#TITLE_ALTERNATIVE#

• Main Author: ACHMAD REZA (15110060 ); Pembimbing : Dr. Ir. Kosasih Prijatna, M.Sc. , AGHA
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/17818
• Institution: Institut Teknologi Bandung

One of the problem that occured when we use the remote sensing data is the appearance of the mixing pixel, especially when we use the image that has lower spatial resolution. Mixing pixel is the problem that appear when we do the conventional classification, such as Maximum Likelihood, which assumes that one pixel only represents an object. Besides, in urban or metropolitan area, one pixel could be a combination of many objects, such as building, road, etc. This research attempts to apply the sub-pixel classification method, that is Spectral Mixture Analysis (SMA) using Landsat 8 image. There are 3 endmembers which is assumed that land cover of the urban area makes up by the linear combination of this 3 endmembers, that is vegetation, impervious surface, and soil. In addition to identify the land cover of Bandung City, this research conducts to compare the result from the developed agorithm of SMA with the Linear Spectral Unmixing (LSU) in ENVI software as well. From the SMA algorithm, the result shows vegetation area by 86853.24 acres, impervious surface by 40556.61 acres, and soil area by 54074.79 acres, and RMS error by 0.005005. From the LSU in ENVI software, the result show vegetation area by 86853.15 acres, impervious surface by 40552.56 acres, and soil area by 54075.69 acres, and RMS error by 0.004966. From the RMS error value and qualitative testing, SMA shows good result to identify land cover in Bandung