Mobile computer vision solution to monitor public space infrastructures.
Cities worldwide have become complex products that compete for a share of the growing population choosing urban living. According to the UN, the world population living in urban areas will increase from 55% in 2018 to 68% in 2050 and is expected to reach 84% in 2100.
Cities are no longer mere agglomerations of the population where public authorities have little to worry about beyond offering basic conditions for life in society. To remain attractive, cities have to embark on a quest to continuously create conditions that increase the well-being of citizens through new services and infrastructures that improve education, mobility, security, leisure, welfare, spatial planning, and utility delivery. These projects are almost always well received by the population and fulfill their function of increasing the quality of life in the city, but they also tremendously increase the number of infrastructures and equipment that needs to be managed, monitored, and maintained every day, a task that requires resources beyond the capacity of public authorities. The only way to keep the situation under control is through active collaboration networks between public and private entities, aided by technological solutions like the one proposed here, that combine data sharing, intelligent analysis, and process automation to ease the entire operation of keeping the modern city working around the clock.
The Tidy City experiment proposes to equip vehicles with mobile devices to take pictures and later process them to automatically detect and classify problems in publicly visible infrastructures. To reduce pollution and CO2 emissions, street-level images will be collected by a wide range of vehicles that already travel the city streets daily, such as waste collection trucks and express delivery vehicles. The collected images are sent to a server for analysis and classification based on AI models. Within the scope of this experiment, the system will be able to detect and record the GPS location of damages in outdoor advertising, lightning and electricity supply infrastructures, and improper waste disposal situations (e.g., waste outside containers, tires, old appliances, and furniture). The detected occurrences data will be shared with each corresponding partner for integration and operations optimization. To comply with the RGPD and other European and National legal requirements, all personally identifiable information (e.g., people’s faces and license plates) will be removed from the images through the same privacy by-design processes already used by Google Street View.
Business Projections – Scalability
KPIs at the end of the experiment:
- Number of occurrences detected: 100 (expected to grow 250% one year after)
- Number of data consumers: 4 (expected to grow 300% one year after)
- Recurrence rate after 2 days: 50% (expected to decrease by 20% one year after)