A quick summary of Smartcitysafety

Smart City Safety

Given the critical importance of security in cities, innovations in the technology are increasingly improving the safety of city inhabitants. New services such as automated detection of suspicious activities allows a quicker response to threats. 5G and Network Function Virtualization enable solutions that help protect city inhabitants in different environments. Video surveillance is an example that combined with artificial intelligence (or machine learning), will help to lower crime rates and to improve life conditions and improve people’s safety. This demo presents a smart city safety system where the combination of Mobile edge Computing (MEC), Network Function Virtualization (VNF) and Machine Learning technologies enable identify suspicious activities in the city. As a blueprint demonstration face detection, face recognition and video transcoder are deployed as virtual functions reducing response delay. In addition, a virtual function that enables speech recognition has been deployed to detect suspicious conversations.

System Architecture

Figure 1 shows the smart city safety architecture deployed at Bristol testbed.


The architecture is composed of four building-blocks.

1. 5GinFIRE Ecosystem: is composed of a web portal2 and an orchestrator module. In 5GinFIRE web portal an experimenter can create or instantiate a VNF and submit its experiment using a Network Service Descriptor (NSD) that describes the network service to be created. The orchestrator module includes the OpenSource MANO Release Two [24] that enables multi-site NFV management and orchestration. Currently, there are three facilities interconnected based on OpenStack3 located in United Kingdom (UK), Spain and Portugal. The system described in this paper was deployed at UNIVBRIS UK facility.

2. University of Bristol 5GinFIRE Facility: this building block contains the NFVI based on OpenStack Pike. The NFVI is composed of two compute nodes (PowerEdge T630, 800GB, 128GB RDIMM, Intel Xeon E5-2680) and one controller node (PowerEdge T630, 480GB, 128GB RDIMM, Intel Xeon E5-2680). This block holds the VNF being used by the smart city safety demo for video transcoder, face detection and face recognition.

3. Smart City Safety Demo: this block consists of four basic elements: a 360 camera model Ricoh Tetha V; a Raspberry PI (Raspi) 3 Model B, running Raspbian Jessi with a 28000mA battery, a microphone SAMSON SAGOMICARR USB Mic, and a bike helmet. The Raspi, the MIC and the 360 camera are attached on the bike helmet and they communicate with each other via WiFi 2.4GHz. The Raspi communicates with the NFVI via WiFi Access Point (AP) 5GHz.

4. 5GinFIRE Multi-site: this block is part of the 5GinFIRE ecosystem that includes multi-site orchestration. The facilities are located in Universidad Carlos III de Madrid (UC3M), Instituto de Telecomunicacao de Aveiro (ITAv) and University of Bristol (UNIVBRIS), 5G-VINO (Greece), WINS-5G (Dublin), eHealth5G (Poland).

VNF Description

The Smart City Safety uses two VNFs:

1. VNF Video transcoder

This VNF enables three functionalities:
a- Receive 360 live streaming video and codify for a rectagular format
b- Execute face detection given the live streaming video in a rectagular format
c- Execute face recognition

The VNF video transcoder is based on OpenCV and it is publicly available at 5GinFIRE VxF repository

The configuration to use the VNF is available here. Read Me Vnf Video

2. VNF Audio transcoder

This VNF enables Speech recognition based on Google Cloud Speech API and Sphinx. It allows to convert audio to text and identify some suspicious words. It also available at 5GinFIRE VxF repository.

The configuration to use the VNF is available here.Read Me Vnf Audio