A Video Dataset of Online Full-frame Co-stabilization of Long and Short Focus Binocular Vision for Bionic FWR

Inspired by the flapping wing flight and central fovea vision structures of raptors, bionic flapping wing robots (FWRs) and their multi-focus airborne vision can transmit video streams online to ground workstations and operators to perform aerial monitoring tasks. However, the raw video stream suffers from strong jitters due to the wing flapping. In this paper, a gimbal-free online full-frame co-stabilization method for long and short focus binocular synchronous vision is proposed. First, dynamic periodic jitters compensation and random jitters filtering are performed on the video motion trajectories based on the periodicity of the wing flapping. Then, the binocular frameexpansions are realized based on the spatial and temporal pixel transfer, and finally the stabilized binocular video streams with almost no edge blank are output online after frame warping. Through outdoor tests with a 1.8 m wingspan FWR and comparisons with existing methods, the proposed method has been verified the advantages of binocular co-stabilization with low-latency online processing and low field of view (FOV) loss, which does not need to take additional loads and power consumption like mechanical gimbals.