Ground Surface Material Classification with Drone Noise

This paper discusses a technique for determining the type of ground material directly

beneath a flying drone by analyzing the noise emitted by its rotors. Identifying ground

material using drone rotor noise reflected from the ground surface, we aim to assess damage

in disaster areas that are difficult for humans to access. In our research, we first collect a

four-hour dataset that contains recordings of drone noise emitted and reflected by five different

types of ground surface materials in an anechoic chamber. We then fed these recorded

drone noise signals into CNN and ResNet models to estimate the ground surface materials,

achieving accuracy rates of 80.8% and 73.3%, respectively.