Runway foreign object debris (FOD) detection

Runway Foreign Object Debris, commonly known as FOD, is a serious aviation safety issue because even relatively small objects can damage aircraft tyres, engines, landing gear or other critical components. Debris may include stones, metal fragments, loose fasteners, rubber, tools, packaging, parts of vehicles or aircraft, and materials left behind after maintenance or construction activity. On an active airfield, the objective is not simply to detect FOD eventually, but to identify and remove it quickly enough to reduce operational risk. Drones can add a mobile inspection layer to existing airport FOD-management procedures. Equipped with high-resolution RGB cameras, optical zoom and AI object-detection software, a drone can scan runway, taxiway and apron surfaces during approved inspection windows and highlight unusual objects for review. The aircraft can also create a georeferenced record showing the exact position of a suspected object so that airport operations teams can remove it quickly. The most important point is that drone FOD detection should complement rather than replace established runway inspections, FOD walks, vehicle patrols and any fixed automated FOD-detection systems already used by the airport. Very small objects can remain below the effective resolution of an aerial camera, so the value of drones is strongest in rapid surface screening, wide-area coverage and automated anomaly detection. ## **What Is Drone-Based FOD Detection?** Drone-based FOD detection uses an unmanned aircraft to inspect pavement surfaces for objects that do not belong there. The drone follows a preplanned route over a runway, taxiway, apron or maintenance area while collecting high-resolution imagery. AI then analyses the video or still images and identifies candidate objects that appear different from the normal pavement background. Each detection can be assigned a geographic position, confidence score and image. Airport personnel review the result and dispatch a vehicle or team to inspect and remove the object if necessary. This turns the drone into a mobile remote-sensing platform rather than simply a flying camera. ## **Why FOD Is a Major Aviation Risk** Aircraft operate at high speeds and with extremely tight engineering tolerances. An object that appears insignificant from ground level can become dangerous when struck by a tyre or ingested into an engine. FOD can damage engine blades, puncture tyres, affect braking components or strike the aircraft structure. It can also create secondary debris when an initial object is broken apart by a vehicle or aircraft. Because the consequences can be severe, airport FOD control relies on prevention, regular inspection and rapid removal. ## **Where FOD Comes From** FOD can originate from many sources around an airport. Aircraft themselves may shed small components, while ground-support equipment, maintenance activity and construction work can introduce loose objects onto operational surfaces. Packaging, tools, cable ties, stones, broken pavement, tyre material and metal fragments are all possible examples. Severe weather can also move debris onto runways or taxiways. Understanding the source is important because repeated detections in one location may indicate an underlying operational problem rather than isolated random events. ## **Why Use Drones for FOD Detection?** Airport surfaces cover large areas, and traditional inspection vehicles need time to drive them systematically. A drone can provide an additional aerial perspective and scan sections rapidly from above. The aircraft can also generate a complete photographic record of the surface rather than relying only on what a driver sees while moving. AI can then review that imagery repeatedly and flag areas that deserve human attention. This is especially useful after maintenance, construction, storms or other events that may increase the likelihood of debris being present. ## **High-Resolution RGB Cameras** High-resolution RGB cameras are the primary sensor for drone FOD detection. The smallest detectable object depends heavily on image resolution, flight altitude, lens quality and lighting. A drone flying high enough to map an entire runway quickly may fail to detect small bolts or metal fragments. Lower-altitude inspection provides more detail but reduces coverage and increases mission time. For this reason, mission planning should begin by defining the smallest FOD size the airport wants the system to detect reliably. ## **Ground Sampling Distance** Ground Sampling Distance, or GSD, is one of the most important technical concepts in FOD inspection. It describes how much real-world surface area each image pixel represents. If one pixel represents several centimetres, a small screw may occupy only a fraction of a pixel and become effectively invisible. If the GSD is much finer, the same object may occupy enough pixels for AI to detect it. The required GSD should therefore be derived from the target object size rather than chosen arbitrarily. ## **AI Object Detection** AI object detection can analyse every frame or photograph and identify objects that appear on the runway surface. The system may classify familiar items directly or simply flag any object that looks different from the pavement. For example, the AI may detect a tyre fragment, metal object or packaging item. In other cases, it may simply mark an unidentified anomaly for human review. This is often more useful than attempting to create a model that recognises every possible type of FOD. ## **AI Anomaly Detection** Anomaly detection is particularly promising because many FOD objects will be unfamiliar. Instead of asking the model to recognise a specific bolt, it learns what normal pavement looks like and flags anything unusual. A new object appearing between two surveys can therefore be identified even if the system has never seen that exact object type before. This approach is strongest when the runway has a high-quality baseline dataset. ## **AI Change Detection** Change detection compares the current runway surface with a previous inspection. Because runway geometry changes very little, a newly appearing object can be highlighted immediately. This reduces the need to distinguish every pavement marking, crack or repair patch from FOD repeatedly. Features that were already present during the previous survey can be ignored unless they changed. For repeat drone operations, this can significantly improve efficiency. ## **Baseline Runway Mapping** A baseline survey records the normal appearance of the runway, taxiway or apron. The dataset includes markings, pavement cracks, lights, drains and other permanent features. Future flights compare new imagery with this baseline. Objects that suddenly appear become much easier to identify. The baseline should be updated after maintenance or pavement work so the system does not continue flagging legitimate changes. ## **Metal Object Detection** Metal FOD can be especially dangerous because even small fragments may damage tyres or engines. RGB cameras can identify visible metal objects where size and contrast are sufficient. However, shiny metal may create reflections, while dark fragments may blend into the pavement. AI needs training across different lighting conditions and pavement types to avoid overestimating performance. ## **Rubber Debris Detection** Tyre fragments and rubber deposits may appear dark against asphalt, making them difficult to distinguish visually. Larger tyre pieces can often be detected by shape and shadow, while thin rubber fragments may be more challenging. Historical comparison and lower-altitude imaging can improve detection. ## **Stone and Gravel Detection** Loose stones can enter operational surfaces from shoulders, construction areas or damaged pavement. Small stones are difficult aerial targets because they may resemble normal surface texture. Drones are more useful for larger stones or areas where gravel has spread across the surf