Abstract: In this paper, a new algorithm of runways detection based on polarimetric decomposition and structural characteristics in complex scenes of fully polarimetric synthetic aperture radar image is proposed. Firstly, training samples are obtained according to the priori of backward scattering mechanism of the terrain and H/α decomposition. Then runway suspected areas are achieved after Bayesian classification combining the statistical characteristic of covariance matrix and the training samples. Finally, binary decision tree is used to identify runways combining with five structural characteristics, to eliminate the suspected runway areas and to finalize the airport runway region. Multi-look fully polarimetric SAR datas acquired by U.S.UAVSAR systems is used to verify the new algorithm. The experimental results show that the novel method can detect runways effectively. Besides, the detected runways own a intact structure, clear outlines and low false alarm rate.