Man-made Structure Extraction Based on Scattering Decomposition Using PolSAR Data

Yasumin Siriprathan, Junichi Susaki*, Yoshie Ishii, Tetsuharu Oba

Abstract

Man-made structure extraction is essential for urban planning, environmental monitoring, and disaster management. Optical sensors often face limitations due to weather and lighting conditions; in comparison, synthetic aperture radar (SAR) provides consistent imaging in all environments. This study leveraged Polarimetric SAR (PolSAR) data and advanced scattering decomposition techniques to enhance the extraction of man-made structures. The methodology involved the collection of microwave scattering data from concrete blocks at various angles within an anechoic chamber, which was then used to train machine learning models. These models were subsequently applied to real-world satellite data from the Advanced Land Observing Satellite-2/Phased Array type L-band Synthetic Aperture Radar-2 (ALOS-2/PALSAR-2) to ensure practical applicability of the approach. Scattering decomposition significantly improved the detection accuracy compared with using the original scattering data alone, offering a clearer identification of man-made structures. Incorporating the polarimetric orientation angle further enhances the classification accuracy, making it a valuable addition to the method. Our approach offers significant insights into the planning and management of sustainable man-made structures and resources.

Keywords: Man-made structure extraction, Polarimetric SAR, Scattering decomposition, Machine learning, Polarimetric orientation angle

Full Text: