FCC Medio Ambiente has incorporated a new project into its street cleansing service in Madrid that will enable inspection vehicles to carry out visual recognition work using artificial intelligence (AI) to detect waste that left beside containers. This pioneering initiative in Spain has been developed by the Information Technology department and the Madrid branch of FCC Medio Ambiente, in charge of the street cleansing contract for lot 6 in the capital, in collaboration with a company specialising in AI, Advisory Experts, and this coming month it will activate the project in Carabanchel, Villaverde and Usera, districts in which an average of 2,100 abandoned waste collection occurences are recorded every day.

To mark the start-up of the project, last Monday FCC Medio Ambiente held a brief event attended by a municipal committee headed by Borja Carabante, Madrid City Council's head of Urban Planning, Environment and Mobility, together with José Antonio Martínez, the councillor responsible for Cleaning and Ground Areas, and Carlos Izquierdo, the councillor for Carabanchel.

With a data feed device through artificial vision, and through the use of visual sensors and the development of algorithms based on AI and Machine Learning, the solution developed allows the proactive and automatic detection of waste abandoned next to containers. The project seeks to eradicate a problem that greatly affects the quality of the service, causing sources of dirt, significant aesthetic problems in the urban landscape and numerous complaints from citizens.

With the implementation of this project, the service reduces the average response time for this type of incidents by more than 50%. It also contributes to reducing the energy consumption of vehicles, which increases efficiency and, consequently, the useful life of the equipment. The activation of this innovative project was possible after intense training and calibration of sensors and the RGB camera to achieve high precision in detecting waste, during which data analysis, algorithm training and parameter adjustment were key to attaining the best possible results. This training was completed with recording sessions of the vehicle in a real environment in the streets of Madrid so that the system could adapt to different scenarios and lighting and weather conditions. In order to meet the requirements of data protection legislation, the system also pixelates people and vehicles that may appear in the images.

Incidents are processed immediately, as all the information arrives in real time at the Service Monitoring Centre of lot 6, set up by the Madrid branch of FCC Medio Ambiente after the award of the last street cleansing contract in November 2021 to centralise the comprehensive management of the service and carry out the control, analysis and transmission of information generated. This immediate communication speeds up the assignment of the incident to the nearest vehicle or the most suitable team to collect each type of waste.

The system classifies the information based on the identification of the anomalies detected around the containers, communicates each case to the data centre and generates a record in the digital platform that includes the type of incident, street and number, GPS coordinates and a photograph. This information, in addition to the resolution of the incident, is used to organise all records in a dashboard that shows the locations, neighbourhoods, as well as days or periods with the highest volume of anomalies.

Picture by Abraham Blázquez