画素レベルディジタルフィルタを用いた高速振動源追跡アルゴリズム

抄録

In this study, we propose a fast algorithm for vision-based vibration source tracking that can significantly reduce the computational complexity in pixel-level digital filters. The algorithm reduces the complexity by adjusting the sampling rate to twice the vibration frequency of the target object to be tracked using frame interpolation. The computational complexity and properties of the pixel-level digital filters for downsampled images are evaluated; the proposed algorithm can accelerate the pixel-level digital filters, while retaining the admissible filter property for vibration source tracking unlike the approach used previously in which the input images captured at a high frame rate were directly processed using pixel-level digital filters. The pixel-wise accuracy in vibration source tracking is demonstrated by several evaluation results of offline high-frame-rate videos, which displayed various flight scenarios of a multicopter with its propellers rotating at 90-100 Hz.