An automated visual recognition system to counteract illegal dumping in smart cities

In this paper we will describe the prototype form of an automated visual recognition system designed to mitigate illegal dumping, as the outcome of an experiential learning activity. The presented solution relies on the sensor networks of a smart city where the waste management system is supposed to be integrated with other municipality services for environment control and management. In particular, we want to take advantage of the pictures, frames and videos continuously recorded by cameras installed in the cities for traffic monitoring, for surveillance or any other reason. Such data are processed by means of cognitive computing techniques and a specific algorithm of image analysis has been trained to identify trash, especially bulky waste, where it should not be, and trigger an alarm to the municipality. Besides, an organization plan is also proposed for intelligent waste collection as well as some organizational ideas for scalability. The learning activity has been conducted within the program “Party Cloud Challenge per Genova” promoted by IBM in collaboration with the city municipality of Genoa, Italy.