Abstract

ABSTRACTOwing to difficulties accessing areas where tidal creeks develop and the frequent morphological changes they undergo, there are very significant advantages in using remote sensing methods to effectively monitor conditions and the dynamics in these zones. The Yellow River Delta, which is the research area of this study, exhibits a very irregular tidal flat, with tidal creeks that vary in width and experience water current anisotropy. This study proposes a method for extracting tidal creeks based on a digital image processing technology that can carry out accurate characterization of tidal creeks against heterogeneous backgrounds. First, the normalized difference water index (NDWI) and Maximum Between-Class Variance (Otsu) classification are used to delineate the wide tidal creeks. Second, the modified fuzzy c-means clustering algorithm (MFCM) is used to suppress the difference between the target and background caused by spatial heterogeneity of the tidal flat environment. Next, the Gaussian matching filter (GMF) is used to enhance the narrow tidal creeks and suppress the influence of anisotropy. Then, adaptive threshold segmentation is conducted to delineate the narrow tidal creeks. Finally, the complete tidal creek networks are delineated by combining the wide and narrow tidal creeks. The overall performance was evaluated by visual interpretation and the result is excellent. In the four test areas, the kappa coefficient (κ) was greater than 0.79, the error of Commission was less than 22%, and the error of omission was less than 15%; the proposed method performed better than the Maximum Likelihood method, Support Vector Machine, and k-means algorithms. The results show that the proposed method can differentiate tidal creeks from different research areas, with good extraction accuracy and stability. The method can provide reference data for real-time dynamic monitoring of tidal creeks and has potential for application to other tidal zones in different regions.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call