Multi-task Unknown-Temporal Trajectory Perception Network for Maneuvering Radar Target Detection

Non-cooperative maneuvering target detection via long-time integration is inevitably constrained by the compound challenge of severe range-Doppler migration and unknown time information. These factors collectively hinder the effective focusing of target energy, thereby severely impairing the detection performance. To address these issues, this paper proposes a multi-task unknown-temporal trajectory perception network (UTTPNet) to extract target trajectories with unknown temporal information, while simultaneously achieving coherent accumulation and parameter estimation. The proposed method employs a multi-task joint learning architecture that not only locates the effective support interval (i.e., target entry and departure times) and detects the target trajectory in the range-slow-time domain but also establishes a mapping from the high-dimensional spatiotemporal features of radar echoes to a low-dimensional motion parameter subspace. This design enables a coarse-to-fine estimation strategy of target velocity and acceleration, yielding highly efficient parameter estimation. Extensive experimental results demonstrate that the proposed method achieves comparable detection performance to state-of-the-art conventional algorithm, but with a significantly lower computational cost.