A student classroom behavior recognition method, device, equipment and storage medium

By combining Intersection over Union (IoU-Tracker) and spatiotemporal graph convolutional neural networks, the problems of identity jumps and tracking loss in classroom behavior recognition are solved, achieving low-overhead and high-accuracy behavior recognition.

CN122200804APending Publication Date: 2026-06-12GUANGZHOU SHIYUAN ELECTRONICS CO LTD +1
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
GUANGZHOU SHIYUAN ELECTRONICS CO LTD
Filing Date
2021-09-16
Publication Date
2026-06-12

AI Technical Summary

Technical Problem

Existing classroom behavior recognition methods are prone to identity changes or tracking loss in situations involving occlusion, rapid posture changes, and multiple people interacting. They also have high computational costs, affecting the accuracy of behavior recognition.

Method used

The Intersection over Union (IoU-Tracker) is used for target tracking. By calculating the IoU of the detection boxes, the same students in multiple consecutive classroom images are identified, avoiding complex optical flow calculations or depth feature extraction. The spatiotemporal graph convolutional neural network is combined to analyze student behavior.

Benefits of technology

It reduces computational overhead, improves the accuracy of behavior recognition, and can suppress identity jumps and tracking loss in complex classroom scenarios with multiple intersecting and occluded targets, providing continuous and accurate student key point sequences.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122200804A_ABST
    Figure CN122200804A_ABST
Patent Text Reader

Abstract

The application discloses a student classroom behavior recognition method and device, equipment and storage medium, through calculating the intersection and union ratio of the detection frame in one frame of classroom image and all detection frames in the adjacent frame of classroom image, the same student in continuous multiple frames of classroom image is determined based on the intersection and union ratio, it is ensured that the detection frame of the same student in different frames can be correctly matched, without complex optical flow calculation or depth feature extraction, the calculation overhead is reduced, the real-time processing demand of the classroom is met, in addition, the same student in continuous multiple frames of classroom image is determined based on the intersection and union ratio, in the complex classroom scene of multiple target interlacing and mutual shielding, the occurrence of identity jump or tracking loss can be effectively inhibited, a continuous and accurate student key point sequence is provided for subsequent behavior analysis, so that the accuracy of the overall behavior recognition is improved.
Need to check novelty before this filing date? Find Prior Art