Landslide Detection Using Polarimetric Sar Data and Genetic Programming (PAP014710)

Código

PAP014710

Autores

Fábio Sato, Marco Aurélio Silva Neto, Sérgio Scheer

Tema

Land use and Floods, landslides and erosion

Resumo

We studied the utilization of polarimetric SAR information and machine learning algorithms to detect landslides. Decomposition methods of the scattering matrix were applied to L-Band SAR data from ALOS satellite and used as input for machine learning classification algorithm based on genetic programming. A case study of a landslide event triggered by rainfall in Brazil was conducted to evaluate and compare decomposition methods and its applicability for landslide detection. Obtained results were compared with manual classification using high-resolution satellite optical images.