Multi-levels classification for tropical tree species identification using in-situ hyperspectral data

Abstract

Multi-Levels classification process has been applied to handle a large number of tropical forest tree species using in-situ Hyperspectral data. Improvement in overall classification accuracy has been achieved with Support Vector Machine and Maximum Likelihood Classifier in identifying twenty tropical forest tree species. The improvement in accuracy for the both classifiers about 5% from the first level classification to the third level classification process in this study. The findings have proven the effectiveness of this multi- levels classification procedure in handling the issue of high diversity is tropical forest environmental during tree species mapping.