UAV-Based IoT Platform: A Crowd Surveillance Use Case

TLDR

Unmanned aerial vehicles are increasingly popular among hobbyists and service providers, and advances in LTE/5G and mobile edge computing are expanding their application scenarios. The article explores the potential of UAVs equipped with IoT devices to deliver IoT services from high altitudes. The authors present a UAV‑based integrative IoT platform with a system orchestrator, demonstrate its use for crowd surveillance via face recognition, and evaluate offloading video processing to a mobile edge computing node versus onboard processing using a testbed and Local Binary Pattern Histogram for recognition. Results show that offloading video processing to a MEC node conserves UAV energy, shortens face‑recognition time, and enables rapid detection of suspicious individuals.

Abstract

Unmanned aerial vehicles are gaining a lot of popularity among an ever growing community of amateurs as well as service providers. Emerging technologies, such as LTE 4G/5G networks and mobile edge computing, will widen the use case scenarios of UAVs. In this article, we discuss the potential of UAVs, equipped with IoT devices, in delivering IoT services from great heights. A high-level view of a UAV-based integrative IoT platform for the delivery of IoT services from large height, along with the overall system orchestrator, is presented in this article. As an envisioned use case of the platform, the article demonstrates how UAVs can be used for crowd surveillance based on face recognition. To evaluate the use case, we study the offloading of video data processing to a MEC node compared to the local processing of video data onboard UAVs. For this, we developed a testbed consisting of a local processing node and one MEC node. To perform face recognition, the Local Binary Pattern Histogram method from the Open Source Computer Vision is used. The obtained results demonstrate the efficiency of the MEC-based offloading approach in saving the scarce energy of UAVs, reducing the processing time of recognition, and promptly detecting suspicious persons.