Optimization of object — based image segmentation in classifying water region

One of the key in classification is the segmentation. Segmenting an image into regions is a problem that has many possible solutions. Object - based segmentation has been very popular in recent years because of its extraordinary ability to separate the flexibility and homogeneity with regards to shape and color from its neighboring pixel cells specifically to the data set with high spatial variability. However, the most critical is the choice of parameter values. This study aims to optimize the segmentation by choosing appropriate filtration method and parameter values. However, to determine the performance of the segmentation process, this study performed Area Fit Index metric using ArcPy package. There are five study area selected with different condition and coverage. Results show that larger areas provide highest accuracy in AFI evaluation. However, this is contrast to the classification results which provides higher accuracy to the smaller dataset.