Effective visual surveillance of human crowds using cooperative unmanned vehicles

摘要

The goal of this work is to propose an effective and efficient visual surveillance system for detection, geolocalization, and data association of moving human crowds using teams of unmanned aerial vehicles (UAVs) and unmanned ground vehicles (UGVs) in a border patrol application. Such complex system suffers from various emerging challenges such as: heterogeneous dynamic data, non-rigid target shapes, dynamic background due to moving sensors, and occlusion. Therefore, different fidelity levels have been considered in this work for UAVs and UGVs based on their different characteristics, and a number of related computer vision algorithms have been proposed based on the dynamic data driven application system (DDDAS) paradigm. Moreover, a testbed involving real hardware (UAVs, UGVs, and cameras) and an agent-based simulation model is developed to verify, validate, and demonstrate the system. The experimental results reveal the effectiveness of the proposed approaches for visual surveillance of human crowds by unmanned vehicles.