

[0110] Feature Category 6: The position of the centroid of a component relative to the centroid of another component, represented by quadrants. The quadrant in which the centroid of a component is located is denoted as 1, and the quadrant in which the centroid of a component is not located is denoted as 0. For example, if the centroid of the other component is set as the origin, the quadrant in which the centroid of the current component is located is determined by the angle relationship between the vectors of the current centroid of the component and the centroid of the other component set as the origin and the X-axis. If the current centroid of the component is above the X-axis, the angle is the angle between the two vectors. If the current centroid of the component is below the X-axis, the angle is 360° minus the angle between the two vectors. Preferably, there are two quadrants: four quadrants or eight quadrants. The first quadrant is [0°, 90°), the second quadrant is [90°, 180°), the third quadrant is [180°, 270°), and the fourth quadrant is [270°, 360°]. The eight quadrants are represented as follows: the first quadrant is [0°, 45°), the second quadrant is [45°, 90°), the third quadrant is [90°, 135°), the fourth quadrant is [135°, 180°), the fifth quadrant is [180°, 225°), the sixth quadrant is [225°, 270°), the seventh quadrant is [270°, 315°), and the eighth quadrant is [315°, 360°].

[0111] Feature category 7: Whether the bounding boxes of the two components contain each other, 1 for containing and 0 for not containing;

[0112] Feature category 8: Whether the overlapping length of the outer frames of the two components in the x-axis and y-axis directions is greater than 2 / 3 of the length and height of the outer frame of the smaller component; if greater, it is 1, otherwise it is 0.

[0113] Feature category 9: Whether the overlapping area of ​​the outer frames of the two components is greater than 2 / 3 of the area of ​​the outer frame of the smaller component, if greater, is 1, otherwise is 0; whether it is greater than 1 / 3 of the area of ​​the outer frame of the larger component, if greater, is 1, otherwise is 0;

[0114] Feature category 10: Whether the length (height) ratio of the outer frame of the two components is greater than 3 or less than 1 / 3, if it meets the condition, it is 1; otherwise, it is 0.

[0115] Feature Category 11: The outer frame of the component in the top left, middle, and bottom right grids of the nine-grid pattern has other components in the positive and negative directions of the x and y axes. If there are other components, it is 1; otherwise, it is 0.

[0116] Furthermore, in step S3, feature classification is automatically performed to obtain necessary features and negligible features. The specific method is as follows:

[0117] The corresponding features of different template Chinese character samples corresponding to the same Chinese character are compared. If the feature value of all samples is equal, the feature is marked as a necessary feature of the corresponding template Chinese character. If different samples have different feature values, the feature is marked as an ignorable feature.

[0118] S4. Obtain the handwritten Chinese characters to be tested and perform data preprocessing;

[0119] For example, write Chinese characters through intelligent devices such as tablets and smartphones, automatically obtain the handwritten Chinese characters, and perform data preprocessing on them. The processing method is similar to that in step S2.

[0120] S5. Extract the features of the preprocessed handwritten Chinese characters to be tested;

[0121] The processing method in this step is similar to the feature extraction in step S3.

[0122] S6. Compare the features of the handwritten Chinese characters to be tested obtained in step S5 with the feature templates obtained in step S3;

[0123] S7. According to the results of the feature comparison, feedback the test results. If all the compared features are equal, feedback that the component structure of the handwritten Chinese character is correct. If there is a certain necessary feature that is not equal, return the corresponding component data and prompt the practitioner that there is an error in the component structure.

[0124] For example, refer to Figure 3 , the test results of the Chinese character "Nie". For example, the components in the incorrect structure can be marked in red:

[0125] In the first example, "ear" and "double" are written in an embedded structure, and "ear" can be marked in red;

[0126] In the second example, it is written in a left-right structure, and "ear" can be marked in red;

[0127] In the third example, the two "you" are written in an up-down structure, and the upper "you" can be marked in red;

[0128] In the fourth example, the component structure is correct.

[0129] The above embodiments only represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the appended claims.

Claims

1. A method for automatically evaluating the structural correctness of handwritten Chinese characters based on template matching, characterized in that, Includes the following steps: S1. Collect template Chinese characters by using correctly written Chinese characters as template characters; S2. Based on the collected template Chinese characters, perform data preprocessing; S3. Extract component-level features from the preprocessed template Chinese characters. By comparing the same feature value of multiple template samples, automatically classify them into necessary features or negligible features, and store the necessary features and negligible features as feature templates. S4. Obtain the handwritten Chinese characters to be tested and perform data preprocessing; S5. Extract features from the preprocessed handwritten Chinese characters to be tested; S6. Compare the features of the handwritten Chinese characters to be tested obtained in step S5 with the feature template obtained in step S3. S7. Based on the feature comparison results, provide feedback on the test results; If the required features do not match, the error type for the specific component is returned, including structural misalignment and proportional mismatch. Step S3 involves extracting features, including calculating component features. The specific calculation method is as follows: S31. Suppose a Chinese character has n strokes. Based on the number of strokes n, the point set data of the Chinese character is represented as follows: ...], ...], ┋ ...]}； S32. Calculate the bounding rectangle of Chinese characters. First calculate the line(s) of each stroke. , , , Then, by calculating the percentage of all strokes in the Chinese character... ( , , , ),Right now, )， )， )， )； S33, the outer rectangle of the calculation component; S34, Centroid Calculation; When writing Chinese characters, the writing speed of each stroke is not consistent, which leads to inaccuracy in calculating the centroid of the stroke. It is necessary to unify the dot density of each stroke. Calculate the length of the strokes: Let density = 10, then the number of points in a stroke is: but After isodense calculation, the new stroke representation is as follows: ； Calculate the centroid M of the uniformly dense strokes, and use it as the centroid of the current stroke: ， ； In step S3, feature classification is performed automatically to obtain necessary features and negligible features. The specific method is as follows: The corresponding features of different template Chinese character samples corresponding to the same Chinese character are compared. If the feature value of all samples is equal, the feature is marked as a necessary feature of the corresponding template Chinese character. If different samples have different feature values, the feature is marked as an ignorable feature.

2. The automatic evaluation method for the correctness of handwritten Chinese character structure based on template matching according to claim 1, characterized in that, The feature extraction in step S3 also includes feature selection based on a pre-defined feature category, specifically using the following strategy: Feature Category 1: The space occupied by the outer frame of the component in each cell of the whole character grid, with 1 for occupied and 0 for unoccupied; Feature Category 2: The space occupied by the center of the component's outer frame in each cell of the whole character grid; occupies 1 space, and does not occupy 0 space. Feature category 3: The space occupied by the component's centroid in each grid cell of the whole character; occupies 1 space and does not occupy 0 space. Feature Category 4: The space occupied by the component strokes in each grid of the whole character; occupies 1 space and does not occupy 0 space. Feature category 5: Whether the boundaries of the component outline and the Chinese character outline are in contact; 1 indicates contact, 0 indicates no contact. Feature category 6: The position of the centroid of a component relative to the centroid of another component, represented by quadrants. The quadrant in which the centroid of a component is located is recorded as 1, and the quadrant in which the centroid of a component is not located is recorded as 0. Feature category 7: Whether the bounding boxes of the two components contain each other, 1 for containing and 0 for not containing; Feature category 9: Whether the overlapping area of ​​the outer frames of the two components is greater than 2 / 3 of the area of ​​the outer frame of the smaller component, if greater, is 1, otherwise is 0; whether it is greater than 1 / 3 of the area of ​​the outer frame of the larger component, if greater, is 1, otherwise is 0; Feature category 10: Whether the length ratio of the outer frames of the two components is greater than 3 or less than 1 / 3, if it meets the condition, it is 1; otherwise, it is 0.

3. The automatic evaluation method for the correctness of handwritten Chinese character structure based on template matching according to claim 2, characterized in that, Feature category 6 includes two quadrants: four quadrants or eight quadrants. The first quadrant is [0°, 90°), the second quadrant is [90°, 180°), the third quadrant is [180°, 270°), and the fourth quadrant is [270°, 360°]. The eight quadrants are represented as follows: the first quadrant is [0°, 45°), the second quadrant is [45°, 90°), the third quadrant is [90°, 135°), the fourth quadrant is [135°, 180°), the fifth quadrant is [180°, 225°), the sixth quadrant is [225°, 270°), the seventh quadrant is [270°, 315°), and the eighth quadrant is [315°, 360°].

4. The automatic evaluation method for the correctness of handwritten Chinese character structure based on template matching according to claim 1, characterized in that, Step S7 includes: If all required features match, the system returns "correct". If any required features do not match, the system returns the corresponding component and indicates to the learner that the component is in an incorrect structure.

5. The automatic evaluation method for the correctness of handwritten Chinese character structure based on template matching according to claim 1, characterized in that, Step S2, data preprocessing, includes data augmentation, data standardization, and interpolation.

Images