Abnormal behavior detection in periodic operations and high-frame-rate (HFR) video logging were realized by introducing a visual machinery surveillance algorithm, which includes a phase-encoding process for periodic operations to match input images with pre-stored reference images efficientl. Assuming that the periodic machinery operations to be monitored are perfectly regular, the surveillance algorithm can encode the phase of a periodic operation just by inspecting temporal changes in the brightnesses at several significan pixels in an input image; abnormal behavior in the periodic operation can be detected by comparing the input image with a reference image synchronized with its encoded phase without the heavy computation needed to search all the reference images. This algorithm was software-implemented on a high-speed vision platform, IDP express, which can record input images of 512 ?? 512 pixels and process the results at 1000 fps. An experiment was performed using a sewing machine with periodic operation at a frequency of 12 Hz, and a vidio of the abnormal behavior was automatically recorded at 1000 fps to verify the effectiveness of HFR-video-based machinery surveillance.