A VisualSfM based Rapid 3-D Modeling Framework using Swarm of UAVs

Abstract

This paper presents an unmanned aerial vehicle (UAV) swarm based rapid 3D modeling framework for objects of interest in unknown environments using Visual Structure from Motion (VisualSfM). The presented technology framework is driven by the goal to provide rapid and accurate situational awareness for mission planning in applications such as disaster control/recovery, search and rescue, industrial infrastructure monitoring, large inventory accounting, military asset movement and so on. Our approach to deliver the necessary and sufficient amount of information using a hierarchical swarm of UAVs. Processed information from UAV onboard sensors is synthesized into easy-to-interpret three dimensional (3D) contents for particular areas of interest. This paper presents the preliminary proof of concept experiments for the presented 3D modeling framework, tested with a heterogeneous swarm of UAVs of varying capabilities. The goal of the presented research and development effort is to deliver a rapid and enhanced surveillance/reconnaissance capability with selective 3D detailing and classification of only the areas of interest in the global map and not the entire map, thereby limiting the computational cost and cognitive load on human agents in the team.