Segmentation and Classification of Airborne GNSS-R Reflectivity Signals with Speckle Noise Mitigation

This article is dedicated to the study of Global Navigation Satellite System Reflectometry (GNSS-R) techniques for remote sensing applications, focusing on classifying the reflectivity of airborne signals to differentiate reflective surfaces along satellite traces. We propose an automatic segmentation algorithm using an online change point detector and off-line change point localization estimate. Given the presence of speckle noise in GNSS signals, a homomorphic log-transformation is applied to mitigate this noise. The system is shown to detect different land-forms in real flight experiments in France, using K-means clustering to identify sand, water bodies, and plain land.