Abstract: The track-before-detect method detects and tracks weak targets by non-coherent integration of target signals over continuous multiple frames. The key to this integration lies in the accurate estimation of the target trajectory and multi-frame iterative filtering. Traditional particle filters rely highly on the proposal distribution, and thus the estimation of the target trajectory is not accurate enough. The newly proposed particle flow filter is a promising alternative. However, it relies significantly on current measurement and neglects the multi-frame iterative filtering. In this paper, a novel track-before-detect strategy is presented. The particle filter is exploited for multi-frame iterative filtering, but within each frame, the filtering process is completed by the Localized Exact Daum-Huang filter. In order to deal with the uncertainty of measurement, the Localized Exact Daum-Huang filter is modified. Each particle finds in its neighborhood the measurement with maximum likelihood, and then the measurement is used to update the particle state. The performance of the proposed algorithm is evaluated by detecting and tracking a Swerling 1 target in Rayleigh clutter.