The Detection Of Airport Runways Based On Bayesian Classification In PolSAR Image

Polarimetric Synthetic Aperture Radar(PolSAR) is a multi-parameter and multi-channel microwave imaging radar system. Meanwhile, it isn't affected by the weather and time and can achieve high resolution imaging for the region of interest(ROI). Comparing to the single-polarization SAR systems, the data collected by PolSAR systems contains more comprehensive surface information and can be used to extract polarimetric scattering characteristic of targets. Therefore, PolSAR has been widely applied in many fields such as targets detection, classification, targets identification and so on.Airport is an important military and civilian transportation facility and contains lots of important information. The automatic detection technology of airport has important application value in many fields, such as precision strike, emergency rescue, military reconnaissance and so on. The runway is one of the airport remarkable features, so its detection can be as the basis of airport identification. In this paper, the runway detection method based on PolSAR image is studied.Based on the research of the classification method and runway's structural features or polarization scattering properties in PolSAR image, two new algorithms of runways detection are proposed. In the first algorithm, different kinds of training samples in image are selected to build templates for each class based on 'H/ S classification firstly. Then, the whole image is classified with supervised Bayesian classification by combining the statistic characteristic of coherency matrix with training samples. Finally, the real runway can be detected by using Morphology filtering and the runways structural features. In order to overcome the dependence of sample information in the former algorithm, the second algorithm applies unsupervised classification to runways detection. It uses h/q decomposition and iterative Bayesian classifier to extract suspected runway areas, and adds the weak backscattering characteristic of runways to discriminate the real runway. Multi-look fully PolSAR datasets acquired by U.S.UAVSAR systems are adopted to verify the efficiency of the new algorithms. The experimental results show that the two novel algorithms can both detect runways effectively and have a low false alarm rate. Besides, the detected runways keep an intact structure and clear outlines. Comparing the results of two algorithms, the former method can detect runways rapidly and accurately and meets the requirement of real-time. The latter method is more practical than the former one. Because it adopts unsupervised classification to extract suspected runway areas without using sample information.