Aldabashi, NawafWilliams, SamEltokhy, AmiraPalmer, EdwardCross, PaulPalego, Cristiano2021-11-132021-11-1325/06/2021https://doi.org/10.1109/IMS19712.2021.9574826https://bit.ly/3oj3kreScopusA 5.8GHz Doppler radar was used to monitor free flying honeybees entering and leaving their hive at a 2m distance. Free falling metal spheres of different size and materials were first used, along with radar cross section (RCS) simulations, for calibration of an in house continuous-wave (CW) radar system. The system was then applied to extract the RCS of free flying honeybees (n=164) at 5.8GHz, which fills a gap in the literature and was found to be in the range of-55 to-60dBsm ± 3dBsm. The Doppler radar was hence integrated with machine learning (ML) techniques to autonomously discriminate the incoming and outgoing flights of honeybees. A neural network built through a random forest algorithm and processing of the data as Line Spectral Pairs (LSPs) achieved a maximum accuracy of 87.83% with a Binary Cross Entropy loss of 0.4274 when interpreting hive departure/entrance events. © 2021 IEEE.en-USBistatic radarDoppler radarMachine learningRadar cross-sectionsSimulationArticlehttps://doi.org/10.1109/IMS19712.2021.9574826