Noise Filtering Algorithm Based on Machine Learning for Identification of Ground Hitting Photons in Jaipur City

Accurate estimation of building height is crucial for urban planning, infrastructure development, and environmental assessments. This research paper investigates the potential of utilizing ICESat-2 photon counting LiDAR data obtained from open altimetry to retrieve building heights in the urban area of Jaipur City. Many times, the elevation data given by ICESat-2 satellite contain noise, which has to be removed beforehand. The initial elevation data is processed for noise removal using the confidence score provided in the dataset. Further, feature extraction is performed using the DBSCAN windowing algorithm, which identifies ground photons and refines the photon set for height estimation. Finally, building height is retrieved by filtering the ground hitting photons from that of building hitting photons and taking the difference between their elevations. The estimated building heights from the proposed noise filtering approach agree well with that of the field measurements for buildings in parts of Jaipur city with mean accuracy of 11 cm.