How the simulation works
A test road with lane markings, patches, sealed cracks and a generated set of pavement distresses is laid out in world coordinates. The aircraft takes off from its dock, follows the flight lines and triggers the camera at the trigger distance. Every frame is rendered from the true camera pose, then passed to a simulated onboard detector whose confidence depends on crack width, ground sample distance and framing.
After landing, frames are matched, aligned and blended into an orthomosaic, and detections from overlapping frames are merged into one list of unique defects with stations, lane offsets and coordinates.
What is generated
Longitudinal cracks
Transverse cracks
Alligator cracking
Edge cracks
Potholes
Patches and utility cuts
Sealed cracks, pole shadows and worn lane paint are also placed on the road as the kind of things a visual model can mistake for damage.