Visual-based automated aircraft inspections

This project aims to develop a system capable of scanning the exterior of an aircraft and detecting 3D shape anomalies. The system must not require contact, operate at safe distance and in real scenarios. The objectives of this project are:

1. To target damage identification and thresholds (size, shape of dents) - for example, as stated by Boeing 737-400 SRM, dents with depths of less than 3.175mm (0.125 ins) on the fuselage are allowable;
2. Critical comparison of 3D reconstruction algorithms and technologies;
3. Identification of state-of-art techniques suitable for building a reliable (less than 5% error on benchmark datasets) one-shot 3D scanning system capable of delivering a point cloud in near-real time (~30 fps);
4. Development and testing of a point set registration algorithm to combine multiple acquired 3D scanning to refine the accuracy of final model;
5. Development and testing of a selection algorithm to automatically identify relevant damages; and
6. Increasing robustness of the algorithms by testing them in real-case scenarios, possibly integrating further data from other sources (sensors, aircraft CAD model).