Application of Unmanned Aerial Vehicles in Smart Cities using Computer Vision Techniques

Abstract

Unmanned aerial vehicle (UAV) systems are being widely used as mobile edge devices to provide mobile edge computing services to the ground terminals. This work presents a typical application of intersection monitoring in smart cities through UAVs equipped with cameras. A deep visual tracking system is developed by utilizing off-the-shelf YOLOv3 along with the Discriminative correlation filter for road user detection and tracking respectively. To deal with the camera movement, which naturally happens in UVA systems, an optical flow method is used to boost the tracking system. The optical method collects additional motion cues to help the tracking system for camera motion compensation as well as vehicles speed measurements. The experimental results show the success of the system for tracking vehicles through UAVs and providing critical traffic measurements for smart cities such as vehicle flow, headway, speed profile and online speed-based activity analysis.