Land cover and vegetation mapping of tropical forest areas using a very high-resolution image in Central Borneo

Abstract

This study analyses and compares land cover (LC) classification results of tropical forested areas in Central Borneo using very high-resolution imagery. The Pleiades image (spatial resolution less than 1m per pixels) is used as the primary input in this study. A neural network -multilayer perceptron (NN-MLP) algorithm is used and is compared with two well-known pixelbased classification algorithms, i.e., Maximum Likelihood classifier (MLC) and ECHO (Extraction and Classification of Homogeneous Objects). The ECHO is varied using (2x2; 4x4; and 6x6) homogeneous pixel group. The study covers an area of 162.60 km2 located in Central Borneo. The classification result produces nine (9) land cover classes, i.e., pavement, sparse vegetation, dense vegetation, bare land, palm oil plantation, mixture grass, sand, mining area, and water body. Classification using NN-MLP, MLC and ECHO produced kappa and overall accuracies of more than 90%. In general, the three algorithms can produce a relatively similar area extend for each class.