Outcome and proof
The working prototype converts one overhead camera feed into a live view of occupied and available parking spaces. It avoids installing and maintaining a sensor in every bay.
Context and problem
Per-space sensors add hardware and maintenance cost across an entire parking facility. A camera-based approach can cover multiple bays with one existing visual source.
My role and responsibility
I built the detection and occupancy pipeline that maps vehicle detections to defined parking-space regions and produces the availability state.
Solution and architecture
- Capture — receive the overhead camera feed.
- Detect — locate vehicles in each frame with YOLO.
- Map — compare detections with predefined parking-space regions.
- Resolve — expose each bay as occupied or available.
Key engineering decisions
Treat occupancy as a spatial problem
Vehicle detection alone does not identify which bay is occupied. The system maps detections into the geometry of the parking layout.
Account for camera perspective
Near and distant bays have different apparent sizes. The occupancy logic operates on per-bay regions rather than assuming one global visual threshold.
State the evidence boundary
The prototype demonstrates the real-time workflow. Accuracy, camera coverage, and processing-rate figures have not been published, so they are not presented as results.
Verified results
- Working prototype
- Single overhead camera input
- Real-time occupied / available output
Visual evidence
The diagram above shows the camera-to-occupancy path without presenting an unrelated stock image as project evidence.
Technology stack
YOLO, OpenCV, and Python.