Assessment of a Landsat 8 image of Atiwa District in Ghana using ECHO, Random Forest, Minimum Distance and Maximum Likelihood Classification Algorithms
Remote sensing for land use land cover classification has become a very important and common method since it is faster and more efficient than some land surveying methods. Most research works in recent times indicate that image classification of satellite images for land cover information using remote sensing is the preferred choice. Various methods for image classification have been developed based on different theories or models. In this study, Extraction and Classification of Homogeneous Objects (ECHO), Random Forest, Minimum Distance Classification (MDC) and Maximum Likelihood Classification (MLC) methods are utilised to classify a Landsat 8 image of Atiwa District in the Eastern Region of Ghana. The study area was grouped into forest, bare land, settlement and vegetation land use land cover classes. After comparing the results of the four classifiers, the Random Forest classifier produced the highest results with an overall accuracy of 92.37% and kappa of 0.9103 as compared to Minimum Distance with an overall accuracy of 81.58% and kappa of 0.7538, Maximum Likelihood with an overall accuracy of 76.32% and kappa of 0.6822, and ECHO
with an overall accuracy of 74.56% and kappa of 0.6587.