A two-stage detection method based on improved DP-TBD for marine weak extended targets

Detecting weak extended targets on the sea surface under cluttered background is very challenging, and traditional detection methods are affected by clutter and low signal-to-clutter-and-noise ratio, resulting in poor detection performance. In order to improve the detection and discovery capabilities of weak extended targets on the sea surface, this paper proposes an improved method based on the dynamic programming track-before-detect (DP-TBD) algorithm, which is called DK-TBD. DK-TBD has two stages. One is the rough TBD detection based on virtual nodes, which solves the problem of target loss caused by environmental factors in classical DP-TBD. The other is the re-tracking based on kernel correlation filtering algorithm, which is used to correct the state of weak extended targets after rough detection. Compared with other DP-TBD-based methods, our method can better detect weak extended targets on the sea surface in simulated and real data with complex clutter as background.