Sistem Cerdas Pelacakan Individu Berbasis YOLO dan Deep SORT untuk Pengawasan Keamanan Kampus

Elvi Rahmi (Politeknik Negeri Bengkalis)
Eva Yumami (Politeknik Negeri Bengkalis)
Isna Yulia (Politeknik Negeri Bengkalis)

Jurnal Elektronika dan Teknik Informatika Terapan
Vol. 4 No. 3 (2026): September: Jurnal Elektronika dan Teknik Informatika Terapan ( JENTIK )

Digital transformation in vocational higher education requires the integration of intelligent technology into campus security systems. This study develops an individual tracking system based on CCTV video using YOLOv8 for object detection and Deep SORT for identity tracking. The system detects people, assigns unique IDs, records movement automatically, and visualizes tracking data through a heatmap and an interactive Streamlit dashboard. Testing was conducted using an uploaded video and CCTV footage. The first scenario produced 1,371 detection records from 47 frames and 36 unique tracks, with an average confidence of 0.458. The CCTV scenario produced 715 records from 60 frames and 14 unique tracks, with an average confidence of 0.486. Overall evaluation yielded 97.9% precision, 100% recall, 99.0% F1-score, and 97.9% accuracy. The results show that the integration of YOLOv8 and Deep SORT can support consistent multi-object tracking and informative real-time visualization for data-driven campus security monitoring.