Real-Time Detection of Terrain Surface Types and K-Factor for Path Profiling

Clients

Techstack

Purpose

Profiling the ground along a route means knowing what kind of surface lies beneath it and how prone that surface is to erosion. Doing this by hand across open terrain is slow and hard to standardize, and the readings that matter most are tied to surface type. A research team wanted to prove that a drone could read the ground below it as it flew and grade each stretch on its own.

Description

We built a system that watches a live aerial video feed and classifies the terrain beneath the drone in real time, sorting each surface into landforms, water, or man-made structures. As it identifies each surface, it looks up the matching erosion factor and logs every detection so the results can be scored for accuracy afterward. A live on-screen interface shows what the model is seeing frame by frame as the drone moves. Turning a moving camera feed into reliable, ground-truth terrain data is exactly the kind of real-time computer-vision problem we take on and ship.