Deep Learning Based On-Street Parking Spot Detection for Smart Cities

Abstract

Parking is a big problem especially in populated cities. It is sometimes very difficult to find free on-street parking spots. Drivers have to do a blind search to find a free spot. Blind searching is not only fuel and time consuming, but also causes traffic congestion. Indoor parking garages have sensors or light systems to indicate free spots. However, an indoor approach cannot be used for the on-street parking problem. To solve this, an image processing system is developed, and available parking spots nearby are detected using deep convolutional neural networks. With the help of roadside cameras, the system is first trained to detect free street parking spots. After the training phase, a frame taken from a roadside camera is analyzed whether there is a free spot or not. A mobile application is also developed which takes the request and triggers the corresponding roadside camera, and then notifies the driver about the available parking spots around the region.