Drone audition: dataset and methods for ground surface material classification using drone noise in outdoor environment

This paper considers a technique to estimate the material of ground surfaces directly below the drone based on the rotor noise generated by the drone. It is anticipated that this technique will enable us to assess the damage in areas that are not easily accessible by humans. In this study, drone noise was recorded while flying over multiple materials, such as asphalt, soil, and water, in a real environment with external noise, such as wind. Following, a machine learning-based model was trained using this dataset to classify the various materials using the recorded audio. Results indicate that while the classification was successful, there is a need for a classification method that is more robust to external noise.