Abstract:
The detection of floating small targets is a challenging problem for marine surveillance radar. To effectively detect the floating small target in a complex marine enviro...Show MoreMetadata
Abstract:
The detection of floating small targets is a challenging problem for marine surveillance radar. To effectively detect the floating small target in a complex marine environment, this letter proposes an innovative multi-featured detection method by leveraging the graph signal processing (GSP) theory. With GSP, we propose two kinds of graph representations of standardized Doppler power spectrum (SDPS) to capture the correlation of the radar data in the Doppler domain. Then, by exploiting the graph representations, three quantitative graph features, graph Laplacian regularizer, trace of Laplacian matrix, and variance of self-loop weight, are developed to distinguish target returns from sea clutter. Finally, a detector based on the graph features is constructed by the fast convex hull learning algorithm. Experiments conducted on the measured radar datasets and comparisons with existing methods confirm the effectiveness of the proposed method.
Published in: IEEE Geoscience and Remote Sensing Letters ( Volume: 20)