Related Experiment Video
Updated: Apr 27, 2026
04:13
Using a Real-Time Locating System to Measure Walking Activity Associated with Wandering Behaviors Among Institutionalized Older Adults
Published on: February 8, 2019
7.2K
Enhancements and On-Site Experimental Study on Fall Detection Algorithm for Students in Campus Staircase
Ying Lu1, Yuze Cui1, Liang Yan2
1College of Resources and Environmental Engineering, Wuhan University of Science and Technology, Wuhan 430081, China.
Sensors (Basel, Switzerland)
|December 11, 2025
Summary
This study introduces an improved YOLOv7 fall detection model for campus stairwells, achieving high accuracy and a lightweight design. On-site experiments confirm its practical feasibility for preventing accidents in crowded environments.
Area of Science:
- Computer Vision
- Artificial Intelligence
- Public Safety
Background:
- Campus stairwells pose significant fall and stampede risks due to high foot traffic.
- Existing fall detection models lack a balance between precision, lightweight design, and practical validation.
- Accurate and efficient fall detection is critical for enhancing safety in crowded public spaces.
Purpose of the Study:
- To develop and validate an enhanced fall recognition model for campus stairwells.
- To address the limitations of existing models in terms of precision, computational cost, and real-world applicability.
- To improve safety by enabling timely detection of falls in dynamic, crowded environments.
Main Methods:
- Developed an enhanced fall recognition model based on YOLOv7, incorporating DO-DConv and Slim-Neck modules.
- Created a specialized dataset of campus stairwell falls with diverse personnel behaviors.
- Conducted numerical comparison experiments and on-site field validation with varying population densities and lighting.
Main Results:
- The enhanced YOLOv7 model achieved 88.1% mean average precision (mAP), outperforming traditional YOLOv7 by 2.41%.
- Model complexity was significantly reduced (GFLOPs from 105.2 to 38.2), and training time decreased by 4 hours.
- Preliminary on-site experiments confirmed acceptable accuracy and analyzed detection confidence under different conditions.
Conclusions:
- The proposed YOLOv7-based model offers a precise, lightweight, and practically validated solution for campus stairwell fall detection.
- The findings provide valuable insights for optimizing fall detection systems in real-world, dynamic environments.
- This research contributes to enhanced public safety through advanced AI-driven surveillance and risk mitigation.
