.png){kind=icon}
.png){kind=icon}
Return
Multimodal Image Processing and Learning Behavior Pattern Visualization for Educational Management
.png){kind=icon}
OA
AI
DOI：10.18280/ts.420522.png)
.png){kind=icon}
.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The development of smart education has created an urgent demand for fine-grained and intelligent management of classroom teaching. Traditional educational evaluation methods are highly subjective and lack objective quantification. Leveraging computer vision techniques to analyze classroom image data provides a new solution for contactless and accurate assessment of learning behaviors. Multimodal image data, with its complementary strengths in capturing appearance, spatial, and physiological information, lays a solid foundation for comprehensively interpreting classroom behavior patterns. However, existing studies are often limited to unimodal analysis, which is vulnerable to environmental interference, or, when employing multimodal data, rely on simplistic fusion strategies that fail to fully exploit the deep complementarity among modalities. Moreover, the interpretability and visualization of analysis results remain insufficient, hindering their practical application in educational management. To address these challenges, this paper investigates multimodal image processing and learning behavior pattern visualization methods tailored for educational management. The main contributions are as follows: (1) a novel multimodal feature integration model is proposed, employing an encoder-decoder architecture that incorporates tri-modal feature fusion, adjacent-layer feature enhancement, and multi-level cascaded feature integration, aiming to generate high-quality saliency maps for robust representation of learning behaviors; (2) a visualization framework for learning behavior patterns is developed, transforming model outputs into intuitive forms such as heatmaps and behavioral trajectories to support educational management decision-making. The key innovations of this study lie in the following: the design of a hierarchical and guided multimodal feature integration model tailored to educational management scenarios, enabling deep complementarity and enhancement across multiple information sources; the development of a visualization paradigm closely coupled with the feature integration model, significantly improving the interpretability and usability of the analysis results; and the deep integration of advanced computer vision technologies with the specific needs of educational management, providing an end-to-end solution from algorithm to application for precise supervision in smart classrooms.
Keywords:
educational management
multimodal images
feature integration
learning behavior patterns
visualization
saliency detection
AI Summary
Key information extracted from the uploaded paper, including a brief overview, abstract, background, key highlights, visual analysis, and future outlook.
Journal
T
IF:
1
Papers:
102
Citations:
1.2K