Design and Evaluation of a Artificial Intelligence Based Vessel Detection System in Pol-SAR Images

Inhalt:
This paper tackles the design and evaluation of a vessel detection system based on intelligent agents that allow the jointly exploitation of statistical information from polarimetric SAR (Syntehtic Aperture Radar) images. Multilayer Perceptron (MLP) and cost-sensitive Support Vector Machine (2C-SVM) are considered to study the suitability of the best architecture. Detection performances are compared to a CFAR (Constant False Alarm Rate) detector based on a Generalized Γ Distribution sea clutter model and an OR-logic post-detection combination. Regardless the intelligent agent architecture, the proposed detectors outperforms the reference one resulting the 2C-SVM detector slightly better than MLP solution.