HFR-Video-Based Honeybee Activity Sensing

Abstract

This study realizes the activity sensing of honeybees flying in a natural outdoor environment by inspecting pixel-level temporal frequency responses in the brightness in a high-frame-rate (HFR) video, computed by executing short-time Fourier transforms (STFTs) of the brightness signals at all pixels. To localize multiple flying honeybees in an HFR video, the periodic brightness changes at pixels around the wings of honeybees that were flapping at hundreds of Hertz were detected in the frequency range below the Nyquist frequency, which is defined as half the frame rate of the HFR video by executing STFTs at all pixels. Vibration pixels around each honeybee were identified as a labeled region with connected components labeling. For each labeled region, STFTs were recomputed for brightness signals in a tracked region-of-interest image, in which its labeled region was located in the center; the trajectory of a honeybee is monitored with the peak frequency of these tracked STFTs as its flight activity. The performance of the HFR-video-based honeybee activity sensing was verified by analyzing several scenes of 12–14-mm-size honeybees flying around a beehive in $1024\times1024$ images captured at 500 fps.