Dual-Camera Drone Pipeline Enables Safe Micro-Inspections from Afar

A team of researchers from multiple institutions has released a modular dual-camera pipeline called aerial_micro_inspection that lets aerial robots capture microscopic details without flying dangerously close to the target. Most existing inspection drones either need to approach within centimeters of the surface—risking collisions—or require complex flight paths and prior knowledge of the target's geometry, position, and orientation. This new pipeline, built on a PX4-powered drone, solves both problems by combining a zoomed, gimbal-mounted inspection camera with a wide-field-of-view stereo navigation camera. The stereo camera acquires the target surface on site, estimates its range in real time, and automatically partitions the surface into smaller inspection regions. While the inspection camera visits each partition, a vision-based feedback loop actively compensates for drone motion and localization inaccuracies, keeping the target in frame.

The pipeline was evaluated in simulation and real-world experiments across two distinct use cases: tree inspection to detect oak processionary caterpillars and their eggs, and greenhouse inspection of sticky traps for whiteflies. Results demonstrated improved coverage robustness under simulated drone disturbances, along with effective detection of caterpillars and eggs and high-detail imaging of insects in the field. The entire system is open-source and built on ROS 2, making it easy to adapt for new applications by replacing only the surface-segmentation and micro-target detection modules. The code is publicly available, and the modular architecture allows researchers or commercial teams to quickly retool the pipeline for inspecting non-structural targets such as vehicles, people, or crops without rewriting the core flight and imaging logic.