Aim

Complex distribution modelling of the sea clutters to improve the target detection performance in terms of ROC, TP, FP figures.

Objectives

1. Complex sea clutter modeling to separate background from targets.
2. Reduce the sample complexity of the distribution modelling
3. Using data-adaptive methods not only for distribution modelling, but also in the decision process chain.
4. Reduce the computational complexity of the method to achieve real-time performance. 

Description

The recent advances in computational methods and data-adaptive approaches, opened a new possibility in using adaptive models for sea-clutter environments as a (non-stationary) stochastic process and design a high-resolution range-azimuth-Doppler detectors (3-dimensional problem). While the conventional sampling paradigm is prohibitively expensive in (very) low signal to interference ratios (SIR), the aim of this project to come up with data-efficient clutter modelling. The modelling of clutter can be in the form of a stochastic process or a deterministic predictor, which the later can be done using the recent deep predictive models [1].

The project aimed at quantifying the strategies of compound gaussian [2] and machine learning [1] based modelling of sea clutter environments under realistic simulation settings and/or real Radar data. The tasks are the investigation of, a) robustness of non-Gaussian detectors to the clutter distribution-shifts and unknown statistical parameters, and b) maritime Radar universal features to provide sensor agnostics solutions.

References

[1] J. Wang and S. Li, "Maritime Radar Target Detection Model Self-Evolution Based on Semisupervised Learning," in IEEE Transactions on Geoscience and Remote Sensing, vol. 62, pp. 1-11, 2024.

[2] K. James Sangston and Alfonso Farina: “Coherent radar detection in compound-Gaussian clutter: Clairvoyant detectors”. IEEE Aerospace and Electronic Systems Magazine, 31, 42–63 (2016). doi:10.1109/MAES.2016.150132.

https://www.sciencedirect.com/topics/engineering/sea-clutter