The idea for an innovative, low-cost traffic monitoring sensor started around 2012. Basic research, technology formulation, applied research and small-scale prototype were carried out within the project ORUSSI - Optimal Road sUrveillance System based on Scalable vIdeo. The main outcome was the first prototype of the traffic monitoring software (VehicleCounter), which was validated in a laboratory environment. Starting in 2014, the sensor was prototyped using the Raspberry PI computing device, and was validated in a relevant environment with installation in the Metropolitan Area of Florence. This was achieved also with support from project CHEST. Within project SOUL-FI FLOUD, the entire platform has been completed and fully demonstrated in operational environment, reaching a sufficient readiness to be marketed. SME Instrument, in 2015, assessed the economic viability of TrafficFlow and ended with a consolidated business plan. Currently, Magenta's participation to Organicity project is helping the technology to spread beyond the Italian border.
The core of our sensing technology is VehicleCounter, a traffic monitoring application initially developed for the Axis Camera Application Platform (ACAP), then ported to the Raspberry Pi computing device. The purpose of VehicleCounter is to collect meaningful statistics regarding number, speed and length of vehicles traveling the observed lanes, and help analyzing how the traffic is shaped, supporting stakeholders and administrations in their decision-making process.
The application allows the user to set a number of virtual sensors by directly drawing them on the images grabbed by a video sensor. Collected data are stored in a local DB for a limited amount of time, allowing them to be queried either manually or by other pieces of software. This allows for an easy integration with third-parties platforms: VehicleCounter accepts simple HTTP requests and responds with JSON documents that can be easily parsed.
VehicleCounter is an Intelligent Video Analysis application: it uses Computer Vision (CV) algorithms to perform its job. Most CV applications are very difficult to setup because there are a lot of parameters to tune, a lot of variables to keep into account, and a lot of problems to deal with. Traffic monitoring applications, particularly, often requires to have a very specific point of view: a zenithal view, that is, with the camera placed above the road to monitor, pointing down. This is undoubtedly the best point of view for car counting, as it reduces occlusions between vehicles and perspective in the scene to the minimum. But this point of view is not always achievable, for various reasons. That is why we made special efforts to give VehicleCounter the capability to work from sub-optimal points of view.
As the graphics shows, accuracy degrades gently as perspective in the scene increases. This allows you to achieve effective vehicle counting even from your apartment's or office's window.
FLOUD is a web platform for road-traffic data collection and analysis. It is the ideal server side companion for our TrafficFlow sensors and for VehicleCounter application in general. It provides the following main functionalities: