Can support vector machine be used to predict and estimate land use and land cover (LULC) mapping using Landsat images?

Authors: Dong Luo*,
Topics: Remote Sensing, Land Use and Land Cover Change
Keywords: support vector machine, land use and land cover (LCLU), predict, Landsat images
Session Type: Poster
Day: 4/13/2018
Start / End Time: 1:20 PM / 3:00 PM
Room: Napoleon Foyer/Common St. Corridor, Sheraton, 3rd Floor
Presentation File: No File Uploaded

Predicting and estimating LULCC is important for government and decision makers. This study wants to test if support vector machine algorithm can predict and estimate LULC using Landsat images for the period of 2000 to 2017. Landsat 7 ETM+ images (2000, 2001 and 2002) and Landsat 8 OLI images (2013, 2014, 2015, 2016 and 2017) for the dry season area of Bento Rodrigues tailing dam, in the Atlantic forest of Brazil were used in this study. To achieve our goals, we developed the following steps: 1) we chose blue, green, red, NIR, SWIR-1 and SWIR-2 bands as 6 features to classify 2000 and 2017 images respectively; 2) we added normalized difference vegetation index (NDVI) as an additional feature and used them to classify the same images; 3) we selected 2001, 2002, 2013, 2015 and 2016 images as training data to predict the same year's images using 7 features. The overall accuracy of classified and predicted images had both good performances. The accuracy of predicted images of 2000 (Landsat 7 ETM+) and 2017 (Landsat 8 OLI) is 83.10% and 87.01%, respectively. The results suggested that support vector machine can be used to predict LULC from Landsat images. The results also showed that for the period of 2000 to 2017, the forest decreased approximately 202.56 km2 but urban area, mining and agriculture areas increased 13.69 km2, 11.00 km2 and 170.84 km2, respectively.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login