The First Assessment of PhiSat-2 Imagery for Monocular Building Height Estimation
The study assessed PhiSat-2 imagery for monocular building height estimation by creating a dataset (PHDataset) and a model (TSONet). TSONet jointly predicts building height and footprint using two streams, improving accuracy over existing methods. Results show PhiSat-2 offers useful spatial-spectral cues for urban mapping but performance is limited by data quality and inherent inconsistencies in the imagery.