Inspiration
The project was inspired by recent security failures such as the Delhi and Mumbai blast incidents and the Coimbatore case involving Rethanya, where suspicious vehicle behavior went unnoticed until it was too late. These events showed that human monitoring alone cannot reliably detect early-stage threats, motivating us to build an automated system that never gets tired.
What it does
The system detects vehicles in real time, tracks how long they remain occupied, and triggers alerts when a vehicle loiters beyond a safe threshold. It highlights threats, sounds an alarm, captures snapshots, and logs license plates for investigation.
How we built it
We used YOLOv8 for vehicle detection, ByteTrack for tracking, OpenCV for frame processing, EasyOCR for license plate reading, MongoDB for log storage, and Streamlit for the dashboard. Together, they convert CCTV footage into proactive threat detection.
Challenges we ran into
We faced issues with tracking stability, OCR accuracy under poor lighting, maintaining real-time speed, and preventing false alerts from short stops. Fine-tuning thresholds and improving logic helped overcome these challenges.
Accomplishments that we're proud of
We built a full end-to-end system that automatically detects loitering vehicles, captures evidence, and alerts operators instantly. The solution is practical, reliable, and addresses a real security gap.
What we learned
We learned how to build real-time computer-vision pipelines, optimize models for live video, handle noisy OCR inputs, and integrate AI with databases and a user dashboard.
What's next for AI Vehicle Loitering Detector
We plan to add thermal camera support, cloud-based alert sharing, and autonomous drone deployment to track suspicious vehicles even after they leave the monitored area.
