MUSAK: A Multi-Scale Space Kinematic Method for Drone Detection

Accurate and robust drone detection is an important challenging task. However, on this issue, previous research, whether based appearance or motion features, has not yet provided a satisfactory solution, especially under complex background. To end, the present work proposes motion-based method termed Multi-Scale Space Kinematic (MUSAK). It fully leverages patterns by extracting 3D, pseudo 3D 2D kinematic parameters at three scale spaces according to keypoints quality builds Gated Recurrent Unit (GRU)-based branches for recognition. The MUSAK evaluated hybrid dataset named multiscale UAV (MUD), consisting of public datasets self-collected data with labels. experimental results show that improves performance large margin, 95% increase in average precision (AP), compared state-of-the-art (SOTA) methods, method, which integrates appearance-based Faster Region-based Convolutional Neural Network (Faster R-CNN), achieves new SOTA AP metrics (AP, APM, APS).