Drones for Precision Scanning of Large Objects and Route Planning in Enclosed Environments

Unmanned aerial vehicles are increasingly integrated into a wide range of applications due to their mobility and ability to reach hard-to-access areas such as inspections of industrial chimneys, stone crushers or bridges. Therefore, a navigational algorithm is needed for the increasingly complex environments the UAV may be introduced to. An example of this is an enclosed environment, which presents both physical and visual challenges for UAV navigation. Addressing these challenges is necessary to enable UAVs to scan large industrial structures autonomously in an enclosed environment, replacing the complex setups required today. For this purpose, a lightweight, compact, and cost-effective vision-based solution is proposed for UAV navigation in such environments. The system consists of a Holybro QAV250 UAV, equipped with an Intel RealSense D435i stereo camera. It uses a Pixhawk 6C Mini as the flight controller and a Raspberry Pi 5 as the companion computer for onboard decision-making. The system is built around two core pipelines: a mapping component and a navigation component, enabling fully autonomous operation. The mapping component uses ORB-SLAM3 to localize the UAV's position and orientation. For environmental mapping, a dense point cloud is generated from the stereo camera depth frame and processed using an octree representation to build a 3D map of its surroundings. The navigation component utilizes a hybrid approach. It combines traditional frontier-based exploration with a fast frontier method that prioritizes visible, reachable frontiers to minimize velocity changes. When no suitable frontier is available within the UAV's field of view, the system falls back to using the RRT* algorithm to compute a safe and optimal path to the next exploration target. The system was validated using Gazebo simulations of an enclosed environment, which demonstrated that the proposed pipeline can safely traverse the space while successfully reconstructing and navigating through previously unknown areas.