Improving the Classification of LANDSAT-8 OLI Images Using Neighborhood Median Pixel Values
Date of Award
Master of Science in Chemistry
Information Systems & Computer Science
Proceso L. Fernandez, Jr., PhD
Image classification in remote sensing is defined by categorizing image pixels or raw data sensed by satellites into a distinct set of labels. In this paper, an improved technique for classifying vegetation, built-up, and water pixels from satellite images is proposed. The technique makes use of the median value of all the pixels in the rectangular neighborhood centered at the given pixel to be classified. A scoring system was developed that compares this median value in relation to the expected median values for each of the different classes. The proposed method was tested on Landsat-8 Operational Land Imager (OLI) bands 1 to 7 images and three index images— Normalized Difference Vegetation Index (NDVI), Normalized Difference Built-up Index (NDBI), and Normalized Difference Water Index (NDWI). The experimental results showed an overall accuracy of 94%, a remarkable improvement from the 84% accuracy of the previous work that uses a distance-based classifier. The obtained results indicate that the proposed method can be a better alternative for classifying images in remote sensing.
(2020). Improving the Classification of LANDSAT-8 OLI Images Using Neighborhood Median Pixel Values. Ateneo de Manila University.