高速ビデオ画像解析に基づくソフトウェア回転センサ

抄録

In this study, we propose a novel concept of real-time software rotation sensor that can simultaneously detect the angles of multiple rotational objects in a high-speed video sequence. Our software sensor can be executed in real time at 500 fps by parallel-implementing digital image correlation processes to inspect the similarities among the input image and 360 reference images at different angles on a GPU high-speed vision system. Its performance in measurement accuracy was verified by showing several experiments results for multiple rotating gears that are captured with partial occlusion in 500-fps videos,including multiple gears fast-rotating at 2400 rpm.