Assessing the Accuracy of Histogram Threshold and Band Ratio as a Shoreline Extraction Techniques from Low-resolution Sentinel-2 Satellite Imagery.

Abstract: Coastal regions worldwide have a significant role for humanity in various aspects. Maintaining coastal zones vitally supports the local and global economy concerning fisheries, tourism, transportation, and human activities. Rapid climate changes due to wind, global warming, sea level rise, and human activities such as construction have negative impacts, such as coastal erosion, which has caused dramatic changes in shorelines. Therefore, adapting coastal management and environmental monitoring will assist in mitigating the adverse effects. To ensure sustainability development in planning and disaster response, it is essential to map and classify shoreline changes. Despite advances in remote sensing satellites, coastal zone management suffered from low resolution in both spatial and temporal aspects from open-source platforms such as MODIS, Landsat, and Sentinel. Moreover, high-resolution drone images are still limited, in addition to many challenges and restrictions. This paper aims to evaluate the accuracy of shoreline extraction techniques such as histogram threshold and band ratio from the low-resolution Sentinel-2 satellite imagery using shoreline delivered from high-resolution drone images as a ground truth. Analysis of Variance (ANOVA) has been applied to Testing the differences in deviations between various techniques. In addition to Root Mean Square Error (RMSE) to measure the discrepancies between the derived and reference shorelines. As a result of the evaluation of both histogram threshold and band ratio as a shoreline extraction method from satellite regarding the reference shoreline extracted from the drone, the main finding showed that the threshold is better than band ratio shoreline extraction technique from low-resolution sentinel-2 satellite imagery.