Smartphone based Driver-Assistance System for Urban Environments

In the 21st century the number of cars and smart phones increases dynamically. The range of available services embeddable into a car is also wider and wider, although this does not mean that every customer is ready to pay for them. So a portion of the latest cars is not equipped by lane detection systems and this is valid for older models as well. Why not to look for a solution for car owners having a smart phone? The idea itself is trivial: smart phones (that have cameras and some computational capacity) could be used to provide some safety services for the drivers. The real challenges are in data storing and effective computing, since the application needs to process the incoming frames in real-time. In this paper we present a smartphone based driver-assistance system focusing on its two main components, namely a Probabilistic Hough transform based lane detection and a road segmentation solved by deep neural network.