Vegetation type classification system using pseudo zernike moments and ELM

摘要

Vegetation classification system is one of the evergreen applications of remote sensing technology. The objective of this classification system is to boost up the agricultural yield by removing the weedy plants. Understanding the benefits of vegetation classification system, this work proposes a vegetation classification system that can distinguish between tree, shrub and grasslands. The goal of the work is achieved by segregating the entire work into three important phases, which are image pre-processing, feature extraction and classification. The satellite images are pre-processed by guided filter. In this work, both global and local features are extracted by pseudo zernike moments and first order features respectively. Finally, Extreme Learning Machine (ELM) is employed as the classifier to differentiate between trees, shrubs and grasslands. The performance of the proposed approach is satisfactory in terms of classification accuracy, sensitivity and specificity.