An image acquisition and analysis method in a classroom teaching process
By collecting and analyzing images in the classroom, identifying drawing board areas, and calculating drawing features, the subjective nature of drawing quality assessment in traditional teaching is solved, enabling objective evaluation of students' drawing process and personalized teaching support.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HANGZHOU WANXIANG POLYTECHNIC
- Filing Date
- 2022-11-15
- Publication Date
- 2026-06-02
AI Technical Summary
Traditional teaching methods make it difficult to objectively and comprehensively assess the quality of students' painting process in art courses, resulting in an inability to provide individualized instruction.
By installing cameras in the classroom, images of the classroom are captured, the drawing board area is identified and segmented, the grid is subdivided, the feature values and color timing jitter of the cells are recorded, the positional offset value is calculated, and the value is compared with the standard vector value to evaluate the quality of the drawing.
It enables objective evaluation of students' drawing process, reduces the influence of subjective judgment, and supports personalized teaching.
Smart Images
Figure CN115661430B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of intelligent teaching, and in particular to a method for image acquisition and analysis during classroom teaching. Background Technology
[0002] Many professional courses in higher education institutions are practice-oriented, where the process and result of students completing in-class assignments are equally important. Art courses, such as sketching, drawing, watercolor, gouache, and oil painting, typically involve the whole class copying and creating together, with the instructor providing demonstrations and guidance. The final assessment of student learning quality relies solely on examining the final product. Instructors can only make highly subjective judgments based on very limited classroom observations regarding the rationality of the student's drawing process, the correctness of the steps, and the presence of any bad techniques. These judgments are often incomplete, inaccurate, and heavily influenced by subjective impressions. Therefore, traditional teaching methods struggle to objectively and comprehensively assess the quality of art classroom instruction, hindering the ability to tailor teaching to individual student needs and provide appropriate guidance. Summary of the Invention
[0003] To address one of the aforementioned technical problems, this invention provides a method for image acquisition and analysis during classroom teaching.
[0004] This invention provides an image acquisition and analysis method for classroom teaching, the method comprising:
[0005] The scene area is photographed at a preset frequency to obtain the captured images;
[0006] The acquired image is identified, and the canvas area is segmented from the acquired image;
[0007] The canvas area is subdivided into multiple cells by gridding;
[0008] Record the feature value for each cell and calculate the color timing jitter of the feature value for each cell;
[0009] The position offset value of each acquired image is obtained based on the color timing jitter of the feature value of each cell, and the actual vector value of each acquired image is obtained based on the position offset value of each acquired image.
[0010] The comparison result is obtained by comparing the actual vector value of each acquired image with the pre-set standard vector value of each acquired image.
[0011] Preferably, the process of identifying the acquired image and segmenting the canvas area from the acquired image includes:
[0012] The objects in the acquired images are identified and classified;
[0013] Based on the classification results, select objects that have pre-defined canvas features;
[0014] Position dimensional boundaries around the object having predefined canvas features;
[0015] The canvas area is divided according to the stated size boundaries.
[0016] Preferably, the process of subdividing the canvas area into multiple cells by gridding includes:
[0017] The horizontal dimension of the drawing board area is divided into N length units on an average basis.
[0018] The vertical dimension of the drawing board area is divided into M length units on an average basis.
[0019] Intersect the dividing lines on the horizontal and vertical dimensions of the canvas area to obtain M×N cells.
[0020] Preferably, the feature values of each cell include coordinates, color blending values, and timestamps, and the process of recording the feature values of each cell includes:
[0021] The coordinates of each cell are identified by the number of rows and columns in the canvas area where each cell is located;
[0022] Perform color mixing on the image within each cell to obtain the average color value within each cell;
[0023] The average color value is converted to HSB color mode, and the color mixing value of each cell is identified by hue, saturation, and lightness.
[0024] Record the shooting time of the scene area and convert the shooting time into a timestamp of the acquired image.
[0025] Preferably, the process of calculating the color timing jitter of the feature values of each cell includes:
[0026] The color blending values of cells at the same coordinate in the images captured at time j and time j-1 are transformed as follows:
[0027] C transform =C H *r H +C B *r B +C B *r B
[0028] Among them, C H CB C B These are the decimal values of the H, S, and B channels, respectively, representing the average color value C; r H r B r B These are the transform coefficients for the H, S, and B channels, respectively; C transform The transformed color blending value;
[0029] The color timing jitter of each cell is obtained by calculating the difference between the transformed color blending values of cells at the same coordinate in the images captured at time j and time j-1, and taking the absolute value of the difference.
[0030] Preferably, when j=0, the color timing jitter of the cell is 0.
[0031] Preferably, the process of obtaining the position offset value of each acquired image based on the color timing dithering of the feature values of each cell includes:
[0032] Compare the color timing jitter of all cells, filter out the cell with the largest color timing jitter, and record the coordinates of the cell with the largest color timing jitter in the canvas area;
[0033] The coordinates of the cell with the largest color timing jitter in the canvas area are reduced in dimension to obtain the position offset value of each acquired image.
[0034] Preferably, the process of obtaining the actual vector value of each acquired image based on the position offset value of each acquired image includes:
[0035] The position offset values of each acquired image are sorted according to the timestamp of each acquired image;
[0036] The actual vector value of each acquired image is obtained based on the sorting results.
[0037] Preferably, after subdividing the canvas area into multiple cells using a grid, the method further includes:
[0038] Perform object recognition on each cell to determine the category of objects within each cell's range;
[0039] Based on the object category, determine whether the object within the range of each cell is an obstruction. If it is an obstruction, mark the cell as an invalid cell; otherwise, mark the cell as a valid cell.
[0040] Preferably, when the cell is an invalid cell, the color blending value of the invalid cell is empty, and when the color blending value of the cell is empty, the color timing jitter of the cell is 0.
[0041] The beneficial effects of this invention are as follows: The image acquisition and analysis method proposed in this invention for classroom teaching processes describes and records the composition and drawing sequence of the drawing board area by capturing images of the actual area. These images are then compared with a standard model to analyze students' learning outcomes in composition and drawing sequence. This invention can utilize the cameras already installed in most classrooms for image acquisition, possessing good implementation conditions and cost advantages. The data volume and computational load of this invention are relatively small, resulting in low cost and ease of promotion. Attached Figure Description
[0042] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0043] Figure 1 This is a flowchart of the image acquisition and analysis method in the classroom teaching process described in an embodiment of the present invention. Detailed Implementation
[0044] To make the technical solutions and advantages of the embodiments of this application clearer, the exemplary embodiments of this application will be described in further detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not an exhaustive list of all embodiments. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.
[0045] like Figure 1 As shown in the figure, this embodiment proposes an image acquisition and analysis method in the classroom teaching process, which includes:
[0046] S101. Take pictures of the scene area at a preset frequency to obtain the acquired images.
[0047] Specifically, most classrooms are now equipped with high-definition cameras. If not, multiple cameras can be installed at relatively low cost. High-definition cameras can capture images of the classroom area at a certain frequency, creating a series of images. The shooting frequency can be adjusted according to the pace of teaching or drawing speed, generally 1-15 seconds is sufficient.
[0048] S102. Recognize the acquired image and segment the canvas area from the acquired image.
[0049] Specifically, the first step is to identify and classify objects in the acquired images. This identification process can be implemented using existing training models based on publicly available technologies such as YOLO V3 and ResNet; this embodiment does not impose any special limitations. Then, objects with predefined canvas features are selected based on the classification results. Canvas features can be defined by shape, such as a cuboid feature. Next, size boundaries are located around the objects with the predefined canvas features. Finally, the canvas region is segmented based on the size boundaries.
[0050] S103. The canvas area is subdivided into grids to obtain multiple cells.
[0051] Specifically, the canvas area is divided into N equal length units horizontally and M equal length units vertically. The intersection of these horizontal and vertical dividing lines yields M×N cells. The values of M and N can be adjusted based on factors such as camera resolution, the size of the area, and the canvas size; generally, a range of 15-50 is sufficient for basic functionality.
[0052] S104. Record the feature value for each cell and calculate the color timing jitter of the feature value for each cell.
[0053] Specifically, in this embodiment, the feature values include coordinates, color blending values, and timestamps.
[0054] The coordinates of a cell can be represented as (m, n). Here, m is the row number of the cell in the canvas area, and n is the column number of the cell in the canvas area.
[0055] For the color blending value of a cell, the image within each cell is first color-blended to obtain the average color value C for each cell. Then, the average color value C is converted to HSB color mode, and the color blending value of each cell is identified by hue, saturation, and lightness.
[0056] The timestamp is the time when the image was captured. It is achieved by recording the time when the scene area was captured and converting the capture time into the timestamp of the captured image.
[0057] In this embodiment, color timing jitter refers to the degree of color change within the same cell when comparing images captured at time j and time j-1. The calculation process for color timing jitter is as follows:
[0058] The color blending values of cells at the same coordinate in the images captured at time j and time j-1 are transformed as follows:
[0059] C transform =C H *r H+C B *r B +C B *r B
[0060] Among them, C H C B C B These are the decimal values of the H, S, and B channels, respectively, representing the average color value C; r H r B r B These are the transform coefficients for the H, S, and B channels, respectively; C transform This represents the transformed color blending value.
[0061] The color timing jitter of each cell is obtained by calculating the difference between the transformed color blending values of cells at the same coordinate in the images captured at time j and time j-1, and taking the absolute value of the difference.
[0062] If no image is acquired at time j-1, meaning the currently acquired image is the first image in the entire shooting process, then the color timing jitter is 0.
[0063] In the above calculations, environmental changes (such as weather changes, shadow coverage, camera aperture changes, etc.) have the greatest impact on color, with brightness (B channel) being the largest, followed by saturation (S channel), and hue (H channel) the smallest. To reduce these effects, the weights of the B and S channels should be appropriately reduced when calculating color contrast. Therefore, C H C S C B The value is C H >C S >C B When C B When C = 1, S The value range is 2-3, C H The value range is 5-10.
[0064] S105. Obtain the position offset value of each acquired image based on the color timing jitter of the feature value of each cell, and obtain the actual vector value of each acquired image based on the position offset value of each acquired image.
[0065] Specifically, after obtaining the color timing jitter of all cells, the color timing jitter of all cells is compared, the cell with the largest color timing jitter is selected, and the coordinates (m) of the cell with the largest color timing jitter in the canvas area are recorded. i n i The coordinates of the cell with the largest color timing jitter in the canvas area are reduced in dimensionality to obtain the position offset value V of each acquired image.i The calculation formula is as follows:
[0066] V i =(m i 2 +n i 2 ) / (M 2 +N 2 )
[0067] After the classroom artwork is completed, the positional offset values of all captured images are recorded. These records are then sorted by timestamps to obtain the actual vector value for each captured image. This actual vector value describes the order in which the students drew the artwork.
[0068] S106. Compare the actual vector value of each acquired image with the pre-set standard vector value of each acquired image to obtain a comparison result.
[0069] Specifically, in this embodiment, the teacher can first use the correct drawing composition and order to draw, and then, after image acquisition and analysis through the steps described in S101 to S105 of this embodiment, obtain a vector value describing the drawing order. This vector value is the standard vector value. By comparing the student's actual vector value with the standard vector value, the degree of conformity between the student's drawing composition and order and the teacher's can be determined. The comparison of vector values can be measured using Euclidean distance. The larger the comparison result, the greater the deviation, and vice versa. A larger deviation indicates that the student's drawing composition and order are far from the ideal state or standard drawing method, requiring targeted teaching or improvement.
[0070] Additionally, vector approximation retrieval can be used to rank the accuracy of all students' drawings in terms of composition and order. Currently, vector similarity retrieval libraries such as Faiss and Milvus support retrieving vector approximations using Euclidean distance. During the retrieval process, the vector values of all students are stored in the database, and the teacher's vector value is used as the search target. The top k approximations are retrieved, arranged in an ordered manner, with higher similarity ranking higher. The number of top k values can be arbitrarily set.
[0071] Furthermore, during the drawing process, students' hands, upper arms, or brushes may obstruct the drawing board. These obstructions are considered distractions for the drawing board area. Therefore, after subdividing the drawing board area into multiple cells using a grid, the method proposed in this embodiment further includes:
[0072] Perform object recognition on each cell to determine the category of objects within each cell's range;
[0073] Based on the object category, determine whether the object within each cell is an obstruction. If it is an obstruction, mark the cell as invalid; otherwise, mark the cell as valid.
[0074] Furthermore, when a cell is invalid, its color blending value is null, and when the color blending value of a cell is null, its color timing jitter is 0.
[0075] The image acquisition and analysis method for classroom teaching proposed in this embodiment describes and records the composition and drawing sequence of the drawing board area by capturing images of the actual area. These images are then compared with a standard model to analyze students' learning outcomes in composition and drawing sequence. This invention can utilize the cameras already installed in most classrooms for image acquisition, offering excellent implementation conditions and cost advantages. The data volume and computational load of this invention are relatively small, resulting in low cost and ease of promotion.
[0076] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.
Claims
1. A method for image acquisition and analysis during classroom teaching, characterized in that, The method includes: The scene area is photographed at a preset frequency to obtain the captured images; The acquired image is identified, and the canvas area is segmented from the acquired image; The canvas area is subdivided into multiple cells by gridding; Record the feature value for each cell and calculate the color timing jitter of the feature value for each cell; The position offset value of each acquired image is obtained based on the color timing jitter of the feature value of each cell, and the actual vector value is obtained based on the position offset value of each acquired image; The actual vector value is compared with a pre-set standard vector value to obtain a comparison result; The feature values of each cell include coordinates, color blending values, and timestamps. The process of recording the feature values of each cell includes: The coordinates of each cell are identified by the number of rows and columns in the canvas area where each cell is located; Perform color mixing on the image within each cell to obtain the average color value within each cell; The average color value is converted to HSB color mode, and the color mixing value of each cell is identified by hue, saturation, and lightness. Record the shooting time of the scene area and convert the shooting time into a timestamp of the acquired image; The process of calculating the color timing jitter of the feature value of each cell includes: The color blending values of cells at the same coordinate in the images captured at time j and time j-1 are transformed as follows: C transform =C H * r H + C S * r S + C B * r B Among them, C H C S C B These are the decimal values of the H, S, and B channels for the color average value C, respectively. H C S C B The value is C H > C S >C B ;r H r S r B These are the transform coefficients for the H, S, and B channels, respectively; C transform The transformed color blending value; The color mixing values of cells at the same coordinate in the images captured at time j and time j-1 are calculated by difference, and the absolute value of the difference is taken to obtain the color timing jitter of each cell. The process of obtaining the position offset value of each acquired image based on the color timing dithering of the feature value of each cell includes: Compare the color timing jitter of all cells, filter out the cell with the largest color timing jitter, and record the coordinates of the cell with the largest color timing jitter in the canvas area; The coordinates of the cell with the largest color timing jitter in the canvas area are reduced in dimension to obtain the position offset value of each acquired image.
2. The method according to claim 1, characterized in that, The process of identifying the acquired image and segmenting the canvas region from the acquired image includes: The objects in the acquired images are identified and classified; Based on the classification results, select objects that have pre-defined canvas features; Position dimensional boundaries around the object having predefined canvas features; The canvas area is divided according to the stated size boundaries.
3. The method according to claim 1, characterized in that, The process of subdividing the canvas area into multiple cells by gridding includes: The horizontal dimension of the drawing board area is divided into N length units on an average basis. The vertical dimension of the drawing board area is divided into M length units on an average basis. Intersect the dividing lines on the horizontal and vertical dimensions of the canvas area to obtain M×N cells.
4. The method according to claim 1, characterized in that, When j=0, the color timing jitter of the cell is 0.
5. The method according to claim 1, characterized in that, The process of obtaining the actual vector value based on the position offset value of each acquired image includes: The position offset values of each acquired image are sorted according to the timestamp of each acquired image; The actual vector value is obtained based on the sorting result.
6. The method according to claim 5, characterized in that, After subdividing the canvas area into multiple cells using a grid, the method further includes: Perform object recognition on each cell to determine the category of objects within each cell's range; Based on the object category, determine whether the object within the range of each cell is an obstruction. If it is an obstruction, mark the cell as an invalid cell; otherwise, mark the cell as a valid cell.
7. The method according to claim 6, characterized in that, When the cell is invalid, the color blending value of the invalid cell is empty, and when the color blending value of the cell is empty, the color timing jitter of the cell is 0.