All projects Next projectLuna — AI-Powered Humanoid Robot
Computer Vision
AI Video Intelligence for Campus Safety
Turning raw CCTV footage into structured, reviewable vehicle events.
The problem
Campus security teams record hours of CCTV footage, but answering simple questions — which vehicles entered, when, and whether riders wore helmets — still means scrubbing video by hand.
The approach
A computer vision pipeline that detects and tracks vehicles across frames, logs entry and exit events against virtual boundaries, runs a number-plate recognition workflow, and routes low-confidence detections to a human for review instead of guessing.
Architecture
System architecture
Stage 1 / 6
CCTV ingest
Recorded or live camera streams are read frame-by-frame with OpenCV.
OpenCV
Key features
- Vehicle detection and multi-object tracking
- Directional entry and exit monitoring
- Number-plate recognition workflow
- Helmet detection concepts
- Structured event logging
- Human review of uncertain detections