Comparative Study of Pixel-Based and Object-Based Classifications in Benthic Mapping

Abstract

Coral reefs have been degrading rapidly throughout the last decade due to climate change and other human activities. Classification and mapping of benthic floors and associated ecosystems such as coral reefs are both inefficient and expensive using traditional ground-based methods. New technologies using publicly available and commercial satellite imageries are crucial for accurate classification and mapping of coral reefs' distribution, management and monitoring. The study utilized the medium (Sentinel 2B with 20 m) and high (SPOT 7 with 1.5m) resolution satellite imageries for benthic mapping of Mabul island’s benthic using pixel-based and object-based classification methods. Results of the study show that the overall accuracy of the pixel-based classification method for Sentinel 2 and SPOT 7 were 97.5% and 90%, respectively. For the object-based technique, the overall classification was slightly lower with 87.05% and 82.81%, respectively. This study suggests pixel-based classification provides better overall accuracy than object-based classification. However, conducting more assessments at different water depths and field surveys is necessary to determine accurate results. This can be achieved in the future by using more advanced technology such as drones and lidar data.