Urban building height extraction accommodating various terrain scenes using ICESat-2/ATLAS data

Although the photon point cloud data acquired from ICESat-2/ATLAS can be efficiently employed in urban building height extraction, its universal applicability in undulating terrain scenarios is constrained, and there are noticeable issues of false positives and false negatives. This research establishes a terrain-adaptive methodological framework based on ICESat-2/ATLAS photon point cloud to extract high-precision, high-density building height data across varied urban topographical conditions. First, a terrain-adaptive elevation buffer is utilized to coarse denoise the photon point cloud, involving the removal of the majority of noise photons in the scene, thereby enhancing the efficiency of subsequent algorithms. Second, urban signal photons are extracted from the remaining original photons using the Adaptive Method Based on Single-Photon Spatial Distribution (SPSD-AM). This approach demonstrates high universality across various urban scenes, while simultaneously ensuring a stable accuracy of urban signal photon extraction. Subsequently, ground photons are extracted from the urban signal photons and fit the ground curve based on the Adaptive Method Based on Spatial Differences of Urban Signal Photons (USPSD-AM), which addresses the challenge of the potential mixing of ground and building photons in complex terrain scenarios. A precise ground curve is then employed to extract building photons from urban signal photons. In order to mitigate issues such as false positives and negatives, post-processing steps, including completion and denoising of building photons, are implemented. Finally, the acquired building photons and ground curve are adopted to extract accurate building height parameters. The precision verification results show that the extracted building heights are considerably consistent with the reference building heights. The mean RMSE and MAE are 0.273 m and 0.202 m for flat terrains and 1.168 m and 0.759 m for undulating terrains, respectively. The proposed method demonstrates superior applicability across diverse urban scenarios, providing a robust theoretical foundation for large-scale urban building height retrieval efforts.