EKF-Based Joint Calibration for Motion Compensation of UAV-Borne 4D Millimeter-Wave Radar Point Clouds

The point cloud acquired by an unmanned aerial 4D millimeter-wave radar suffers from motion distortion due to high-speed displacement and attitude variations of the detected targets. To reconstruct the true spatial distribution and kinematic state of the point cloud, this paper proposes a multi-sensor spatiotemporal joint calibration and motion compensation method based on the Extended Kalman Filter (EKF). First, a spatiotemporal alignment model between external sensors and the radar is established to correct timebase synchronization errors. The external sensor measurements are then fed into the EKF to construct a state transition matrix incorporating position, velocity, and attitude, enabling optimal estimation of the airframe motion parameters through radar pose observation updates. Subsequently, a multi-coordinate transformation chain is derived to rectify the global point cloud distribution. Additionally, based on a velocity projection model, the 4D point cloud is decomposed to compensate for spatial and velocity bimodal errors before mapping to the global coordinate space. Experimental validation on millimeter-wave radar data confirms that the proposed method achieves superior motion compensation compared to non-calibration-based approaches.