Integration of DeepSORT and RFID Technology for Enhanced Human Tracking
Human activity tracking enhances safety and reduces the risk of people getting lost or kidnapped. This paper presents a human tracking system using the YOLOv8n detection model, DeepSORT tracking algorithm, and RFID technologies on low-power devices like the Raspberry Pi. The passive RFID tags, wh...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Northern Technical University
2024-12-01
|
| Series: | NTU Journal of Engineering and Technology |
| Subjects: | |
| Online Access: | https://journals.ntu.edu.iq/index.php/NTU-JET/article/view/1095 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Human activity tracking enhances safety and reduces the risk of people getting lost or kidnapped. This paper presents a human tracking system using the YOLOv8n detection model, DeepSORT tracking algorithm, and RFID technologies on low-power devices like the Raspberry Pi.
The passive RFID tags, which do not require batteries, make the system lightweight and maintenance-free. The Raspberry Pi Model V3 camera, with an 8-megapixel Sony IMX219 sensor, captures video at 640x480p90 resolution.
The YOLOv8n algorithm was trained on 2292 images, achieving an accuracy of 0.992 for mAP50 and 0.902 for mAP50-95. After integrating it with DeepSORT, the system achieved MOTA = 0.973684 and MOTP = 0.438766 at 30 fps.
In real time, tracking for 20 frames yielded MOTA = 1.0 and MOTP = 0.13. The UHF RFID reader detected tags at a distance of 1.5 meters.
|
|---|---|
| ISSN: | 2788-9971 2788-998X |