Comparison of ML and DL Based Methods for Forest Height Estimation Through Multi-Channel SAR Data

Forest measurement plays a vital role in monitoring climate change and quantifying the global carbon cycle. Synthetic Aperture Radar Tomography (TomoSAR) has long been an effective technique for reconstructing three-dimensional (3D) forest structures. With recent advancements in Artificial Intelligence (AI), methods based on Machine Learning (ML) and Deep Learning (DL) have been proposed for estimating parameters in forested areas. Both approaches address the height retrieval problem as a classification task. In this paper, two methodologies-ML-based CatBoost and DL-based TomoSAR Neural Network (TSNN)-are introduced and compared in terms of accuracy and computational load, demonstrating that they can effectively replace traditional model-based TomoSAR methods.