Track-Before-Detect for Automotive Radar Using Multiple Frames with Time-Varying Fields of View

Track-before-detect (TBD) has been widely studied due to its superior detection performance for low-observable targets in complex environments. In existing works, traditional TBD methods are always considered for stationary radars and are based on an implicit assumption: the presence of the target is unchanged in the sliding window. However, for the extension of TBD to the automotive radar system, on the one hand, movements of the ego-vehicle result in the field of view misalignment; on the other hand, when the vehicle executes sudden and violent maneuvers, the target will fail to fall into the limited field of view temporarily in the sliding window. In order to solve the problems mentioned above, this paper presents an energy accumulation strategy based on motion compensation, which would be processed across coordinate systems, and a multiple hypothesis detection method with adaptive threshold (AT) to avoid detection performance degradation. Simulation results for various scenarios are provided to corroborate that the AT-TBD method enables effective detection even in complex cases where target presence changes throughout the time horizon corresponding to the current sliding window.