Robinson Ridge Lichen Segmentation Maps Using UAV imagery and Machine Learning

This is a child record, see the parent record AAS_4628_UAS for more information.This work was completed as part of the program Securing Antarctica's Environmental Future (SAEF).Dataset Components (Open Access)1.    Output Segmentation MapsThis record includes vegetation segmentation outputs generated by standalone deep learning models (DeepLabv3+, FCN, U-Net) and an ensemble stacking approach using XGBoost. These models were used to classify vegetation communities in the study area. The results provide detailed segmentation maps that highlight the distribution of different vegetation types. The output segmentation maps are critical for understanding and analysing the vegetation communities in Robinson Ridge, Antarctica. These maps, generated through advanced machine learning techniques, support further research, validation, and applications in environmental monitoring. The inclusion of Vegetation Indices (VIs) helps in deriving additional insights into vegetation health and distribution.Data Collection and AnalysisData was collected in January 2023 using a BMR3.9RTK UAV developed by SaiDynamics Australia, equipped with a MicaSense Altum multispectral sensor and a Sony Alpha 5100 RGB camera. Flights at 70 m altitude yielded a ground sampling distance of 2.93 cm/pixel. A total of 2,814 images were collected, covering around 5.15 hectares.Usage NotesRefer to the readme.txt files in each record for further details on data formats.Files:-    .dat, .hdr, .tif – Segmentation maps and vegetation indices-    readme.txt – Description and metadata for data usage and formats