Parking guidance systems have the potential to reduce the congestion in crowded areas by providing real-time indications of occupancy of parking spaces. Such systems are mostly implemented in indoor environments using costly sensor-based techniques. Consequently, there is an increasing demand for such systems in outdoor environments as well. Finding inexpensive methods to detect unoccupied parking spaces has been thoroughly researched.
We use classical computer vision methods to detect vacant parking spots from surveillance footages. Our database is then updated and fee the data to our react native application.
Using YOLO v3 to learn bounding boxes of parking locations, we then manually feed these coordinates to a classical computer vision methods based on corner detection and BIRCH clustering to classify whether each location busy/free. We also use homography transformation to account for minor camera movement.