Open AccessDetection and Measure Carstensz Glacier Area Changes Using Machine Learning Technique
Author(s) -
Rizaldi Suwandi,
Sani Muhamad Isa
Publication year - 2020
Publication title -
international journal of innovative technology and exploring engineering
Language(s) - English
Resource type - Journals
ISSN - 2278-3075
DOI - 10.35940/ijitee.c8531.019320
Subject(s) - glacier , thresholding , artificial intelligence , pixel , geology , remote sensing , satellite , feature (linguistics) , snow , computer science , physical geography , geomorphology , geography , image (mathematics) , engineering , linguistics , philosophy , aerospace engineering
Using satellite data for acquiring glacier outlines has become more popular in the last decade. Glacier change assessment is the main goal for deriving glacier outlines. It's important to make the best method to generate the glacier outline as there most of the glacier outline is made with manual delineation and spectral thresholding. This research used a machine learning model to deriving the glacier pixels from satellite data. The model trained using more than 80 thousand of a glacier and non-glacier pixels. The model that trained has been proved to able classified a glacier pixel with more than 99% accuracy in one of the best experiments. The NDSI (Normalized Difference Snow Index) proved to be the key feature to classifying glaciers and shown to be one the best combination with NDSI + GLCM + TIFF (Band 4). This model hopefully can be further expanded and installed directly in satellite so we can instantly make a glacier outline without any manual delineation or spectral thresholding needed