[Paper Review] Design, Implementation and Simulation of a Cloud Computing System for Enhancing Real-time Video Services by using VANET and Onboard Navigation Systems

Karim Hammoudi, Nabil Ajam+6 more|arXiv (Cornell University)|Nov 25, 2014

Vehicular Ad Hoc Networks (VANETs)7 references3 citations

TL;DR

This paper proposes a novel cloud computing architecture integrating VANETs, onboard vision-based navigation systems, and real-time video processing to enhance dynamic road monitoring. By leveraging edge processing on vehicles, RSU communication, and cloud-based analytics, the system enables real-time detection and tracking of vehicles and individuals for police surveillance, achieving face and license plate recognition in under 5 seconds per image with low-latency data transfer.

ABSTRACT

In this paper, we propose a design for novel and experimental cloud computing systems. The proposed system aims at enhancing computational, communicational and annalistic capabilities of road navigation services by merging several independent technologies, namely vision-based embedded navigation systems, prominent Cloud Computing Systems (CCSs) and Vehicular Ad-hoc NETwork (VANET). This work presents our initial investigations by describing the design of a global generic system. The designed system has been experimented with various scenarios of video-based road services. Moreover, the associated architecture has been implemented on a small-scale simulator of an in-vehicle embedded system. The implemented architecture has been experimented in the case of a simulated road service to aid the police agency. The goal of this service is to recognize and track searched individuals and vehicles in a real-time monitoring system remotely connected to moving cars. The presented work demonstrates the potential of our system for efficiently enhancing and diversifying real-time video services in road environments.

Motivation & Objective

• To design a generic, scalable cloud computing architecture that enhances real-time video services for road environments.
• To integrate onboard vision systems, VANETs, and cloud computing to enable cooperative monitoring and data sharing among vehicles.
• To experimentally validate the system's capability in real-time detection and tracking of vehicles and individuals for law enforcement applications.
• To demonstrate the feasibility of using embedded systems and cloud platforms for dynamic, georeferenced video analytics in mobile road environments.
• To develop a simulator for testing the system's performance before real-world deployment on moving vehicles.

Proposed method

• The system uses onboard cameras in vehicles to capture real-time video streams, which are processed locally using Emgu CV for face and license plate extraction.
• Extracted features (faces, license plates) are transmitted via VANET to Roadside Units (RSUs) and then forwarded to cloud computing nodes for further analysis.
• A cloud-based processing pipeline matches extracted features against a reference database to identify searched individuals or stolen vehicles.
• The system employs a hybrid architecture with edge processing on Raspberry Pi devices, RSU relays, and centralized cloud workstations for data correlation and visualization.
• Data is visualized in real time on a Google Maps-based interface, with timestamps and geolocation metadata for tracking.
• The system was tested using a simulated in-vehicle embedded system and a small-scale VANET simulator, with performance measured in data transfer and processing times.

Experimental results

Research questions

• RQ1 Can a cloud-based system effectively integrate VANETs, onboard vision systems, and real-time video analytics to enhance road monitoring services?
• RQ2 How efficiently can real-time face and license plate recognition be performed in a distributed vehicular-cloud architecture?
• RQ3 What are the latency and throughput characteristics of transferring and processing video data across a VANET-to-cloud pipeline in a simulated road environment?
• RQ4 To what extent can cooperative monitoring using mobile vehicles improve the detection and tracking of searched individuals or vehicles?
• RQ5 How scalable and reliable is the proposed architecture for deployment in real-world mobile scenarios involving multiple moving nodes?

Key findings

• The system achieved face extraction in 1.08 seconds and license plate extraction in 3.29 seconds per image on a 2.4GHz Intel Core i5 workstation.
• Data transfer from Raspberry Pi to RSU took 1.33 seconds, and from RSU to cloud nodes took 1.12 seconds for a 16.5KB image (640x480 resolution).
• The entire pipeline, including image transfer and processing, was completed in under 5 seconds per image, enabling quasi real-time monitoring.
• The system successfully demonstrated the feasibility of using mobile vehicles as distributed sensing nodes for police surveillance applications.
• The architecture was validated through simulation and prototype implementation, showing potential for deployment in real mobile environments.
• The system supports dynamic Point of Interest (POI) detection, such as available parking spots and hazardous road conditions, and enables information sharing among vehicles and agencies.