Improvement of ship detection accuracy by sar multi-look cross-correlation technique using adaptive CFAR

This paper describes a novel technique to improve the accuracy of ship detection by synthetic aperture radar (SAR). The methodology is to apply the algorithm of constant false alarm rate (CFAR) to the coherence images produced by multi-look cross-correlation (MLCC) of sub-images. The experiment was carried out in the Tosa bay, Kochi, Japan using three known fishing boats and the simultaneous data acquired by the Phased Array L-band SAR (PALSAR) on board of the Japanese Land Observing Satellite (ALOS) at four different modes. The inter-look coherence images were found to fit best to Gamma distribution, and using this distribution function, CFAR was applied to the coherence image. The results showed substantial improvement in signal to noise ratio and false alarm rate as compared with the coherence image alone. In the followings, the principle of MLCC-CFAR and experimental procedure are summarized, and the results are discussed.