Parking slot occupancy detection

by Bharti Patil

This page contains information about a final project from the Embedded and Distributed AI course of the Spring 2020 semesterlänk till annan webbplats, öppnas i nytt fönster. The focus has been developing real-time intelligent algorithms which can run on embedded systems.

Finding parking space for your vehicle is a major problem in big cities. The rise of car ownership has created an imbalance between parking space demand and supply. In the current situation, a parking slot management system that can track parking spots has become a necessity for all major cities.

Mask-RCNN is used on every frame and it will return a dictionary that contains the bounding box coordinates, masked value of detected objects, confidence values for each prediction so far, and various class labels of detected objects. Now using the labels we will filter out the bounding boxes of the labels cars, humans, dogs. Then we use these boxes in the evaluation to compute the Intersection over Union(IOU).

MASK RCNN(Image segmentation algorithm) is trained for detecting cars as below:

View of the algorithm detecting cars on the road using different colours

IOU gives us a measure of how much a car’s bounding box is overlapping a parking space bounding box. Using this information, we can easily work out if a car is in a parking slot or not. Putting all this together gives us the result shown in the image below.

A car driving off from a parking space


For more information, you can contact to Bharti Patil: