A Smart Parking System: An IoT Based Computer Vision Approach for Free Parking Spot Detection Using Faster R-CNN with YOLOv3 Method

Nowadays, parking is much costlier and time consuming process in almost every big city, all over the world. The issue is that, the user couldn't find available parking space at time, and it found that cars may seek very limited parking space which causes severe traffic congestion. Thus, Smart parking systems are necessary to find the near term parking on demand. In this work, Smart Parking System is proposed to assign a free parking place for people, who need parking lot. This system will be able to process the images of parking area and its free slots in real-time, to notify user about all free slots which are available for parking. The user can choose the available parking lot as per their needs. This system will store the activity logs for further analysis to determine parking trends on different days. The parking lot detection is implemented using Faster Recurrent Convolutional Neural Networks (Faster R-CNN) with YOLOv3 technique. This scheme trains a model with car image dataset, which will support the system to recognize car in parking lot. This approach is very constructive, as it won't confuse with other temporary objects in the parking lot, the proposed system only spot parked cars in the parking lot. This system is more robust, energy-efficient and has the competence for further improvements to be done in it.