Marine Target Detection from Nonstationary Sea-Clutter Based On Topological Data Analysis

Due to the instinct complexity and the large scale non-stationary of so-called sea-clutter, radar backscatters from ocean surface, it is always challenging to detect the weak marine target. In classical statistical approaches, the seaclutter is modeled as several kinds of stochastic processes, which are found inadequate, especially in high sea-state circumstances. Therefore it is reasonable to discover the underlying dynamics that is responsible for generating the time series of sea-clutter. In this work, we take into account of the marine target detection from the X-Band seaclutter datasets with low Signal-Clutter-Ratio, and propose adequate methods to process these non-stationary data, including Empirical Mode Decomposition and Topological Data Analysis. Both theoretical simulation and experimental results indicate the proposed method's usefulness of for marine target detection, which is implemented by extract different structural features from measured sea-clutter data.