A Vehicle-Mounted Radar Clutter Suppression Method on the Move Integrating IMU-Assisted Measurement and Dynamic Time–Frequency Ridge Trackings

The proliferation of small unmanned aerial vehicles (UAVs) operating at low altitude and in swarms has raised increasing concerns for low-altitude airspace safety and poses a central challenge to ground-based low-altitude radar monitoring. When a vehicle-mounted radar operates while in motion, terrain-induced three-axis attitude perturbations give rise to a time-varying clutter Doppler spectrum within each coherent processing interval (CPI), manifested as spectral peak drift and dynamic spectral broadening. Moreover, the radial velocity of slow-flying low-altitude UAVs frequently overlaps the clutter Doppler band, so such targets tend to be masked by strong ground clutter, which severely degrades the performance of conventional clutter suppression methods. To address this problem, a clutter suppression method that combines dynamic time–frequency (TF) ridge tracking with inertial measurement unit (IMU) auxiliary measurements is proposed, which shifts the clutter suppression problem from static filtering in the range–Doppler (RD) domain to dynamic ridge tracking in the TF domain. The clutter Doppler centroid and spectral spread are first predicted from IMU data through a kinematic inversion chain, yielding an echo-independent prior that remains reliable under low signal-to-clutter conditions. After a short-time Fourier transform (STFT), this prior constrains the search space of the Viterbi tracker, enabling frame-by-frame localization of the clutter ridge and frame-level adaptive adjustment of the notch parameters, thereby achieving accurate suppression of the non-stationary clutter while preserving the energy of slow-moving targets. Simulations of a close-range, slow-moving, and low-altitude UAV-swarm scenario demonstrate that the proposed method attains effective clutter suppression and maintains a high detection probability for slow-moving UAV targets at a low false alarm rate, allowing it to support low-altitude drone swarm monitoring from a moving radar platform. The complete processing chain requires only about 3×108 operations per coherent processing interval, so that real-time operation on an embedded platform is feasible. An error budget analysis shows that this prior requires only a calibrated consumer-grade MEMS inertial sensor and that the installation alignment is the binding requirement rather than the sensor stability. The validation presented here is based on simulated clutter and platform motion; validation against measured radar echoes is left for future work.