Abstract:

A new effective algorithm for complex scenes of polarimetric synthetic aperture radar (PolSAR) image runway detection based on multi-stage classification is developed. Firstly, training sample templates are selected according to h/q classification with the priori of backward scattering mechanism of the terrain. Then, the statistical characteristic of polarized coherent matrix of PolSAR images are used in the second stage classification. Thirdly, according to weak backscattering characteristic of the runways, the third stage classification is done by using the span detector. Finally, the airport runway region is identified with the size and structural characteristics of the runways. The datasets obtained by the U. S. UAVSAR airborne systems are used to do the experiments. Results show that comparing with the existing two methods, the new algorithm can detect the runways effectively with intact structures, clear outlines and low false alarm rates.