Built-Up Area Detection based on Degree of Polarization Analysis in Frequency Domain Using Fully PolSAR Data

This work proposes a built-up area detecting method by analyzing the fluctuation of degree of polarization in frequency domain. The averaged detrended amplitude of the two dimensional fast Fourier transform is proposed for describing the fluctuating information, and serving as the new discriminator. The comparison results for ALOS2-PALSAR2 data of Hakodate area, San Francisco area, and Ebetsu area prove that the proposed method has stably higher detecting accuracy, basically higher than 90%, for various types of built-up areas, much fewer error responses, basically lower than 5%, in natural area, and better performance on recognizing small buildings from vegetation background. Two application examples, i.e., urbanization process observation of Weihe area, and collapsed buildings detection of Mashikimachi area after the Kumamoto earthquake further confirm its good performance.