Classification of Benthic Habitat Based on Object with Support Vector Machines and Decision Tree Algorithm Using Spot-7 Multispectral Imagery in Harapan and Kelapa Island

The research of object based image classification (OBIA) with machine learning algorithm for high resolution image in Indonesia is still limited especially for coral reef mapping, therefore further research needed for comparison in method and application of algorithms as alternative of classification. This research aims to map benthic habitat based on multiscale classification using OBIA method with support vector machine and decision tree algorithm in Harapan Island and Kelapa Island, Kepulauan Seribu. Segmentation was performed using a multiresolution segmentation algorithm with a scale factor of 15. The OBIA method is applied to atmospheric corrected images with a predefined benthic habitat classification scheme. The overall accuracy of SVM and DT algorithm implementations are 76.68% and 60.62%, respectively. The Z statistic value analysis obtained from the application of two algorithms used is 2.23, where this value indicates that the classification with SVM algorithm is significantly different from the DT algorithm. This research suggest that the OBIA technique could be a promise approach for mapping benthic habitats