An agile multi-frame detection method for targets with time-varying existence

Existing solutions for multi-frame detection (MFD) of dim targets commonly include a sliding window batch processing routine, in which multiple consecutive measurement frames are jointly processed at each measurement time. While the sliding window batch processing improves the detection performance, it can lead to serious performance deterioration. In particular, the batch processing approach is known to cause latencies in detection decisions once abrupt changes to target existence (such as target spawning or death) occur. The detection latency is fundamentally due to an implicit assumption in derivation of the test statistics: the status of target existence is assumed unchanged throughout the time horizon corresponding to the current sliding window. This paper presents a solution called Agile MFD that is immune to the performance degradation caused by unpredicted changes in target existence. Agile MFD is designed to discover a change in target existence rapidly by modeling target existence as a time-varying random variable, incorporating it in the test statistics, and estimating it along with the target states. The results of extensive numerical simulations demonstrate that the proposed Agile MFD method can significantly reduce the decision latency and achieve stable performance even in extreme cases where target existence randomly changes in frequent times. (C) 2019 Elsevier B.V. All rights reserved.