Automatic Classification of Road Traffic with Fiber Based Sensors in Smart Cities Applications

Low cost monitoring of road traffic can bring a significant contribution to use the smart cities perspective for safety. The possibility of sensing and classifying vehicles and march conditions by means of simple physical sensors may support both real time applications and studies on traffic dynamics, e.g. support and assistance for car crashes and prevention of accidents, and maintenance planning or support to trials in case of litigation. Optical fibers technology is well known for its wide adoption in data transmissions as a commodity component of computer networks: its popularity led to large availability on the market of high quality fiber at affordable price. As a purely physical application, its optical properties may be exploited to monitor in real time mechanical solicitations the fiber undergoes. In this paper we present a novel approach to using optical fibers as road sensors. As quite popular in literature, fiber is used to sense the vibrations caused by vehicles on the road: in our case, signals are processed by functional classification techniques to obtain a higher quality and a larger flexibility for the reuse of results. Classification aims at enabling profiling of road traffic. Moreover in our approach we would like to optimise the analysis and classification computations by splitting the process among edge nodes and cloud nodes according to the available computation capacity. Our solution has been tested by an experimental campaign to show the suitability of the approach.