Host: The Japan Society of Mechanical Engineers
Name : [in Japanese]
Date : June 05, 2019 - June 08, 2019
This study proposes a novel method for drone surveillance that can simultaneously analyze time-frequency responses at all pixels of a high-frame-rate video. Propellers of a flying drone are rotating at hundreds of Hz, and their principal vibration frequency components are much higher than those of their background objects. In order to separate pixels around a drone’s propellers from its background scene, we utilize these time-series features for vibration source localization with pixel-level short-time Fourier transform (STFT). We verify the relationship between the number of taps in the STFT computation and the performance of our algorithm such as execution time and localization accuracy by conducting experiments under various conditions such as degraded apparency, weather condition, and defocused blur. Its robustness is also verified by localizing a flying multicopter in real time in an outdoor scenario.
