An OBIA and Rule Algorithm for Coastline Extraction from High- and Medium-Resolution Multispectral Remote Sensing Images

Abstract

This paper aims to develop and test a method that enable to semi-automatically extract coastline independent of the beach type and study location. The proposed technique is based on the object-oriented approach of image feature extraction and takes into account the spectral as well as spatial variability of land features. The method combines object-based image analysis (OBIA) technique and spectral attribute information for the generation of classification rules for landside-seaside feature separation through satellite images. To evaluate the efficacy of the proposed approach and the performance of the developed rules, two different study locations with totally different coastal geomorphic features (coastal plains, bay beaches, rocky cliffs, etc.) have been used to extract coastline. The method is tested with two different sensor-driven images having medium (Sentinel-2) and high (Orbview-3) spatial resolution. The produced results are quantitatively evaluated by comparison with manually digitized coastline features. The distance measurements between the OBIA and manually extracted coastlines are used to measure the degree of consistency and inconsistency. The proposed method is found to be successful in the coastline extraction from both the datasets with the consistency of 95 to 99%. The higher agreement between the extracted coastlines for each type of coastal location indicates the higher precision and efficiency of the proposed workflow.