PENERAPAN ALGORITMA SPECTRAL ANGLE MAPPER (SAM) UNTUK KLASIFIKASI LAMUN MENGGUNAKAN CITRA SATELIT WORLDVIEW-2

Abstract

Remote sensing technology has been developed for monitoring and identification of coastal environment and resources, such as seagrasses. In Indonesia, particularly seagrass mapping spectrometer utilizing spectral library has not been done. This study aimed to determine the spectral signature based in situ measurement and image analysis, analyze the implementation of the algorithm Spectral Angle Mapper (SAM) and test accuracy in mapping seagrass to species level based on spectral libraries. Research conducted in seagrass Tunda Island, Banten. Satellite imagery used is WorldView2 and the seagrass spectral reflectance was measured using a spectrometer USB4000. SAM classification algorithm utilizing spectral libraries and classify objects in a single pixel can be homogeneous. Classification results in the form of class Enhalus acoroides, Cymodocea rotundata, Thalassia hemprichii, and Halophila ovalis. The resulting accuracy of 35.6%. The area of each class is 0.8 hectares for the class Cymodocea rotundata, 2.79 hectares for Enhalus acoroides, class Thalassia hemprichii 3.7 hectares, and 3.5 hectares for Halophila ovalis. Classification of seagrass to species level yet produce good accuracy. Seagrass area with a variety of species and number of channels on a multispectral satellite image is assumed to be the cause of the low value of accuracy.