Accuracy assessment of several classification algorithms with and without hue saturation intensity input features on object analyses on benthic habitat mapping in the Pajenekang Island Waters, South Sulawesi

Abstract

Object-based image analysis (OBIA) method has been proven to improve the accuracy value on benthic habitat mapping. The purpose of this study was to assess the accuracy of several classification algorithms on benthic habitat mapping based on OBIA method with and without input feature of Hue Saturation Intensity (HSI) in the Pajenekang island waters, South Sulawesi, Indonesia. Sentinel-2A satellite imagery with 10 m2 spatial resolution acquired on 3 September 2018 was used in this study. During OBIA analyses, we segmented the object into 5, 10, and 15 classes and treated each of them with input features of mean+ratio vs mean+ratio+hue saturation intensity (HSI). We later classified the benthic habitat by applying several classification algorithms such as the Bayes, K-Nearest Neighbour (KNN), Support Vector Machine (SVM), and Decision Tree (DT). The results showed that the Bayes algorithm produced highest accuracy of 78.35% within 10 segmentation classes and input features of mean+ratio+HSI followed by the KNN of 71.13% with 5 segmentation classes and input features of mean+ration+HSI. The addition of HSI input features into OBIA analyses increased the accuracy of benthic habitat classification mapping of 4.13% with the Bayes classification algorithm.