Automated detection of coastline using Landsat TM based on water index and edge detection methods

This paper presents a comprehensive and automated approach to detect coastline accurately and efficiently, which integrates MNDWI (Modified Normalized Difference Water Index), Otsu algorithm and canny edge detection. Landsat TM data was employed in Qinhuangdao coastal zone as a case study. First, by analyzing the ground truth data, the coasts in Qinhuangdao were classified into four classes: sandy coast, bedrock coast, artificial coast and muddy coast, and then detection result of coastline was achieved on the basis of these types. Finally, the coastlines detected in September 2006 and September 2011, were compared together. Some changes caused by anthropogenic activities were monitored. The overall accuracy of coastline detection, verified by using GPS data was 92%.