Bernoulli Track-Before-Detect Filter With Pulse-to-Pulse Correlated Amplitude Model for Maritime Swerling I Targets

The Bernoulli track-before-detect (BTBD) filter with amplitude measurements has achieved significant success in maritime radar surveillance for joint target detection and tracking. However, its independence assumption across target units limits its ability to accumulate energy from the extension of weak Swerling I targets, leading to underestimations of target existence and unexpected target misses. This article aims to effectively accumulate energy from short-exposure extensions of Swerling I targets within the BTBD framework. We first analyze the inconsistency between the measurement model of the existing BTBD filter and the pulse-to-pulse correlation characteristics of Swerling I targets. Based on this, we derive the theoretical likelihood function of the correlated amplitude model and propose two approximation methods for computational simplification. The first method employs maximum likelihood estimation (MLE) to approximate the theoretical likelihood while retaining the core architecture of the original BTBD filter, allowing for a straightforward comparison with the original method. The second method adopts an intensity resampling strategy based on the particle filtering implementation of the BTBD filter. This strategy seamlessly integrates a standard Bayesian filter to maintain a recursive estimation of the mean intensity of the target, thus providing adaptations to the fluctuations of Swerling I targets between scans. Experimental results using both simulated and measured radar data demonstrate the superior performance of the proposed filters in terms of trajectory integrity, interruption reduction, and state estimation accuracy.