5g-Cage: A Context And Situational Awareness System For City Public Safety With Video Processing At A Virtualized Ecosystem

In this article we present 5G-CAGE, an ongoing project aimed to deploy a city safety solution that enables monitoring and analytics of video streams collected from distributed sources of a Smart City. Unlike current proposals based on inflexible architectures or limited networks, 5G-CAGE leverages 5G's high throughput and low latency, as well as its enhanced dynamism and adaptability with advanced virtualization-based technologies. In this context, 5G-CAGE defines a virtualization-enabled solution called City Object Detection (CODet), which allows recognizing interest objects in safety related situations, such as vehicles (e.g. license plates or brands), obstacles in emergency settings, or human faces recognition, to name a few. It can process multiple streams collected from fixed and moving cameras used as a distributed visual sensing system, adequately combining image processing and computer vision algorithms in a virtualized ecosystem. This paper presents initial tests in the specific task of locating and recognizing vehicle license plates, where the CODet virtualized solution has been successfully integrated and tested in the 5G-INFIRE platform, an EU-funded project which provides a playground wherein new components, architectures, and APIs may be tried and proposed before being ported to 5G networks.