Real-time High-speed Vision-based Vibration Spectrum Imaging

Abstract

This paper describes a real-time vibration spectrum imaging system that uses a pixel-level short-term Fourier transform (STFT). This pixel-level STFT function for high-frame-rate images at 1000 fps or greater is accelerated through its parallel-implementation on a GPU-based high-speed vision system, and 512×512 peak-frequency images in a wide frequency band can be obtained in real time at more than 50 fps. The effectiveness of our real-time vibration spectrum imaging method was demonstrated using the vibration results for a fan, multicopter, and guitar string, which had rotating or vibrating frequencies in the range of hundreds of Hertz.