Comparing and Analizing IsoData and K-Means Method to Classifying Shallow Water Objects.
Indah Kartika (a*), Nurjannah Nurdin (b), Syafyuddin Yusuf (b), Banda Selamat (c), Farid Samawi (c), Chair Rani (c).
Hasanuddin University
Abstract
This research was conducted on March 2019 in Barrang Caddi Island, Spermonde Archipelago. The acquisition of Sentinel-2A satellite imagery was on February 24, 2019. DOS method (Dark Object Substraction) was used to atmospheric correction , water column correction using Lyzenga algorithm with values of ki / kj = 0.679844349 and a = -0,39554029. Image classification analysis using IsoData and K-Means unsupervised classification method. Gound truthing using the UPT (Underwater Photo Transect) method. Basic information on spectral values is obtained from Sentinel-2A satellites using band 2 with 490nm, band 3 with 560nm, band 4 with 665nm, and band 8 with 842nm. The results show that there are 4 dominant objects including live coral, rubble, algae, sand and seagrass. The highest spectral reflectance can be detected by green band (560nm) with IsoData and K-Means classifications. Unsupervised classification using IsoData method has capability to detect live coral and while the K-Means method is capable to explore the seagrass.
Keywords: Sentinel-2A, isodata, k-means, spectral, Spermonde
Topic: Marine Geodesy and Satellite Oceanography